VALIDATION OF NOVEL TECHNIQUES TO EVALUATE NUTRITIONAL 
STATUS IN ADULT FEMALES IN RURAL, HIGHLAND ETHIOPIA 
 
 
 
 
 
 
 
 
A Dissertation 
Presented to the Faculty of the Graduate School 
of Cornell University 
In Partial Fulfillment of the Requirements for the Degree of 
Doctor of Philosophy 
 
 
 
 
 
by 
Jenna Marie Golan 
May 2022
 
 
 
 
 
 
 
 
 
 
 
 
 
© 2022 Jenna Marie Golan
 
 
VALIDATION OF NOVEL TECHNIQUES TO EVALUATE NUTRITIONAL 
STATUS IN ADULT FEMALES IN RURAL, HIGHLAND ETHIOPIA 
Jenna Marie Golan, Ph. D. 
Cornell University 2022 
ABSTRACT 
Addressing malnutrition in Ethiopia is complex because the prevalence 
of underweight persists as overweight/obesity increases.  A limitation in 
understanding and addressing malnutrition in adults is that body mass index 
(BMI) is frequently the only measure of nutritional status.  Therefore, 
researchers need tools to assess body composition, measure physical activity, 
and understand the functional significance of malnutrition.  However, a 
significant limitation to using these tools is that they have not been validated 
in many low- and middle-income settings.   
This research evaluated existing bioelectrical impedance analysis (BIA) 
and skinfold thickness (SFT) prediction equations that calculated fat mass 
(FM), fat-free mass (FFM), and percent body fat using air displacement 
plethysmography (ADP).  The study participants consisted of 125 females and 
129 males residing in Jimma City, Ethiopia.  The second aim evaluated the 
Global Physical Activity Questionnaire (GPAQ) and a 24-hour recall of time 
use and perceived exertion in measuring the proportion of time spent at 
moderate and vigorous physical activity (MVPA) using accelerometry.  
The third aim evaluated the sit-to-stand test (STS), usual gait speed 
(UGS), and activities of daily living (ADLs) questionnaire as measures of 
 
physical function for feasibility, reliability, and validity.  The second and third 
aims took place in rural Tigray, Ethiopia.  The study population consisted of 
females between 18 and 45 years.  
One existing BIA prediction equation was validated for adult males.  
No BIA prediction equations were valid for females, and no SFT prediction 
equations were valid for males or females.  New equations were created.  The 
GPAQ was found to have low validity.  The 24-hour recall had a fair 
agreement with accelerometry.  The agreement improved by controlling for 
BMI.   STS was a feasible, reliable, and valid measure of physical function.  
UGS lacked feasibility and reliability.  The validity of the ADL questionnaires 
was inconclusive. 
This research will provide researchers with better tools to understand 
and address malnutrition in rural highland Ethiopia.  The body composition 
equations will improve the identification of people who are malnourished.  
Quantifying physical activity and assessing physical function will enable 
researchers to understand the causes and consequences of malnutrition, 
guiding effective interventions to address malnutrition. 
 
 
 
 
BIOGRAPHICAL SKETCH 
Jenna Marie Golan was born in Chicago, Illinois.  She became interested in nutrition 
during her undergraduate studies.  She received a B.A. in anthropology from 
Northwestern University, Evanston, Illinois in 2006, a M.S. in nutrition from Columbia 
University, New York, New York in 2007, and a M.S. in epidemiology from Harvard 
T.H. Chan School of Public Health in 2010.  In 2016, she entered the graduated program 
in the division of nutritional sciences where she worked with Professor John Hoddinott.   
iii 
 
 
 
 
 
 
 
 
 
 
 
 
To my family, for always loving and supporting me while following my dreams. 
To Adam and our future. 
iv 
 
ACKNOWLEDGMENTS 
 
Many people supported me during my years are Cornell.  First and foremost, I 
would like to thank my advisor, Dr. John Hoddinott, for his mentorship, patience, and 
encouragement.  Before starting at Cornell, I overlapped with John at the International 
Food Policy Research Institute. Where he was known widely regarded not only as a 
thoughtful researcher but also a great mentor.  His previous mentees were his biggest 
advocates when deciding where I wanted to pursue my doctorate.  I am grateful for his 
thoughtful advice and guidance during my time as his student.  Additionally, I was 
provided with many opportunities to conceive and execute my own research that many 
doctoral students are not.  As a result, I have gained significant knowledge and 
experience that has helped me become a competent, independent, and passionate 
researcher. I cannot express enough thanks and appreciation.   
I would also like to thank the members of my doctoral committee: Drs. Anna 
Thalacker-Mercer, James Booth, and Kathleen Rasmussen.  My research is stronger for 
the guidance they provided.  First, I would like to thank Jim for pushing me to explore 
new and different statistical analyses to ensure sound methods.  Next, I would like to 
thank Anna for all the guidance she provided regarding physical function.  I always 
enjoyed our conversations and learned so much about diet, exercise, and muscle.  
Finally, I would like to thank Kathy for her mentorship, first as her teaching assistant 
and then as a researcher.  I asked Kathy to be on my committee because her approach to 
work and life contrasted my own.  I needed her exacting standards to become the 
researcher that I wanted to be. While I still have a significant way to go, I have made 
immeasurable progress in getting there over the last several years because of Kathy.  I 
v 
 
would like to thank Dr. Patricia Cassano, the primary investigator of the translational 
science training grant that supported me.  I would also like to that the Cornell Statistical 
Consulting Unit for their help while learning partial least squares regression. 
I would like to thank my collaborators at IFPRI, including Dr. Guush Berhane, 
Dr. Kalle Hirvonen, and Nahume Yadene.  I would like to thank my research team in 
Tigray, especially Equbay Gebrehiwet.  I would also like to thank all the kind research 
participants in Tigray who welcomed me into their home, fed me injera, and made me 
fresh coffee.  I would like to thank my collaborators at Jimma University, including Dr. 
Tefera Belachew, Getu Gizwa, and Abonesh Taye.  Thank you, Abonesh, for ensuring 
that I was fed, entertained by her children, and had internet access while in Jimma.  I 
would like to thank JUCAN for allowing me access to their Bod Pod.  This research was 
supported by several generous grants, including the NIH/NIDDK T32 Nutrition 
Training Program in Translational Science, the H.E. Babcock Fund, the Bill and Melinda 
Gates Foundation, the van Veen travel fund, the AWARE travel grant, and Cornell’s 
graduate student travel fund. 
I would like to thank my family for always supporting me even when they didn’t 
fully understand why I made the life choices I did.  Finally, I would like to thank the 
love and support of the fantastic individuals who supported me in Ithaca.  First and 
foremost, my wonderful and endlessly supportive husband.  Adam never questioned 
why I would take off for Ethiopia for weeks a time, even early in our relationship.  I 
would also like to thank my in-laws, especially Sonia, Parker, Binna, and Monica, for 
their friendship.  Additionally, I would like to thank my lab mates, Gargi Wable and 
Fiona Coleman, and my friends Amy Fothergill, Josh Nassif, Jocelyn Boiteau, Emily 
Riddle, and Laura Pompano.  You made my time at Cornell so special.  
vi 
 
 
 
TABLE OF CONTENTS 
ABSTRACT .................................................................................................................................. iii 
BIOGRAPHICAL SKETCH ....................................................................................................... iii 
ACKNOWLEDGMENTS ............................................................................................................ v 
LIST OF FIGURES ....................................................................................................................... xi 
LIST OF TABLES ....................................................................................................................... xii 
LIST OF ABBREVIATIONS ..................................................................................................... xiii 
CHAPTER 1 ................................................................................................................................... 1 
INTRODUCTION AND LITERATURE REVIEW ............................................................................. 1 
Background ............................................................................................................................................. 1 
Malnutrition ............................................................................................................................................ 2 
Body composition ............................................................................................................................................ 11 
Physical activity ............................................................................................................................................... 12 
Physical function .............................................................................................................................................. 15 
REFERENCES ....................................................................................................................................... 18 
CHAPTER 2 ................................................................................................................................. 29 
VALIDATION OF SKINFOLD THICKNESS AND BIOELECTRICAL IMPEDANCE 
EQUATIONS IN ETHIOPIAN ADULTS ................................................................................ 29 
Abstract .................................................................................................................................................. 29 
Introduction ........................................................................................................................................... 30 
vii 
 
Materials and methods ........................................................................................................................ 32 
Study participants ............................................................................................................................................ 32 
Air-displacement plethysmography ............................................................................................................. 33 
Anthropometric measures .............................................................................................................................. 33 
Bioelectrical impedance analysis ................................................................................................................... 34 
Skinfold thickness ............................................................................................................................................ 34 
Existing equations ............................................................................................................................................ 35 
Statistical analysis ............................................................................................................................................ 35 
Results .................................................................................................................................................... 38 
Discussion .............................................................................................................................................. 53 
REFERENCES ....................................................................................................................................... 56 
CHAPTER 3 ................................................................................................................................. 64 
VALIDITY OF THE GLOBAL PHYSICAL ACTIVITY QUESTIONAIRE (GPAQ) AND 
24-HOUR RECALL OF TIME USE AND PERCEIVED EXERTION IN FEMALES IN 
RURAL TIGRAY, ETHIOPIA ................................................................................................... 64 
Abstract .................................................................................................................................................. 64 
Introduction ........................................................................................................................................... 66 
Material and Methods .......................................................................................................................... 68 
Participants and study setting ....................................................................................................................... 68 
Sample size ........................................................................................................................................................ 68 
Data collection .................................................................................................................................................. 69 
Physical activity ............................................................................................................................................... 71 
Data processing ................................................................................................................................................ 73 
GPAQ analysis ................................................................................................................................................. 74 
24-hour recall of perceived exertion analysis .............................................................................................. 76 
viii 
 
Ethical approval ............................................................................................................................................... 77 
Results .................................................................................................................................................... 77 
Study population ............................................................................................................................................. 78 
GPAQ ................................................................................................................................................................. 80 
24-hour recall .................................................................................................................................................... 86 
Discussion .............................................................................................................................................. 90 
REFERENCES ....................................................................................................................................... 94 
CHAPTER 4 ............................................................................................................................... 102 
RELIABILITY AND VALIDITY OF PHYSICAL FUNCTION TESTS AND ADL SURVEY 
QUESTIONS IN FEMALES LIVING IN RURAL HIGHLAND ETHIOPIA .................... 102 
Abstract ................................................................................................................................................ 102 
Introduction ......................................................................................................................................... 103 
Materials and methods ...................................................................................................................... 105 
Study setting ................................................................................................................................................... 105 
Study subjects ................................................................................................................................................. 106 
Data collection ................................................................................................................................................ 107 
Physical function instruments ...................................................................................................................... 108 
Feasibility, reliability, and validity .............................................................................................................. 110 
Results .................................................................................................................................................. 112 
Discussion ............................................................................................................................................ 118 
REFERENCES ..................................................................................................................................... 122 
CHAPTER 5 ............................................................................................................................... 128 
CONCLUSION ......................................................................................................................... 128 
ix 
 
Summary of results ........................................................................................................................................ 128 
Contribution to research and future directions ......................................................................................... 132 
APPENDIX 1: Supplementary body composition tables .................................................... 139 
APPENDIX 2: Aim 2 sensitivity analyses ............................................................................. 142 
APPENDIX 3: Physical function ............................................................................................. 145 
 
x 
 
LIST OF FIGURES 
 
Figure 1. 1 Maps of Ethiopia ...................................................................................... 5 
 
Figure 2. 1: Bland-Altman plots of ADP and BIA ................................................. 42 
Figure 2. 2: Bland-Altman plots of ADP and SFT ................................................. 47 
Figure 2. 3: Bland-Altman plots of ADP and new equations .............................. 52 
 
Figure 3. 1: Scatter plot comparing the proportion of time spent at MVPA ..... 89 
xi 
 
LIST OF TABLES 
Table 2. 1: Characteristics of participants ............................................................... 39 
Table 2. 2: Evaluations of the validity of existing BIA equations ....................... 40 
Table 2. 3: Evaluation of the validity of existing SFT equations ......................... 46 
Table 2. 4: Evaluation of new BIA and SFT equations ......................................... 51 
 
Table 3. 1: Summary of study participants' characteristics ................................. 79 
Table 3. 2: Reliability of the GPAQ for dichotomous responses ......................... 81 
Table 3. 3: Reliability of the GPAQ for continuous responses – Spearman’s r 82 
Table 3. 4: Validity of GPAQ .................................................................................... 83 
Table 3. 5: GPAQ Discordancy ................................................................................ 85 
Table 3. 6: Validity of 24-hour recall ....................................................................... 87 
Table 3. 7: Calibration of the 24-hour recall model ............................................... 88 
 
Table 4. 1: Study participants characteristics ....................................................... 112 
Table 4. 2: Reliability of physical function tests .................................................. 114 
Table 4. 3: Correlation of ADL questions ............................................................. 115 
Table 4. 4: Validity of ADLs in relationship to Psts .............................................. 116 
Table 4. 5: Validity of ADLs in relationship to Psts .............................................. 118 
xii 
 
  LIST OF ABBREVIATIONS 
 
Abbreviations: 
ADL – activities of daily living 
ADP – air displacement plethysmography 
BIA – bioelectrical impedance analysis 
BIC - Bayesian information criterion 
BD – body density 
BMI – body mass index 
BMR – basal metabolic rate 
DHS – Demographic and Health Survey 
DXA – dual-energy x-ray absorptiometry 
FFM- fat-free mass 
FM- fat mass 
GPAQ – Global Physical Activity Questionnaire 
JUCAN - Jimma University Clinical and Nutrition Research Center  
KS – Kolmogorov Smirnov  
LMIC – low- and -middle-income countries 
MDD-W – Minimum dietary diversity in women 
METS – metabolic equivalents (Chapter 3) 
METS - Modeling Epidemiologic Transition Study (Chapter 2) 
MUAC – Mid-upper arm circumference 
MVPA – moderate or vigorous physical activity 
PEE – physical activity-related energy expenditure 
xiii 
 
PLS - partial least-squares 
PSNP – Productive Safety Net Program 
Psts – Power sit-to-stand 
SEE – standard error of estimates 
SFT – skinfold thickness 
SSN – social safety nets 
STS – sit-to-stand 
TE – total error 
TEF – thermic effect of food 
UGS – usual gait speed
xiv 
 
CHAPTER 1 
INTRODUCTION AND LITERATURE REVIEW 
Background  
 Like many low- and middle-income countries, addressing malnutrition in 
Ethiopia is complex.  Undernutrition and its causes persist, while the prevalence of 
overweight and obesity is increasing, this is known as the double burden of 
malnutrition [1,2].  One tool to address malnutrition in Ethiopia is a national safety net 
program, the Productive Safety Net Programme (PSNP), which addresses poverty, a 
basic determinant of malnutrition.  The PSNP began in 2005 and has been rigorously 
evaluated.  Despite its impacts on the underlying causes of malnutrition, including 
income and food security, the PSNP has not had the expected impact on females’ 
nutritional status [3].  However, nutritional status is only measured using body mass 
index (BMI), which previous research has demonstrated to be problematic in classifying 
Ethiopians as malnourished [4,5].  The limited impact of the PSNP and other 
interventions to improve nutritional status motivated this research.  Additionally, there 
are concerns that addressing undernutrition by solely focusing on increasing body mass 
could increase the prevalence of obesity [6].  To effectively combat the double burden of 
malnutrition in Ethiopia, nutrition research needs more breadth and depth. 
Limiting the definition of nutritional status to BMI is insufficient [7] to address 
malnutrition, given the scale of the problem and the far-reaching impact on a person’s 
health and well-being [8].  Better tools exist to evaluate nutritional status, including a 
two-compartment body composition model [9].  Additionally, the evaluations of the 
PSNP have not included other causes and consequences of malnutrition, including 
physical activity and physical function.  Previously, it was not possible to measure body 
1 
 
composition, physical activity, and physical function in Ethiopian adults because the 
tools had not been validated in rural highland Ethiopia.  This dissertation validated 
body composition prediction equations, a survey tool to quantify physical activity, and 
a direct measure of functional status.  The inclusion of these tools in future research on 
malnutrition in Ethiopia, including the evaluation of the PSNP, will provide vital 
insight into the drivers of malnutrition, allowing for the development of effective 
interventions and programs. 
Malnutrition 
Undernutrition and obesity are two sides of malnutrition, referred to as the 
double burden of malnutrition, with interconnected common determinants, including 
early life nutrition, diet quality, food environments, and socioeconomic factors [10].  
However, for decades public health and development efforts took a siloed approach 
that tackled undernutrition and obesity separately.  The perspective was that 
undernutrition is linked to poverty, food insecurity, and infection and was considered a 
problem of low- and middle-income countries (LMIC).  It encompasses both 
underweight and micronutrient deficiencies.  Obesity is linked with dietary excess and 
sedentary behavior.  Historically, it was viewed as a problem in high-income countries 
[2].  However, as obesogenic environments expand and the causes of undernutrition 
persist, these two nutritional problems can co-occur within communities, households, 
and individuals.  This co-occurrence results in the double burden of malnutrition.  Both 
overnutrition and undernutrition have adverse health impacts throughout the lifespan, 
particularly in females of reproductive age and their children. 
Underweight females of reproductive age and their children are at greater risk of 
adverse health outcomes.  The World Health Organization (WHO) defines underweight 
2 
 
as having a body mass index (BMI) less than or equal to 18.5 kg/ 2m , overweight as 
having a BMI greater than 25 kg/ 2m , and obesity as having a BMI greater than 30 kg/ 2m  
[11].  Females of reproductive age who are underweight have an increased risk of death 
[12] and illness, especially infections of the respiratory tract [13–15], compared to 
normal-weight females (18.5 kg/ 2m < BMI £ 25 kg/ 2m ). In addition, the nutritional demands 
of pregnancy and lactation can make females more nutritionally vulnerable [16,17].  
Being born to an underweight mother can have lifelong effects.  A low pre-pregnancy 
BMI and undernutrition during pregnancy are associated with fetal growth restriction, 
responsible for 12% of child deaths [8].  Additionally, fetal growth restriction leads to 
stunting in childhood [18] and developmental deficits [19–21], which negatively affects 
schooling and cognitive performance [22–24].  In adulthood, it can lead to short stature, 
obesity, non-communicable diseases [8,25], depression [26], and decreased economic 
productivity [27,28].  This creates an intergenerational cycle of poverty and 
malnutrition, which is difficult to disrupt [2].  This research focused on energy-related 
malnutrition.  Micronutrient deficiencies are beyond its scope. 
Obesity in adults is associated with all-cause mortality, diabetes mellitus, 
hypertension, coronary heart disease [11,29], many types of cancer [30], and poorer 
mental health [31,32].  Additionally, obesity is associated with decreased wages, 
economic productivity [33–35], and increased health-related expenses [36,37].  During 
pregnancy, females who are obese may be at increased risk for pregnancy-associated 
hypertension, gestational diabetes, complications during labor and delivery, 
postpartum weight retention, and subsequent obesity [38].   Adverse birth outcomes 
associated with high pre-pregnancy BMI and weight gain during pregnancy include 
macrosomia [39], preterm birth [40,41], and stillbirth [42].  Additionally, females that are 
3 
 
obese have difficulty initiating breastfeeding [38].  Exclusive and appropriate 
breastfeeding is protective again all forms of malnutrition in the child and obesity in the 
mothers [2].  
Worldwide, 40.8% of females are overweight or obese, and 9.1% are underweight 
[43].  While the double burden of malnutrition exists in almost every country, it is a 
more significant problem in LMICs in Asia and Africa than in other regions because the 
prevalence of underweight people is higher [8].  These countries are also experiencing a 
rapid increase in the prevalence of overweight and obesity.  An example is Ethiopia, 
which has a high prevalence of underweight females and an increasing prevalence of 
overweight and obesity.  According to the 2016 Ethiopia Demographic and Health 
Survey (DHS), the prevalence of underweight among females between the ages of 15 
and 49 years is 22% [1].  The prevalence of underweight among females was higher in 
rural areas (25%) than urban areas (15%) [1].   
Meanwhile, overweight and obesity in Ethiopia rose from 3% in 2000 to 8% in 2016.  In 
contrast to underweight, overweight and obesity in females between the ages of 15 and 
49 years were higher in urban areas (21%) than rural areas (4%) [1].  Development and 
urbanization can lead to more sedentary lifestyles and inexpensive calories, increasing 
overweight and obesity. 
Ethiopia 
Ethiopia is a landlocked, low-income country in East Africa (Figure 1.1) with a 
high prevalence of underweight in reproductive-age females with a national public 
works SSN program. According to the 2016 DHS, the wealthiest households are 
4 
 
Figure 1. 1 Maps of Ethiopia 
The map of the left shows the location of Ethiopia in East 
Africa.  The map of the right is a map of Ethiopia that points 
out the location of the capital city of Addis Ababa, Jimma City 
where aim 1 took place, and the Tigray region where aim 2 and 
3 took place. 
concentrated in the urban areas; 89% of urban households compared to 7% of rural 
households were in the wealthiest quintile.  Additionally, 46% of rural households were  
in the lowest two wealth quintiles [1].  The disparities between urban and rural 
households are can also be seen in nutritional status, education attainment, and housing 
quality. 
Ethiopia has a high prevalence of underweight in females of reproductive age 
(18-49 years old), stunting in children under age five years, and an increasing rate of 
overweight and obesity.  In 2016, 22% of females aged 15-49 were underweight, which 
decreased from 30% in 2000 [1]. In addition, only 2% of females were of short stature 
(below 145 cm).  In 2016, stunting in children under age five remained high at 38%, 
despite decreasing from 58% in 2000 [1]. 
5 
 
Undernutrition in Ethiopia is caused by many factors, including food security 
which is significantly impacted by seasonal variations in rainfall and income.  Ethiopia 
has mountainous highland areas in the center and north of the country and flat lowland 
areas east and south.  Within the highland regions of the country, food production and 
livelihoods depend on rain-fed smallholder agriculture. A short rainy season (belg) 
occurs between March and May, and the main rainy season (meher) occurs between 
June and October.  During the main rainy season, more than 90 percent of the total 
crops in the country were produced.  Crops were harvested between October and 
December [44].  Between December and February, the harvest was stored or sold at 
local food markets [3].   
The seasonality of crop production resulted in variations in local food availability 
and prices [45,46].  Belg was the post-harvest season characterized by drier conditions 
but better food availability and higher incomes and savings due to the recent harvest. 
On the other hand, meher was the lean season, and households were busy with 
agricultural activities. Twenty-eight percent of females in rural areas were employed, 
with 55% in agricultural work.  Additionally, 67% of female agricultural workers were 
employed seasonally. Despite the increased energy demands associated with 
agricultural activities, household energy intake fell due to lower incomes [3].   
Throughout the year, dietary diversity in females and children was low.  A study 
that looked at females in rural highland regions of Ethiopia found that less than 6% of 
females met their minimum dietary diversity indicating they had consumed at least five 
out of ten foods groups the previous day [3,47].  According to the 2016 DHS, only 14% 
of children met the minimum dietary diversity score, which required consuming foods 
from at least four out of seven different foods groups the previous day.  More urban 
children (19%) met their minimum dietary diversity compared to rural children (6%) 
6 
 
[1,48].  During the main rainy season, dietary diversity improved in rural regions as 
more affordable vegetables were available [45]. 
In addition to food security and dietary diversity, there was seasonal variation in 
infection diseases, including malaria [49] and tuberculosis [50], that corresponded to 
increased rainfall during the main rainy season [51].  Seasonal changes in food 
availability, energy exertion, and infection rates increased energy demands during the 
main rainy season, which is likely a driver of the increased prevalence of underweight 
[52–56].  These seasonal variations in food and infection also cause considerable 
fluctuations in acute malnutrition rates in children across seasons [57].   
In Ethiopia, religion plays a vital role in shaping diets, especially for the 
Orthodox Christians, which is the main religion (43% of females and 45% of men) [1].  
The Ethiopian Tewahedo Orthodox Church has several fasting periods throughout the 
year.  While fasting, observers must refrain from consuming all animal-sourced foods 
and from eating until midday [58,59].  With 55 days, Lent, which typically occurs 
between February and April, is the most prolonged fasting period.  Other significant 
fasting periods occur in August (16 days) and before Ethiopian Christmas in December 
and January (40 days).  While very young children and pregnant females are exempt 
from fasting, previous research indicates that fasting reduced the availability of animal 
source foods in households and local food markets [60]. 
Educational attainment is associated with undernutrition in females [61].  
According to the 201 DHS, females with no education decreased from 66% in 2005 to 
48% in 2016.  Similarly, the percentage of females with no education increased by age 
group; only 14% of females aged 15-19 years had no education, while 79% of females 
aged 45-49 years had no education.  Indicating there was an improvement in females’ 
education over time.  Education was better in urban areas than rural areas; 57% of rural 
7 
 
females compared to 16% of urban females had no formal education. The urban-rural 
difference in educational attainment was more significant as years of schooling 
increased.  Only 1% of rural areas compared to 21% of urban females had more than 
secondary education.  Educational attainment also varied by wealth quintile. Seventy-
four percent of females in the lowest wealth quintile compared to 19% in the highest 
quintile had no education. Similarly, less than 1% of females in the lowest wealth 
quintile had more than secondary education, compared with 18% of those in the highest 
quintile.   
The average household size in Ethiopia was 4.6 persons.  Urban households were 
slightly smaller than rural households (3.5 versus 4.9 persons).  Males were the head of 
most households; a female-headed only 1 in 4 households [1].  Only 8% of rural 
households had access to electricity, compared to 93% of urban households.  Ninety-
seven percent of urban households had access to an improved source of drinking water, 
meaning it is piped into their or their neighbor’s dwelling or plot, or they had access to 
a public tap, borehole, or protected well. 
Meanwhile, only 57% of rural households had access to an improved water 
source.  More than half of rural households travel 30 minutes or longer to fetch drinking 
water and return.  This chore was most likely the responsibility of adult females (68% of 
rural households).  In rural areas, females under 15 years were three times more likely 
than their male counterparts to fetch water (13% versus 4%) [1].  Only 16% of urban and 
4% of rural households had access to improved toilet facilities which include non-
shared flush/pour flush toilets piped to sewer systems, septic tanks, or pit latrines; 
ventilated improved pit latrines; pit latrines with slabs; or composting toilets.  More 
than half, 56%, had access to unimproved facilities in rural areas, and 39% had no access 
to facilities [1].   
8 
 
PSNP 
Poverty is an underlying cause of food insecurity and subsequent malnutrition.  
Globally, governments address poverty, food insecurity, and poor nutritional outcomes 
through social assistance and social safety net (SSN) programs.  These programs help 
individuals and households cope with chronic poverty and vulnerability [62].  
Examples of these programs include the unconditional transfer of cash and goods, 
conditional cash transfers which reward actions such as enrolling children in school, 
food and in-kind transfers, school feeding programs, and fee waivers and targeted 
subsidies.  Sixty-seven percent of countries have public works programs that provide 
cash or food payments for employment on public projects that reverse environmental 
degradation, improve water control, and increase road access.  These projects build 
community assets while protecting household assets.  Millions of people participate in 
public works programs in countries including Afghanistan, Argentina, Ethiopia, India, 
Malawi, Mexico, Nepal, and Niger [62,63]. 
The Productive Safety Net Programme (PSNP) is the Government of Ethiopia’s 
principal social protection program beginning in 2005.  The PSNP focused on reducing 
food insecurity and household asset depletion by using labor during the lean season to 
build essential community assets and address environmental degradation.  The PSNP 
provides six months of employment on public works projects to its beneficiaries, 
though a small number of individuals receive unconditional cash transfers.  There are 
no national criteria to determine who qualifies for benefits; the local governments 
determine who qualifies.  Depending on location, payments are cash or food staples, 
predominantly grain.  The PSNP provides benefits to approximately 8 million 
Ethiopians which is 10% of the population, making it one of sub-Saharan Africa’s most 
extensive social protection programs.   
9 
 
The PSNP has been rigorously evaluated since 2006, demonstrating that the 
program has successfully improved its economic outcomes.  It improved household 
food insecurity, wealth, and economic resiliency [64–67].  The PSNP had spillover 
benefits into local economies [68].  The economic outcomes should have improved the 
nutritional status of females and children [25].  However, the program did not have the 
expected effect on stunting rates among children and underweight among their mothers 
[3,67].   
The evaluation of the PSNP only includes two indicators of nutritional status for 
adults, BMI and mid-upper arm circumference (MUAC) for females of reproductive 
age.  The outcomes of interest include the prevalence of stunting in children and the 
proportion of females classified as underweight using the WHO BMI cutoffs [11].  A 
significant limitation to understanding and addressing malnutrition in adults is that 
BMI only measures nutritional status.  The WHO BMI cutoffs are an inadequate 
indicator of malnutrition in females living in rural highland Ethiopia [5,7,69].  A better 
definition is needed to identify females that are malnourished.  
Like other evaluations of nutrition-specific and nutrition-sensitive programs, the 
evaluation of the PSNP did not include measurements of several of the causes and 
consequences of malnutrition, including physical activity and physical function.  
Therefore, it is difficult to understand why the program did not improve 
undernutrition.  Innovative tools are needed to characterize the impact of the PSNP on 
intermediary causes of malnutrition, which may prevent the program from having its 
expected impact.   
10 
 
Body composition 
BMI as an indicator of nutritional status has remained relatively unchanged since 
the 1850s [70], despite increased knowledge about nutrition and health.  A criticism of 
BMI [71] is that it frequently misclassifies excess adiposity, overestimating the rates of 
obesity-associated health risks [72,73].  One reason is that BMI is dependent upon a 
constant relationship between stature and body mass, and it assumes a constant 
relationship with adiposity regardless of age, physical activity, and genetics [74], 
resulting in obesity misclassification varying with age and across ethnic groups [75]. 
However, accounting for ethnicity was insufficient to correct this, and obesity 
misclassifications still occur within genetically similar populations [7].  
Attempts to remedy this have resulted in Asian population-specific cutoffs [76].  
While the same issues have been identified in African populations, the BMI cutoffs have 
not changed [7]. For example, a study in Ethiopia found that adults experienced many 
obesity-related adverse health outcomes at a significantly lower BMI [4]. Therefore, 
Sinaga et al. concluded that the optimal cutoff for obesity using BMI in adult Ethiopians 
was 22.2 kg/ 2 m for males and 24.5 kg/ 2m  for females.  These cutoffs may be adequate for an 
urban dwelling population like the study sample who resided in Jimma City and were 
employed at Jimma University.  However, they are not generalizable to other Ethiopian 
adults, including those residing in rural regions and engaged in physically demanding 
labor for income-generating activities.  The fat-free mass in physically active 
populations is lower than in sedentary populations, even when their body mass is the 
same [77].  Therefore, using Ethiopian-specific BMI cutoffs is still inadequate. 
Despite the criticisms of BMI, it is still used globally to define underweight, 
normal weight, overweight, and obesity, especially in resource-constrained settings.  Its 
usage has been justified at a population level because its components, height and 
11 
 
weight, are easy to measure and do not require specialized equipment [2].  A better 
method to assess nutritional status would be a two-compartment body composition 
model that measures FM, FFM, and percent body fat.  Both insufficient and excess body 
fat are associated with adverse health.  Adipose tissue plays essential roles in the 
storage of energy [78], endocrinology [79], and immune function [80,81]. The quantity 
and location of excessive adipose are associated with mortality [82] and morbidity 
[7,83–85].  Including body composition into research and programs addressing 
malnutrition in LMIC would better identify undernourished females and those with 
excess adipose tissue than BMI.   
Several methods can be used to measure FM and FFM.  However, the reference 
measures, such as dual-energy x-ray absorptiometry (DXA) and air displacement 
plethysmography (ADP), require access to laboratories and highly trained personnel.  
Other methods, including bioelectrical impedance (BIA) and skinfold thickness (SFT), 
are suitable for use in resource-limited settings, including rural, highland regions of 
Ethiopia.  However, these two methods are dependent on prediction equations, and the 
existing prediction equations have not been validated in Ethiopian adults. Therefore, 
the first aim of this research was to validate existing prediction equations that use BIA 
and SFT.  New, gender-specific equations were created and validated when existing 
equations were not valid. 
Physical activity 
Energy intake and expenditure are the energy balance equation's two 
components determining body mass and subsequent malnutrition. Immediate 
determinants of nutritional status in adults include adequate dietary intake and access 
to health care [8,86]. However, the guidelines regarding adequate dietary intake are 
12 
 
vague and usually just recommend an adequate and diverse diet.  Existing research has 
determined what diet classifies a diet as diverse.  Several survey questionnaires have 
been developed to assess dietary diversity among adult females, such as the Minimum 
Dietary Diversity for Women (MDD-W) [47].  The precise definition and assessment of a 
diverse diet enabled the creation of interventions that focus on improving diet diversity 
[25]. Unfortunately, the same clear guidance does not exist for adequate intake in 
resource-limited settings, meaning that it is impossible to determine whose diet is 
inadequate and what interventions effectively improve dietary intake.    
Determining an adequate intake is dependent upon understanding the caloric 
needs of a population.  The energy balance equation that determines body mass 
includes energy intake and expenditure. Therefore, understanding adequate dietary 
intake requires an understanding of energy expenditure. Unfortunately, energy 
expenditure can only be accurately measured using indirect or direct calorimetry.  Like 
DXA and ADP, indirect and direct calorimetry depend upon access to a laboratory, 
expensive equipment, and highly trained personnel [87], making it unsuitable for free-
living adults in resource-limited environments.   
However, energy expenditure has three components, basal metabolic rate (BMR), 
thermic effect of food (TEF), and physical activity-related energy expenditure (PEE), 
which can be divided into exercise-related thermogenesis, and non-exercise activity 
thermogenesis [88].  BMR accounts for the most significant proportion of energy 
expenditure in most people and TEF the least.  However, unlike BMR, PEE is 
modifiable and dependent on a person’s movement and activity.  PEE changes based on 
the amount of time spent at different exertion levels.  Exertion can be divided into four 
levels, sedentary, light, moderate, and vigorous.  Physical activity that requires 
moderate or vigorous exertion can significantly increase energy expenditure.  However, 
13 
 
unlike dietary intake, there are no validated survey tools to estimate physical activity in 
populations in LMIC, including rural Ethiopia [89]. Therefore, the second aim of this 
study was to validate survey tools to quantify the proportion of time spent in MVPA 
using triaxial accelerometry. 
Physical activity was self-reported using the Global Physical Activity 
Questionnaire (GPAQ) and a 24-hour recall of time use and perceived energy exertion.  
Previous efforts to validate the GPAQ in Ethiopia were unsuccessful; however, the 
study used pedometers, which only measure steps [89].  This research will evaluate the 
GPAQ using triaxial accelerometers.  An accelerometer improves over a pedometer by 
measuring movement in three planes, forward and back, side to side, and up and down 
[90].  However, the GPAQ still requires that participants adequately recall their average 
physical activity over an entire week.  This may be difficult because of the long recall 
period or weekly variations in physical activity. In addition, due to the dependence on 
weather to complete agricultural-related activities, there may not be a typical week of 
physical activity for a participant in rural Ethiopia. Therefore, rather than characterize 
physical activity or an average week, a better method may be to have more data points 
that recall what participants did over the previous 24 hours, similar to the 24-hour 
dietary recall [91,92]. 
Combining the theory behind the 24-hour dietary recalls and time use survey in 
the Women’s Empowerment in Agriculture Index [93], the research team developed the 
24-hour recall of time use and perceived exertion was developed for this study.  It 
provided valuable details on how females spend their time and which activities 
required greater energy exertion, which is not captured by the GPAQ [94].  Participants 
reconstructed their day in 30-minute periods, reporting primary and secondary 
activities, their perceived exertion, and whether they were concurrently engaged in 
14 
 
childcare.  For this research, the proportion of a person's time at MVPA was quantified.  
Quantifying the frequency and duration of MVPA was the first step in determining 
who had increased energy needs and thus determining adequate dietary intake.  
Physical function 
 The functional significance of being underweight in females in LMIC is 
unknown.  Physical function is a person’s ability to physically fulfill the tasks necessary 
for their daily lives, expected roles, and general well-being.  Together, these tasks can be 
classified as activities of daily living (ADLs) [95].   Physical function reflects physical 
activity and nutritional intake [96–98].  Intrinsically, physical function is essential for 
overall health and quality of life [99,100].  Instrumentally, physical function can identify 
people at risk for further declines in health [101–103] and increased future health 
expenditures [104].  Outside of health measures, physical function impacts social 
networks [105,106] and income-generating activities [107–109].  Extended periods of 
vigorous activity without appropriate nutritional intake are associated with poor 
physical function [15,110,111]. Low physical function is one mechanism through which 
underweight and obesity may negatively affect well-being and economic productivity, 
especially when income depends on physical labor. 
Decrease physical function lowers females’ ability to engage in labor-intensive 
income-generating activities and complete their household duties in rural highland 
Ethiopia.  Females are frequently the caretakers in this location. Their expected roles 
include caring for young children, purchasing and preparing food for the household, 
fetching and sanitizing water, and keeping a household free of potential vectors of 
diseases, including human and animal feces [3].  Diseases are an immediate cause of 
poor nutritional status.  A hygienic household environment can prevent many diseases 
15 
 
[112,113], while adequate access to health care will limit the long-term impact of disease 
on nutritional and health status [114,115].   The inability of females to fulfill their roles 
in the household can adversely affect the underlying and direct causes of nutritional 
status, lowering the health status and economic productivity of all household members. 
Despite the importance of physical function, very little research has assessed it in 
females of reproductive age in LMICs.  Research on physical function has focused 
primarily on elderly and ill populations in industrialized countries [116–118].  The third 
aim of this study was to assess the feasibility, reliability, and validity of direct physical 
function tests, including the sit-to-stand test (STS), the usual gait speed test (UGS), and 
a context-specific ADL questionnaire.  
To the best of our knowledge, neither the STS nor the UGS has been used in this 
or similar populations.  The results of the STS were used to calculate the power sit-to-
stand variable (Psts).  Prior research has measured grip strength using a dynamometer 
[119,120].  This is an inadequate measure of physical function because it focuses on one 
group of muscles in the upper body.  Physical function combines muscle mass, strength, 
and power [121–123].  Muscle power is more relevant than muscle mass and muscle 
strength for many tasks of daily living [122,124], making it the preferred indicator for 
evaluating physical function.  Additionally, lower extremities’ function is more strongly 
associated with physical function than upper extremities [122]. Therefore, the Psts and 
UGS are better measures of physical function than grip strength because they measure 
power and coordination in the lower body.  ADL questionnaires are one self-reporting 
method used by researchers to evaluate physical function in populations.  However, the 
questions in most ADL questionnaires are not relevant to this population.  This study 
adapted existing questionnaires to make them appropriate for this context.  The ADL 
questionnaires were then evaluated for validity using the physical function test.   
16 
 
Research aims 
 To address malnutrition in rural highland Ethiopia, researchers need to measure 
nutritional status using body composition, quantify physical activity, and assess 
physical function.  Unfortunately, the tools to do this are either not suitable for use in 
community-based research or have not been validated for use outside of the laboratory. 
Therefore, this research aims to validate innovative tools to evaluate nutritional status 
in rural highland Ethiopia. 
The overall goal of this dissertation is to validate tools that will allow researchers 
to understand malnutrition in rural highland Ethiopia better.  Chapter 2 evaluates 
existing BIA and SFT prediction equations of body composition to improve researchers’ 
abilities to classify people as malnourished using percent body fat.  We determined that 
the METS BIA equation for males was valid and created a new BIA equation for females 
and new sex-specific equations that use SFT.  Chapter 3 evaluates the validity of the 
GPAQ and the 24-hour recall of perceived exertion.  We concluded that the GPAQ was 
not valid. However, the 24-hour recall was valid when calibrated with BMI. Finally, 
chapter 4 assessed the feasibility, reliability, and validity of STS, UGS, and the ADL 
questionnaire.  We determined that only the STS was suitable for evaluating physical 
function in rural, highland Ethiopia 
 
 
  
17 
 
REFERENCES 
1.  Central Statistical Agency (CSA) Ethiopia and ICF Ethiopia Demographic and 
Health Survey; CSA and ICF: Addis Ababa, Ethiopia and Rockville, Maryland, 
USA, 2016; 
2.  Wells, J.C.; Sawaya, A.L.; Wibaek, R.; Mwangome, M.; Poullas, M.S.; Yajnik, C.S.; 
Demaio, A. The Double Burden of Malnutrition: Aetiological Pathways and 
Consequences for Health. Lancet 2020, 395, 75–88, doi:10.1016/S0140-
6736(19)32472-9. 
3.  Berhane, G.; Golan, J.; Hirvonen, K.; Hoddinott, J.F.; Kim, Sunny; Seyoum Tafesse, 
Alemayehu; Abay, Kibrewossen; Assefa, Thomas; Habte, Yetmwork Evaluation of 
the Nutrition-Sensitive Features of the Fourth Phase of Ethiopia’s Productive 
Safety Net Programme; 0 ed.; International Food Policy Research Institute: 
Washington, DC, 2020; 
4.  Sinaga, M.; Worku, M.; Yemane, T.; Tegene, E.; Wakayo, T.; Girma, T.; Lindstrom, 
D.; Belachew, T. Optimal Cut-off for Obesity and Markers of Metabolic Syndrome 
for Ethiopian Adults. Nutri J 2018, 17, 109, doi:10.1186/s12937-018-0416-0. 
5.  Sinaga, M.; Yemane, T.; Tegene, E.; Lidstrom, D.; Belachew, T. Performance of 
Newly Developed Body Mass Index Cut-off for Diagnosing Obesity among 
Ethiopian Adults. J Physiol Anthropol 2019, 38, 14, doi:10.1186/s40101-019-0205-2. 
6.  Uauy, R.; Kain, J. The Epidemiological Transition: Need to Incorporate Obesity 
Prevention into Nutrition Programmes. Public Health Nutr. 2002, 5, 223–229, 
doi:10.1079/PHN2001297. 
7.  Long, A.E.; Prewitt, T.E.; Kaufman, J.S.; Rotimi, C.N.; Cooper, R.S.; McGee, D.L. 
Weight–Height Relationships among Eight Populations of West African Origin: 
The Case against Constant BMI Standards. Int J Obes 1998, 22, 842–846, 
doi:10.1038/sj.ijo.0800670. 
8.  Black, R.E.; Victora, C.G.; Walker, S.P.; Bhutta, Z.A.; Christian, P.; Onis, M. de; 
Ezzati, M.; Grantham-McGregor, S.; Katz, J.; Martorell, R.; et al. Maternal and Child 
Undernutrition and Overweight in Low-Income and Middle-Income Countries. 
Lancet 2013, 382, 427–451, doi:10.1016/S0140-6736(13)60937-X. 
9.  Heymsfield, Steven B.; Lohman, Timothy G.; d; Going, Scott B. Human Body 
Composition; 2nd ed.; Human Kinetics: Champaign, IL, 2005; ISBN 0736046550. 
10.  Hawkes, C.; Ruel, M.T.; Salm, L.; Sinclair, B.; Branca, F. Double-Duty Actions: 
Seizing Programme and Policy Opportunities to Address Malnutrition in All Its 
Forms. Lancet 2020, 395, 142–155, doi:10.1016/S0140-6736(19)32506-1. 
11.  WHO Consultation on Obesity; Organization W.H. Obesity: Preventing and 
Managing the Global Epidemic: Report of a WHO Consultation; World Health 
Organization: Geneva, Switzerland, 2000; 
18 
 
12.  Flegal, K.M.; Graubard, B.I.; Williamson, D.F.; Gail, M.H. Excess Deaths Associated 
With Underweight, Overweight, and Obesity. JAMA 2005, 293, 7, 
doi:10.1001/jama.293.15.1861. 
13.  Harpsøe, M.C.; Nielsen, N.M.; Friis-Møller, N.; Andersson, M.; Wohlfahrt, J.; 
Linneberg, A.; Nohr, E.A.; Jess, T. Body Mass Index and Risk of Infections Among 
Women in the Danish National Birth Cohort. Am J Epidemiol 2016, 183, 1008–1017, 
doi:10.1093/aje/kwv300. 
14.  Winter-Jensen, M.; Afzal, S.; Jess, T.; Nordestgaard, B.G.; Allin, K.H. Body Mass 
Index and Risk of Infections: A Mendelian Randomization Study of 101,447 
Individuals. Eur J Epidemiol 2020, 35, 347–354, doi:10.1007/s10654-020-00630-7. 
15.  Dobner, J.; Kaser, S. Body Mass Index and the Risk of Infection - from Underweight 
to Obesity. CMI 2018, 24, 24–28, doi:10.1016/j.cmi.2017.02.013. 
16.  King, J.C. The Risk of Maternal Nutritional Depletion and Poor Outcomes Increases 
in Early or Closely Spaced Pregnancies. J Nutr 2003, 133, 1732S-1736S, 
doi:10.1093/jn/133.5.1732S. 
17.  Adair, L.S.; Popkin, B.M. Prolonged Lactation Contributes to Depletion of Maternal 
Energy Reserves in Filipino Women. J Nutr 1992, 122, 1643–1655, 
doi:10.1093/jn/122.8.1643. 
18.  Christian, P.; Lee, S.E.; Donahue Angel, M.; Adair, L.S.; Arifeen, S.E.; Ashorn, P.; 
Barros, F.C.; Fall, C.H.; Fawzi, W.W.; Hao, W.; et al. Risk of Childhood 
Undernutrition Related to Small-for-Gestational Age and Preterm Birth in Low- 
and Middle-Income Countries. Int J Epidemiol 2013, 42, 1340–1355, 
doi:10.1093/ije/dyt109. 
19.  Cordero, M.E.; D’Acuña, E.; Benveniste, S.; Prado, R.; Nuñez, J.A.; Colombo, M. 
Dendritic Development in Neocortex of Infants with Early Postnatal Life 
Undernutrition. Pediatr Neurol 1993, 9, 457–464, doi:10.1016/0887-8994(93)90025-8. 
20.  Alam, M.A.; Richard, S.A.; Fahim, S.M.; Mahfuz, M.; Nahar, B.; Das, S.; Shrestha, 
B.; Koshy, B.; Mduma, E.; Seidman, J.C.; et al. Impact of Early-Onset Persistent 
Stunting on Cognitive Development at 5 Years of Age: Results from a Multi-
Country Cohort Study. PLoS One 2020, 15, e0227839, 
doi:10.1371/journal.pone.0227839. 
21.  Xie, W.; Jensen, S.K.G.; Wade, M.; Kumar, S.; Westerlund, A.; Kakon, S.H.; Haque, 
R.; Petri, W.A.; Nelson, C.A. Growth Faltering Is Associated with Altered Brain 
Functional Connectivity and Cognitive Outcomes in Urban Bangladeshi Children 
Exposed to Early Adversity. BMC Medicine 2019, 17, 199, doi:10.1186/s12916-019-
1431-5. 
22.  Strauss, J.; Thomas, D. Health, Nutrition, and Economic Development. J Econ Lit 
1998, 36, 766–817. 
19 
 
23.  Wolde, T.; Belachew, T. Chronic Undernutrition (Stunting) Is Detrimental to 
Academic Performance among Primary Schools of Adolescent Children: A 
Randomized Cross Sectional Survey in Southern Ethiopia. BMC Research Notes 
2019, 12, 142, doi:10.1186/s13104-019-4160-0. 
24.  Woldehanna, T.; Behrman, J.R.; Araya, M.W. The Effect of Early Childhood 
Stunting on Children’s Cognitive Achievements: Evidence from Young Lives 
Ethiopia. Ethiop J Health Dev 2017, 31, 75–84. 
25.  Black, R.E.; Allen, L.H.; Bhutta, Z.A.; Caulfield, L.E.; Onis, M. de; Ezzati, M.; 
Mathers, C.; Rivera, J. Maternal and Child Undernutrition: Global and Regional 
Exposures and Health Consequences. Lancet 2008, 371, 243–260, 
doi:10.1016/S0140-6736(07)61690-0. 
26.  Walker, S.P.; Chang, S.M.; Powell, C.A.; Simonoff, E.; Grantham-McGregor, S.M. 
Early Childhood Stunting Is Associated with Poor Psychological Functioning in 
Late Adolescence and Effects Are Reduced by Psychosocial Stimulation. J Nutr 
2007, 137, 2464–2469, doi:10.1093/jn/137.11.2464. 
27.  Victora, C.G.; Adair, L.; Fall, C.; Hallal, P.C.; Martorell, R.; Richter, L.; Sachdev, 
H.S.; Maternal and Child Undernutrition Study Group Maternal and Child 
Undernutrition: Consequences for Adult Health and Human Capital. Lancet 2008, 
371, 340–357, doi:10.1016/S0140-6736(07)61692-4. 
28.  Hoddinott, J.; Behrman, J.R.; Maluccio, J.A.; Melgar, P.; Quisumbing, A.R.; 
Ramirez-Zea, M.; Stein, A.D.; Yount, K.M.; Martorell, R. Adult Consequences of 
Growth Failure in Early Childhood. Am J Clin Nutr 2013, 98, 1170–1178, 
doi:10.3945/ajcn.113.064584. 
29.  NHLBI Obesity Education Initiative Expert Panel on the Identification, Evaluation, 
and Treatment of Overweight and Obesity in Adults Clinical Guidelines on the 
Identification, Evaluation, and Treatment of Overweight and Obesity in Adults; 
National Institutes of Health, 1998; 
30.  Bhaskaran, K.; Douglas, I.; Forbes, H.; dos-Santos-Silva, I.; Leon, D.A.; Smeeth, L. 
Body-Mass Index and Risk of 22 Specific Cancers: A Population-Based Cohort 
Study of 5.24 Million UK Adults. Lancet 2014, 384, 755–765, doi:10.1016/S0140-
6736(14)60892-8. 
31.  Roberts, R.E.; Deleger, S.; Strawbridge, W.J.; Kaplan, G.A. Prospective Association 
between Obesity and Depression: Evidence from the Alameda County Study. Int J 
Obes Relat Metab Disord 2003, 27, 514–521, doi:10.1038/sj.ijo.0802204. 
32.  Luppino, F.S.; de Wit, L.M.; Bouvy, P.F.; Stijnen, T.; Cuijpers, P.; Penninx, B.W.J.H.; 
Zitman, F.G. Overweight, Obesity, and Depression: A Systematic Review and 
Meta-Analysis of Longitudinal Studies. Arch Gen Psychiatry 2010, 67, 220–229, 
doi:10.1001/archgenpsychiatry.2010.2. 
20 
 
33.  Goettler, A.; Grosse, A.; Sonntag, D. Productivity Loss Due to Overweight and 
Obesity: A Systematic Review of Indirect Costs. BMJ Open 2017, 7, e014632, 
doi:10.1136/bmjopen-2016-014632. 
34.  Kudel, I.; Huang, J.C.; Ganguly, R. Impact of Obesity on Work Productivity in 
Different US Occupations: Analysis of the National Health and Wellness Survey 
2014 to 2015. JOEM 2018, 60, 6–11, doi:10.1097/JOM.0000000000001144. 
35.  Nugent, R.; Levin, C.; Hale, J.; Hutchinson, B. Economic Effects of the Double 
Burden of Malnutrition. Lancet 2020, 395, 156–164, doi:10.1016/S0140-
6736(19)32473-0. 
36.  Cawley, J.; Biener, A.; Meyerhoefer, C.; Ding, Y.; Zvenyach, T.; Smolarz, B.G.; 
Ramasamy, A. Direct Medical Costs of Obesity in the United States and the Most 
Populous States. J Manag Care Spec Pharm 2021, 27, 354–366, 
doi:10.18553/jmcp.2021.20410. 
37.  Biener, A.; Cawley, J.; Meyerhoefer, C. The High and Rising Costs of Obesity to the 
US Health Care System. J Gen Intern Med 2017, 32, 6–8, doi:10.1007/s11606-016-
3968-8. 
38.  Rasmussen, K.M.; Yaktine, A.L.; Institute of Medicine (US); National Research 
Council (US) Committee to Reexamine IOM pregnancy Weight Guidelines 
Consequences of Gestational Weight Gain for the Mother. In Weight Gain During 
Pregnancy: Reexamining the Guidelines; National Academies Press (US): 
Washington DC, 2009 ISBN 978-0-309-13113-1. 
39.  Crowther, C.A.; Hiller, J.E.; Moss, J.R.; McPhee, A.J.; Jeffries, W.S.; Robinson, J.S. 
Effect of Treatment of Gestational Diabetes Mellitus on Pregnancy Outcomes. N 
Engl J Med 2005, 352, 2477–2486, doi:10.1056/NEJMoa042973. 
40.  Cnattingius, S.; Villamor, E.; Johansson, S.; Bonamy, A.-K.E.; Persson, M.; 
Wikström, A.-K.; Granath, F. Maternal Obesity and Risk of Preterm Delivery. 
JAMA 2013, 309, 2362–2370, doi:10.1001/jama.2013.6295. 
41.  Su, X.J.; Huang, S.J.; Li, X.; Du, Q.L. Prepregnancy Overweight and Obesity Are 
Associated with an Increased Risk of Preterm Birth in Chinese Women. OFA 2020, 
2, 237–244, doi:10.1159/000506688. 
42.  Flenady, V.; Koopmans, L.; Middleton, P.; Frøen, J.F.; Smith, G.C.; Gibbons, K.; 
Coory, M.; Gordon, A.; Ellwood, D.; McIntyre, H.D.; et al. Major Risk Factors for 
Stillbirth in High-Income Countries: A Systematic Review and Meta-Analysis. 
Lancet 2011, 377, 1331–1340, doi:10.1016/S0140-6736(10)62233-7. 
43.  Di Cesare, M.; Ghosh, S.; Osendarp, S.; Mozaffarian, D. A World Free from 
Malnutrition: An Assessment of Progress towards the Global Nutrition Targets. In 
2021 Global Nutrition Report: Development Initiatives Poverty Research Ltd.; 
Development Initiatives: Bristol UK, 2021. 
21 
 
44.  Taffesse, A.S.; Dorosh, P.; Gemessa, S.A. 3 Crop Production in Ethiopia: Regional 
Patterns and Trends. In Food and Agriculture in Ethiopia; Dorosh, P., Rashid, S., 
Eds.; University of Pennsylvania Press: Philadelphia, 2013 ISBN 978-0-8122-0861-0. 
45.  Hirvonen, K.; Taffesse, A.S.; Worku Hassen, I. Seasonality and Household Diets in 
Ethiopia. Public Health Nutr. 2016, 19, 1723–1730, doi:10.1017/S1368980015003237. 
46.  Gilbert, C.L.; Christiaensen, L.; Kaminski, J. Food Price Seasonality in Africa: 
Measurement and Extent. Food Policy 2017, 67, 119–132, 
doi:10.1016/j.foodpol.2016.09.016. 
47.  FAO and FHI 360 Minimum Dietary Diversity for Women- A Guide to 
Measurement; FAO: Rome, 2016; 
48.  World Health Organization Indicators for Assessing Infant and Young Child 
Feeding Practices Part 2 Measurement; WHO: Geneva, 2010; ISBN 978 92 4 159929 
0. 
49.  Taddese, A.A.; Baraki, A.G.; Gelaye, K.A. Spatial Modeling, Prediction and 
Seasonal Variation of Malaria in Northwest Ethiopia. BMC Research Notes 2019, 
12, 273, doi:10.1186/s13104-019-4305-1. 
50.  Gashu, Z.; Jerene, D.; Datiko, D.G.; Hiruy, N.; Negash, S.; Melkieneh, K.; Bekele, D.; 
Nigussie, G.; Suarez, P.G.; Hadgu, A. Seasonal Patterns of Tuberculosis Case 
Notification in the Tropics of Africa: A Six-Year Trend Analysis in Ethiopia. PLoS 
One 2018, 13, e0207552, doi:10.1371/journal.pone.0207552. 
51.  Altizer, S.; Dobson, A.; Hosseini, P.; Hudson, P.; Pascual, M.; Rohani, P. Seasonality 
and the Dynamics of Infectious Diseases. Ecology Letters 2006, 9, 467–484, 
doi:10.1111/j.1461-0248.2005.00879.x. 
52.  Branca, F.; Pastore, G.; Demissie, T.; Ferro-Luzzi, A. The Nutritional Impact of 
Seasonality in Children and Adults of Rural Ethiopia. Eur J Clin Nutr 1993, 47, 840–
850. 
53.  Ferro-Luzzi, A.; Morris, S.S.; Taffesse, S.; Demissie, T.; D’Amato, M. Seasonal 
Undernutrition in Rural Ethiopia: Magnitude, Correlates, and Functional 
Significance. Food Nutr Bull 2002, 23, 227–228, doi:10.1177/156482650202300211. 
54.  Maleta, K.; Virtanen, S.; Espo, M.; Kulmala, T.; Ashorn, P. Seasonality of Growth 
and the Relationship between Weight and Height Gain in Children under Three 
Years of Age in Rural Malawi. Acta Paediatrica 2003, 92, 491–497, 
doi:10.1111/j.1651-2227.2003.tb00584.x. 
55.  Vaitla, B.; Devereux, S.; Swan, S.H. Seasonal Hunger: A Neglected Problem with 
Proven Solutions. PLOS Medicine 2009, 6, e1000101, 
doi:10.1371/journal.pmed.1000101. 
22 
 
56.  Abay, K.; Hirvonen, K. Does Market Access Mitigate the Impact of Seasonality on 
Child Growth? Panel Data Evidence from Northern Ethiopia. J Dev Stud 2017, 53, 
1414–1429, doi:10.1080/00220388.2016.1251586. 
57.  Baye, K.; Hirvonen, K. Seasonality: A Missing Link in Preventing Undernutrition. 
Lancet Child Adolesc Health 2020, 4, e3, doi:10.1016/S2352-4642(19)30343-8. 
58.  Sinaga, M.; Teshome, M.S.; Kidane, R.; Yemane, T.; Tegene, E.; Lindstrom, D.; 
Belachew, T. Metabolic Effects of Fasting and Animal Source Food Avoidance in an 
Ethiopian Adult Cohort. Sci Rep 2019, 9, 16964, doi:10.1038/s41598-019-53185-3. 
59.  Seleshe, S.; Jo, C.; Lee, M. Meat Consumption Culture in Ethiopia. Korean J Food 
Sci Anim Resour 2014, 34, 7–13, doi:10.5851/kosfa.2014.34.1.7. 
60.  Kim, S.S.; Nguyen, P.H.; Tran, L.M.; Abebe, Y.; Asrat, Y.; Tharaney, M.; Menon, P. 
Maternal Behavioural Determinants and Livestock Ownership Are Associated with 
Animal Source Food Consumption among Young Children during Fasting in Rural 
Ethiopia. Matern Child Nutr 2018, 15, e12695, doi:10.1111/mcn.12695. 
61.  Geda, N.R. Inequalities in Maternal Malnutrition in Ethiopia: Evidence from a 
Nationally Representative Data. BMC Women’s Health 2021, 21, 3, 
doi:10.1186/s12905-020-01154-8. 
62.  World Bank The State of Social Safety Nets 2018; World Bank: Washington, DC, 
2018; 
63.  Subbarao, K.; Ninno, C. del; Andrews, C.; Rodríguez-Alas, C. Public Works as a 
Safety Net : Design, Evidence, and Implementation; The World Bank Group, 2013; 
64.  Ruducha, J.; Mann, C.; Singh, N.S.; Gemebo, T.D.; Tessema, N.S.; Baschieri, A.; 
Friberg, I.; Zerfu, T.A.; Yassin, M.; Franca, G.A.; et al. How Ethiopia Achieved 
Millennium Development Goal 4 through Multisectoral Interventions: A 
Countdown to 2015 Case Study. Lancet Glob Health 2017, 5, e1142–e1151, 
doi:10.1016/S2214-109X(17)30331-5. 
65.  Berhane, G.; Hoddinott, J.; Kumar, N.; Margolies, A. The Productive Safety Net 
Programme in Ethiopia: Impacts on Children’s Schooling, Labour and Nutritional 
Status; 2017th ed.; International Initiative for Impact Evaluation (3ie): New Delhi, 
2017; 
66.  Shigute, Z.; Mebratie, A.D.; Sparrow, R.; Yilma, Z.; Alemu, G.; Bedi, A.S. Uptake of 
Health Insurance and the Productive Safety Net Program in Rural Ethiopia. Soc Sci 
Med 2017, 176, 133–141, doi:10.1016/j.socscimed.2017.01.035. 
67.  Berhane, G.; Gilligan, D.O.; Hoddinott, J.; Kumar, N.; Taffesse, A.S. Can Social 
Protection Work in Africa? The Impact of Ethiopia’s Productive Safety Net 
Programme. Econ Dev Cult Change 2014, 63, 1–26, doi:10.1086/677753. 
68.  Filipski, M.; Taylor, J.E.; Abegaz, G.A.; Ferede, T.; Taffesse, A.S.; Diao, X. Synopsis: 
Economy-Wide Impacts of the Productive Safety Net Programme (PSNP). 2. 
23 
 
69.  Goacher, P.J.; Lambert, R.; Moffatt, P.G. Can Weight-Related Health Risk Be More 
Accurately Assessed by BMI, or by Gender Specific Calculations of Percentage 
Body Fatness? Med Hypotheses 2012, 79, 656–662, doi:10.1016/j.mehy.2012.08.003. 
70.  Mandel, D.; Zimlichman, E.; Mimouni, F.B.; Grotto, I.; Kreiss, Y. Height-Related 
Changes in Body Mass Index: A Reappraisal. J Am Coll Nutr 2004, 23, 51–54, 
doi:10.1080/07315724.2004.10719342. 
71.  Bray, G. What Is the Ideal Body Weight? J Nutr Biochem 1998, 9, 489–492, 
doi:10.1016/S0955-2863(98)00007-2. 
72.  Ode, J.J.; Pivarnik, J.M.; Reeves, M.J.; Knous, J.L. Body Mass Index as a Predictor of 
Percent Fat in College Athletes and Nonathletes. Med Sci Sports Exerc 2007, 39, 
403–409, doi:10.1249/01.mss.0000247008.19127.3e. 
73.  Goh, V.H.H.; Tain, C.F.; Tong, T.Y.Y.; Mok, H.P.P.; Wong, M.T. Are BMI and Other 
Anthropometric Measures Appropriate as Indices for Obesity? A Study in an Asian 
Population. J Lipid Res 2004, 45, 1892–1898, doi:10.1194/jlr.M400159-JLR200. 
74.  Bagust, A.; Walley, T. An Alternative to Body Mass Index for Standardizing Body 
Weight for Stature. QJM-Int J Med 2000, 93, 589–596, doi:10.1093/qjmed/93.9.589. 
75.  Jackson, A.S.; Stanforth, P.R.; Gagnon, J.; Rankinen, T.; Leon, A.S.; Rao, D.C.; 
Skinner, J.S.; Bouchard, C.; Wilmore, J.H. The Effect of Sex, Age and Race on 
Estimating Percentage Body Fat from Body Mass Index: The Heritage Family 
Study. Int J Obes 2002, 26, 789–796, doi:10.1038/sj.ijo.0802006. 
76.  WHO Expert Consultation Appropriate Body-Mass Index for Asian Populations 
and Its Implications for Policy and Intervention Strategies. Lancet 2004, 363, 157–
163, doi:10.1016/S0140-6736(03)15268-3. 
77.  McArdle, W.D. Exercise Physiology: Nutrition, Energy, and Human Performance; 
Eighth edition.; Wolters Kluwer Health/Lippincott Williams & Wilkins: Baltimore, 
2015; ISBN 978-1-4511-9155-4. 
78.  Mora, S.; Pessin, J.E. An Adipocentric View of Signaling and Intracellular 
Trafficking. Diabetes Metab Res Rev 2002, 18, 345–356, doi:10.1002/dmrr.321. 
79.  Trayhurn, P. Endocrine and Signalling Role of Adipose Tissue: New Perspectives 
on Fat. Acta Physiol Scand 2005, 184, 285–293, doi:10.1111/j.1365-
201X.2005.01468.x. 
80.  Bouillanne, O.; Dupont-Belmont, C.; Hay, P.; Hamon-Vilcot, B.; Cynober, L.; 
Aussel, C. Fat Mass Protects Hospitalized Elderly Persons against Morbidity and 
Mortality. Am J Clin Nutr 2009, 90, 505–510, doi:10.3945/ajcn.2009.27819. 
81.  Trayhurn, P.; Beattie, J.H. Physiological Role of Adipose Tissue: White Adipose 
Tissue as an Endocrine and Secretory Organ. Proc Nutr Soc 2001, 60, 329–339, 
doi:10.1079/PNS200194. 
24 
 
82.  Sedlmeier, A.M.; Baumeister, S.E.; Weber, A.; Fischer, B.; Thorand, B.; Ittermann, 
T.; Dörr, M.; Felix, S.B.; Völzke, H.; Peters, A.; et al. Relation of Body Fat Mass and 
Fat-Free Mass to Total Mortality: Results from 7 Prospective Cohort Studies. Am J 
Clin Nutr 2021, 113, 639–646, doi:10.1093/ajcn/nqaa339. 
83.  Popkin, B.M.; Gordon-Larsen, P. The Nutrition Transition: Worldwide Obesity 
Dynamics and Their Determinants. Int J Obes 2004, 28, S2–S9, 
doi:10.1038/sj.ijo.0802804. 
84.  Szymańska, E.; Bouwman, J.; Strassburg, K.; Vervoort, J.; Kangas, A.J.; Soininen, P.; 
Ala-Korpela, M.; Westerhuis, J.; van Duynhoven, J.P.M.; Mela, D.J.; et al. Gender-
Dependent Associations of Metabolite Profiles and Body Fat Distribution in a 
Healthy Population with Central Obesity: Towards Metabolomics Diagnostics. 
OMICS 2012, 16, 652–667, doi:10.1089/omi.2012.0062. 
85.  Rosen, E.D.; Spiegelman, B.M. What We Talk about When We Talk about Fat. Cell 
2014, 156, 20–44, doi:10.1016/j.cell.2013.12.012. 
86.  Nutrition, For Every Child, UNICEF Nutrition Strategy 2020-2030, In Brief; 
UNICEF: New York, 2020; 
87.  Psota, T.; Chen, K. Measuring Energy Expenditure in Clinical Populations: 
Rewards and Challenges. Eur J Clin Nutr 2013, 67, 436–442, 
doi:10.1038/ejcn.2013.38. 
88.  Chung, N.; Park, M.-Y.; Kim, J.; Park, H.-Y.; Hwang, H.; Lee, C.-H.; Han, J.-S.; So, J.; 
Park, J.; Lim, K. Non-Exercise Activity Thermogenesis (NEAT): A Component of 
Total Daily Energy Expenditure. J Exerc Nutrition Biochem 2018, 22, 23–30, 
doi:10.20463/jenb.2018.0013. 
89.  Bull, F.C.; Maslin, T.S.; Armstrong, T. Global Physical Activity Questionnaire 
(GPAQ): Nine Country Reliability and Validity Study. J Phys Act Health 2009, 6, 
790–804, doi:10.1123/jpah.6.6.790. 
90.  Ndahimana, D.; Kim, E.-K. Measurement Methods for Physical Activity and 
Energy Expenditure: A Review. Clin Nutr Res 2017, 6, 68–80, 
doi:10.7762/cnr.2017.6.2.68. 
91.  Gunzer W.; Peterseil M.; Pilz R.; Pail E.; Fuchs-Neuhold B. Assessment of Maternal 
Nutrition during Pregnancy and [Early] Lactation Period in Austria: Comparison 
of a Food- Frequency Questionnaire with a 24-Hour Recall. Obes. Facts 2018, 11, 
204, doi:10.1159/000489691. 
92.  Savard C.; Lemieux S.; Lafrenière J.; Laramée C.; Robitaille J.; Morisset A.-S. 
Validation of a Self-Administered Web-Based 24-Hour Dietary Recall among 
Pregnant Women. BMC Pregnancy Childbirth 2018, 18, doi:10.1186/s12884-018-
1741-1. 
25 
 
93.  Alkire, S.; Meinzen-Dick, R.; Peterman, A.; Quisumbing, A.; Seymour, G.; Vaz, A. 
The Women’s Empowerment in Agriculture Index. World Dev 2013, 52, 71–91, 
doi:10.1016/j.worlddev.2013.06.007. 
94.  Surveillance and Population-Based Prevention Global Physical Activity 
Questionnaire (GPAQ) Analysis Guide; World Health Organization: Geneva, 
Switzerland, 2005; 
95.  Leidy, N.K. Functional Status and the Forward Progress of Merry-Go-Rounds: 
Toward a Coherent Analytical Framework. Nurs Res 1994, 43, 196–202. 
96.  Gleeson, M.; Sherrington, C.; Keay, L. Exercise and Physical Training Improve 
Physical Function in Older Adults with Visual Impairments but Their Effect on 
Falls Is Unclear: A Systematic Review. J Physiother 2014, 60, 130–135, 
doi:10.1016/j.jphys.2014.06.010. 
97.  Wolfe, R.R. The Underappreciated Role of Muscle in Health and Disease. Am J Clin 
Nutr 2006, 84, 475–482, doi:10.1093/ajcn/84.3.475. 
98.  Trivedi, H.D.; Tapper, E.B. Interventions to Improve Physical Function and Prevent 
Adverse Events in Cirrhosis. Gastroenterol Rep (Oxf) 2018, 6, 13–20, 
doi:10.1093/gastro/gox042. 
99.  Tomey, K.M.; Sowers, M.R. Assessment of Physical Functioning: A Conceptual 
Model Encompassing Environmental Factors and Individual Compensation 
Strategies. Phys Ther 2009, 89, 705–714, doi:10.2522/ptj.20080213. 
100.  O’Neill, D.; Forman, D.E. The Importance of Physical Function as a Clinical 
Outcome: Assessment and Enhancement. Clin Cardiol 2020, 43, 108–117, 
doi:10.1002/clc.23311. 
101.  Seeman, T.E.; Charpentier, P.A.; Berkman, L.F.; Tinetti, M.E.; Guralnik, J.M.; Albert, 
M.; Blazer, D.; Rowe, J.W. Predicting Changes in Physical Performance in a High-
Functioning Elderly Cohort: MacArthur Studies of Successful Aging. J Gerontol 
1994, 49, M97–M108, doi:10.1093/geronj/49.3.M97. 
102.  Wolinsky, F.D.; Miller, D.K.; Andresen, E.M.; Malmstrom, T.K.; Miller, J.P.; Miller, 
T.R. Effect of Subclinical Status in Functional Limitation and Disability on Adverse 
Health Outcomes 3 Years Later. J Gerontol A Biol Sci Med Sci 2007, 62, 101–106, 
doi:10.1093/gerona/62.1.101. 
103.  Viccaro, L.J.; Perera, S.; Studenski, S.A. Is Timed Up and Go Better Than Gait Speed 
in Predicting Health, Function, and Falls in Older Adults? J Am Geriatr Soc 2011, 
59, 887–892, doi:10.1111/j.1532-5415.2011.03336.x. 
104.  Fried, T.R.; Bradley, E.H.; Williams, C.S.; Tinetti, M.E. Functional Disability and 
Health Care Expenditures for Older Persons. Arch Intern Med 2001, 161, 2602–
2607, doi:10.1001/archinte.161.21.2602. 
26 
 
105.  Imamura, K.; Kamide, N.; Sakamoto, M.; Sato, H.; Shiba, Y.; Matsunaga, A. 
Association Between Social Network and Physical Function in Community-
Dwelling Older Adults in Japan. Phys Ther Res 2020, 23, 153–159, 
doi:10.1298/ptr.E10028. 
106.  Rogers, S.N.; Lowe, D.; Yueh, B.; Weymuller Jr, E.A. The Physical Function and 
Social-Emotional Function Subscales of the University of Washington Quality of 
Life Questionnaire. Arch Otolaryngol Head Neck Surg 2010, 136, 352–357, 
doi:10.1001/archoto.2010.32. 
107.  Noppert, G.A.; Brown, C.S.; Chanti-Ketterl, M.; Hall, K.S.; Newby, L.K.; Cohen, 
H.J.; Morey, M.C. The Impact of Multiple Dimensions of Socioeconomic Status on 
Physical Functioning Across the Life Course. Gerontol Geriatr Med 2018, 4, 
2333721418794021, doi:10.1177/2333721418794021. 
108.  Nilsen, C.; Agahi, N.; Kåreholt, I. Work Stressors in Late Midlife and Physical 
Functioning in Old Age. J Aging Health 2017, 29, 893–911, 
doi:10.1177/0898264316654673. 
109.  Palumbo, A.; De Roos, A.; Cannuscio, C.; Robinson, L.; Mossey, J.; Weitlauf, J.; 
Garcia, L.; Wallace, R.; Michael, Y. Work Characteristics Associated with Physical 
Functioning in Women. IJERPH 2017, 14, 424, doi:10.3390/ijerph14040424. 
110.  McLoughlin, D.M.; Spargo, E.; Wassif, W.S.; Newham, D.J.; Peters, T.J.; Lantos, 
P.L.; Russell, G.F.M. Structural and Functional Changes in Skeletal Muscle in 
Anorexia Nervosa. Acta Neuropathol 1998, 95, 632–540, 
doi:10.1007/s004010050850. 
111.  Russell, D.M.; Prendergast, P.J.; Darby, P.L.; Garfinkel, P.E.; Whitwell, J.; 
Jeejeebhoy, K.N. A Comparison between Muscle Function and Body Composition 
in Anorexia Nervosa: The Effect of Refeeding. Am J Clin Nutr 1983, 38, 229–237, 
doi:10.1093/ajcn/38.2.229. 
112.  Clarke, N.E.; Clements, A.C.A.; Bryan, S.; McGown, J.; Gray, D.; Nery, S.V. 
Investigating the Differential Impact of School and Community-Based Integrated 
Control Programmes for Soil-Transmitted Helminths in Timor-Leste: The 
(S)WASH-D for Worms Pilot Study Protocol. Pilot Feasibility Stud 2016, 2, 69, 
doi:10.1186/s40814-016-0109-4. 
113.  Abate, K.H.; Belachew, T. Care and Not Wealth Is a Predictor of Wasting and 
Stunting of ‘The Coffee Kids’ of Jimma Zone, Southwest Ethiopia. Nutr Health 
2017, 23, 193–202, doi:10.1177/0260106017706253. 
114.  Hangoma, P.; Aakvik, A.; Robberstad, B. Explaining Changes in Child Health 
Inequality in the Run up to the 2015 Millennium Development Goals (MDGs): The 
Case of Zambia. PLoS One 2017, 12, e0170995, doi:10.1371/journal.pone.0170995. 
115.  Abebe, Z.; Haki, G.D.; Baye, K. Health Extension Workers’ Knowledge and 
Knowledge-Sharing Effectiveness of Optimal Infant and Young Child Feeding Are 
27 
 
Associated With Mothers’ Knowledge and Child Stunting in Rural Ethiopia. Food 
Nutr Bull 2016, 37, 353–363, doi:10.1177/0379572116651209. 
116.  Vaidya, T.; Chambellan, A.; de Bisschop, C. Sit-to-Stand Tests for COPD: A 
Literature Review. Respir Med 2017, 128, 70–77, doi:10.1016/j.rmed.2017.05.003. 
117.  Donoghue, O.A.; Savva, G.M.; Cronin, H.; Kenny, R.A.; Horgan, N.F. Using Timed 
Up and Go and Usual Gait Speed to Predict Incident Disability in Daily Activities 
Among Community-Dwelling Adults Aged 65 and Older. Arch Phys Med Rehabil 
2014, 95, 1954–1961, doi:10.1016/j.apmr.2014.06.008. 
118.  Landi, F.; Russo, A.; Liperoti, R.; Tosato, M.; Barillaro, C.; Pahor, M.; Bernabei, R.; 
Onder, G. Anorexia, Physical Function, and Incident Disability Among the Frail 
Elderly Population: Results From the IlSIRENTE Study. J Am Med Dir Assoc 2010, 
11, 268–274, doi:10.1016/j.jamda.2009.12.088. 
119.  Hjorth, M.F.; Kloster, S.; Girma, T.; Faurholt-Jepsen, D.; Andersen, G.; Kæstel, P.; 
Brage, S.; Friis, H. Level and Intensity of Objectively Assessed Physical Activity 
among Pregnant Women from Urban Ethiopia. BMC Pregnancy Childbirth 2012, 
12, 154, doi:10.1186/1471-2393-12-154. 
120.  Payne, C.F.; Gómez-Olivé, F.X.; Kahn, K.; Berkman, L. Physical Function in an 
Aging Population in Rural South Africa: Findings from HAALSI and Cross-
National Comparisons With HRS Sister Studies. J Gerontol B Psychol Sci Soc Sci 
2017, 72, 665–679, doi:10.1093/geronb/gbx030. 
121.  Cormie, P.; McGuigan, M.R.; Newton, R.U. Developing Maximal Neuromuscular 
Power. Sports Med 2011, 41, 17–38, doi:10.2165/11537690-000000000-00000. 
122.  Barbat-Artigas, S.; Rolland, Y.; Zamboni, M.; Aubertin-Leheudre, M. How to 
Assess Functional Status: A New Muscle Quality Index. J Nutr Health Aging 2012, 
16, 67–77, doi:10.1007/s12603-012-0004-5. 
123.  Lindemann, U.; Claus, H.; Stuber, M.; Augat, P.; Muche, R.; Nikolaus, T.; Becker, C. 
Measuring Power during the Sit-to-Stand Transfer. Eur J Appl Physiol 2003, 89, 
466–470, doi:10.1007/s00421-003-0837-z. 
124.  Gray, M.; Paulson, S. Developing a Measure of Muscular Power during a 
Functional Task for Older Adults. BMC Geriatrics 2014, 14, 145, doi:10.1186/1471-
2318-14-145. 
 
 
  
28 
 
CHAPTER 2 
VALIDATION OF SKINFOLD THICKNESS AND BIOELECTRICAL IMPEDANCE 
EQUATIONS IN ETHIOPIAN ADULTS 
Abstract 
Methods to measure fat mass (FM) and fat-free mass (FFM) in resource-
constrained settings are limited and dependent on prediction equations that have not 
been validated in several populations, including Ethiopian adults.  Valid prediction 
equations are critical tools as the rates of overweight and obesity increase in low- and 
middle-income countries (LMIC).  We aimed to validate existing bioelectrical 
impedance analysis (BIA) and skinfold-thickness (SFT) equations using air 
displacement plethysmography (ADP).  Subsequently, we created a new BIA equation 
for females and sex-specific SFT equations.  Body composition as measured by several 
existing, commonly used BIA and SFT prediction equations were evaluated against the 
results of ADP.  A total of 126 females and 129 males residing in Jimma City, Ethiopia, 
volunteered.  Study participants were 18-45 years old and without complicating 
medical conditions.  New prediction equations were created using partial least-squares 
regression.  One existing equation was valid for measuring FFM in males using BIA.  
No existing equations were valid for measuring FFM using BIA in females or SFT in 
males or females. Therefore, new, sex-specific equations were created.  The validated 
equation and the new proposed equations will be valuable tools in identifying those 
with insufficient or excessive adiposity and can be used to identify those at risk for the 
associated adverse health consequences.  Additionally, this work adds to the body of 
evidence that the current tools used to screen for overweight and obesity are inadequate 
in sub-Saharan African populations. 
29 
 
Introduction 
Overweight and obesity are increasing in low- and middle-income countries 
(LMIC) and are prevalent alongside undernutrition resulting in a double burden of 
malnutrition [1–3].  Development in low-income countries fosters an obesogenic 
environment, especially in urban areas, where people have a sedentary lifestyle and 
greater access to calorically dense foods [4].  Increased obesity is associated with the 
rising prevalence of non-communicable diseases, including type II diabetes and 
cardiovascular disease (1,4–7), and negative economic impacts through lost wages, 
productivity, and high medical expenses [8–11].   Obesity and overweight are 
determined using body mass index (BMI) because its components, height and total 
body mass, are easy to collect and use in calculations.   
BMI assumes a consistent relationship between fat mass (FM) and fat-free mass 
(FFM) across age, sex, physical activity levels, and ethnicity [12], and it assumes 
universal cutoff points for obesity-related health risks [13].  These assumptions ignore 
people with a normal BMI with excess adipose tissue and the associated adverse health 
outcomes [14–16] and metabolically healthy people with a high BMI [17].  The 
inadequacy of BMI to identify those at risk of adverse health outcomes is even greater 
in non-white populations [13,18]. One study found that Ethiopian adults experienced 
the signs and symptoms of metabolic syndrome at significantly lower BMI [19].  The 
researchers concluded that the obesity cutoff should decrease to 22.2 kg/ 2m  for males and 
24.5 kg/ 2m  for females [20].   
A better approach to screen for health risks is a two-compartment body 
composition model that quantifies FM, FFM, and percent body fat [21].  While some 
adipose tissue is necessary due to its roles in the storage of energy [22], endocrinology 
[23], and immune function [24,25]. The quantity and location of excessive adipose are 
30 
 
associated with mortality [26] and morbidity [4,27–29]. Thus, limiting nutritional status 
to BMI oversimplifies the relationship between nutritional status and adverse health 
outcomes.   
When available, clinicians and researchers use criterion methodologies to 
measure body composition, such as dual-energy x-ray absorptiometry (DXA), 
hydrodensitometry, and air displacement plethysmography (ADP).  Unfortunately, the 
equipment needed for these methodologies is cost-prohibitive [30] and too heavy and 
cumbersome to transport, limiting their use in LMIC [21].  Bioelectrical impedance 
analysis (BIA) and skinfold thickness (SFT) are two methods used to measure FM and 
FFM that are less expensive and more transportable.   
Unfortunately, BIA and SFT prediction equations are not generalizable because 
they are dependent on age, sex, physical activity, and ethnicity [31–33].  Most SFT 
prediction equations [34] were created using populations of European descent [35–38], 
and previous studies have concluded that they are not valid in populations of African 
descent [39–41]. On the other hand, several BIA equations have been validated in non-
white populations [33,42–44].  
Ethiopia is one country coping with the double burden of malnutrition that does 
not have a validated method to calculate FM and FFM from SFT or BIA.  According to 
the 2016 DHS, nationally, 22% of adult females and 33% of adult males are underweight 
[45].  In urban areas, 22% of adult females and 12% of adult males had a BMI classifying 
them as overweight or obese [45].  Previous studies on body composition in Ethiopia 
have either focused on infants and children [46–48], used incomplete cross-validation 
analysis techniques [49], or stopped short of validating new prediction equations [50].   
This study evaluates existing prediction equations that use BIA to calculate 
percent body fat and SFT to calculate body density.  After determining that the BIA 
31 
 
equations do not adequately assess FM and FFM in adult Ethiopian females and that the 
SFT equations do not adequately assess FM and FFM in adult Ethiopian males or 
females, new equations are proposed and validated.   
Materials and methods 
Study participants 
Ethiopia, like many countries, has limited access to reference measures, such as 
hydrodensitometry and no access to DXA.  However, the Jimma University Clinical and 
Nutrition Research Center (JUCAN) operates a Bod Pod, making it possible to use ADP 
to measure body composition. Given its size and reputation, Jimma University attracts 
students from all over Ethiopia, thus providing an ethnically diverse population, an 
essential consideration as Ethiopia has over 80 different ethnicities.  The majority of the 
population of Ethiopia identifies as either Oromo or Amhara [45].  
Participants (129 males and 125 females) were recruited from the community 
surrounding Jimma University in Jimma City in the Oromia region of Ethiopia. They 
were recruited using multilingual posters (English, Oromo, and Amharic) around 
Jimma University and the local municipal government office.  Additionally, study 
participants were recruited by existing study participants and locally influential 
community members.  Participants were included if they were between 18 and 45 years 
old [51,52].  Participants were excluded if they reported that they were pregnant or 
lactating, diabetic, or hypertensive.  Pregnancy and lactation lead to changes in weight, 
body density, and hydration that require condition-specific equations [53,54].  Diabetes 
and hypertension also require condition-specific equations due to fluid retention 
[55,56].     
32 
 
 All measurements were performed in the morning before the participants had 
broken their overnight fast.  If participants had broken their fast, they were asked to 
return at a later date.  Three participants did not meet the inclusion criteria (1 was 
lactating, 1 had diabetes mellitus, and 1 had hypertension).   
The Internal Review Board of Cornell University (Protocol ID# 1806008048) and 
Jimma University (Ref. No. IHRPGD\454\2018) approved the protocol.  All 
participants were informed of its purpose, requirements, and procedures before 
providing oral consent to participate in the study. 
Air-displacement plethysmography 
ADP was the reference method for this study.  ADP was measured using a Bod 
Pod GS (COSMED, Rome, Italy).  A trained technician operated the Bod Pod.  During 
the measurement process, participants wore underpants and a bathing cap.  The 
participants were first weighed using the Bod Pod system electronic scale (Tanita Corp., 
Tokyo, Japan). Afterward, the participants sat quietly in the Bod Pod test chamber and 
breathed normally.   
Anthropometric measures 
Five anthropometric measurements were taken by trained personnel using 
standard procedures [57].  Measurements were performed in random sequential order 
on the same day.  The participants’ height was measured to the neared 0.01 cm (763, 
Seca Germany).  Waist circumference was measured on a horizontal plane taken from 
the narrowest part of the umbilicus, while hip circumference was measured on a 
horizontal plane at the level of the greater trochanter using a standard tape measure.  
Mid-upper arm circumference (MUAC) was measured halfway between the olecranon 
process and acromion.  All anthropometric measurements were taken in duplicate, and 
33 
 
the mean of the two measurements was used for the analysis.  From these values, BMI 
and waist to hip ratio were calculated.  Age was collected in years, and physical activity 
was recorded on a scale of 1 to 4.  One indicated a sedentary life, whereas four indicated 
that the study participant regularly engaged in vigorous physical activity. 
Bioelectrical impedance analysis 
FM and FFM have different electrical conductivity, resulting in different 
resistances to electrical currents.  BIA uses an analyzer to send a harmless electrical 
current through the body to measure the tissue’s resistance.  The resistance is inputted 
into a predictive equation that estimates percent body fat [58].  BIA measurements were 
taken using Biodynamics BIA 450 Bioimpedance Analyzer (Shoreline, WA) according to 
the manufacturer’s instructions [58].  Measurements were taken in duplicate with a 5-
min wait in between measurements.  The mean of these measurements was used for 
data analysis. 
Skinfold thickness 
SFT depends on the relationship between subcutaneous fat, visceral fat, and 
whole-body density [34].  SFT uses calipers to measure subcutaneous fat and then 
inputs the measurements into predictive equations to calculate body density.  SFT 
measurements were made by trained personnel using standard procedures [59].  SFT 
was measured at five sites on females, the bicep, triceps, subscapular, suprailiac, and 
thigh.  SFT was measured at the same five sites on males with two additional sites, the 
chest and abdomen.  All SFT measurements were taken using Harpenden Calipers 
(Baty International, West Sussex, UK), on the right side of the body, in triplicate, and to 
the nearest millimeter.  The mean of the SFT measurements was used for data analysis. 
34 
 
Existing equations 
For males and females, five BIA equations were evaluated for validity.  Two 
equations,  the modified Segal fatness-specific equation [33,60] and the Sun et al. 
equation [42], were previously found to be acceptable for calculating percent body fat in 
African Americans [43,61].  Three equations were developed using populations of 
African descent, the Modeling Epidemiologic Transition Study (METS) equations [44], 
the Luke equation [62], and the Zillikens equation [63]. 
These three existing skinfold thickness equations were evaluated for both 
females and males.  The equations were selected because they frequently appear in the 
literature about using skinfold thickness measurements for calculating percent FM. For 
females, these were Sloan [64]; Jackson, Pollack, & Ward [37]; and Durnin and 
Wormsley [38].  For males, the three equations were Sloan [35], Jackson and Pollock 
[36], and Durnin and Wormsley [38].  Percent FM was calculated from body density 
using the Schutte equation [65]. 
Statistical analysis 
The number of study participants was determined based upon previous studies, 
which developed skinfold thickness equations using sample sizes ranging from 50 
participants [35,64] to 400 [36]. Statistical analysis was conducted R 4.0.4 (The R 
Foundation for Statistical Computing Platform) and RStudio V1.44.1106 (RStudio, PBC).   
Existing BIA equations were evaluated for validity by comparing FFM calculated 
by the BIA equations to FFM calculated from body density measured by the Bod Pod.  
Existing SFT equations were evaluated for validity by comparing body density 
calculated by the SFT equations to body density measured by the Bod Pod.  BIA 
equations and SFT equations were evaluated using Lohman’s cross-validation 
35 
 
procedures [66].  These included 1) having a substantial correlation (r > 0.80), 2) not 
having means that significantly differ according to a Student’s t test, 3) having the 
standard error of estimates (SEE) and the total error (TE) not exceed 2.8 kg for females 
or 3.5 kg for males, 4) having the slope of the line describing the relationship between 
the criterion values and predicted values not differ significantly from one and the 
intercept of the line not significantly differ from zero, 5) a Bland-Altman plot where the 
y-axis was centered at zero and little or no systematic prediction error across degrees of 
body fatness [67].   
Additionally, the Kolmogorov-Smirnov (KS) test was used to determine if the 
values measured by the criterion method had the same distribution as those calculated 
by the predicted equations.  An a of 0.05 was set as the cutoff for significance for both 
the Student’s t test and the Kolmogorov-Smirnov test. Thus, an equation was 
considered valid if it met all six cross-validation criteria.   
Below we show that existing BIA equations were not valid for adult Ethiopian 
females and that existing SFT equations were not valid for males or females; therefore, 
new sex-specific equations needed to be established.  The new BIA equation and the 
two new SFT equations were created using the same methodologies. First, participants 
were randomly split into a training set (67%) and a testing set (33%). Then, new 
equations were created using the training set of participants, while the testing set was 
used to validate the new equations.   
Four variables evaluated for inclusion into the new females’ specific BIA 
equation were height in cm, weight in kg, resistance (R), and height squared divided by 
resistance (H2/R).  Each variable was evaluated against body density as measured by 
the Bod Pod to determine if they were significantly associated.   
36 
 
The SFT equations for males and females were created separately, and the 
variables were evaluated without considering the results from the other equation.  
Fourteen variables were evaluated for inclusion into the females’ SFT equation, and 16 
variables were evaluated for inclusion in the males’ SFT equation [66].  First, age, 
physical activity, seven anthropometric measurements, and skinfold thickness 
measurements (five for females and seven for males) were regressed individually 
against body density as calculated by the Bod Pod to determine if they were 
significantly associated.  They were retained to be tested for inclusion in the full model 
if their p-value was less than 0.20. Next, variables were transformed as appropriate to 
achieve a linear relationship with body density as measured by the Bod Pod.   
Next, all variables that passed the initial screening were fit into a partial least-
squares (PLS) regression using the leave-one-out validation method [68].  Given the 
collinearity of the variables of interest, a dimension-reducing method was needed to 
create a prediction model.  An iterative method was used to eliminate the variables [69].  
The variables were first evaluated for inclusion in the SFT equations based on feedback 
from the research assistants who took the measurements.  They reported that thigh 
skinfold thickness was challenging to measure due to the difficulty and sensitivity of 
accessing the mid-thigh.  Additionally, many participants were uncertain of their exact 
age.   
The remaining variables were evaluated for elimination based upon the percent 
of variation they were able to explain. The variables that explained the least amount of 
variance were evaluated first, continuing through those that explained the most amount 
of variance.  If the total variation explained by the model increased by more than 5% 
after the variable was removed, the variable remained in the model.  The number of 
components for the PLS regression was determined using both a one-sigma heuristic 
37 
 
and randomization [70].  The goal was to have the fewest variables and the fewest 
components in the final equations.   
The new equations were validated using the training set of participants.  They 
were evaluated using the same criteria used for the existing equations.  
Results 
Data was collected on 125 females and 129 males.  The mean BMI of the study 
participants was higher than the national average.  Female participants had an average 
BMI of 24.3 kg/ 2 m (Table 2.1) compared to the national average of 20.7  kg/ 2 m [45].  Male 
study participants had an average BMI of 22.1 kg/ 2m  (Table 2.1) compared to a national 
average of 19.6 kg/ 2 m [45].  Additionally, the percent of the study population classified as 
overweight or obese according to their BMI was greater than the average in urban areas 
of Ethiopia.  Approximately 45% of the female participants and 16% of the male 
participants were overweight or obese (Table 2.1).  According to the 2016 DHS, 21% of 
females and 12% of males residing in urban areas were overweight or obese [45].    
FFM from all five of the females’ BIA equations was highly correlated with the 
results of the Bod Pod.  Pearson’s r ranged from 0.86 to 0.91 (Table 2.2).  The modified 
Segal fatness-specific equation and the Luke equation did not meet any of the other 
criteria.  In addition to the correlation coefficient, the Sun et al. equation only met the 
criteria regarding the intercept of the best fit line for the plotted points.  The METS 
equation and the Zillikens equation met the criteria for the correlation coefficient, 
Student’s t test, SEE, TE, and the Kolmogorov-Smirnov test.  However, the slope of the 
best fit line for the plotted points was significantly different from one and the intercept 
from zero.  The p-values for both BIA equations for both slope and intercept were less 
than 0.01. 
38 
 
 
 
Table 2. 1: Characteristics of 
participants
39 
 
 
Table 2. 2: 
Evaluations of the validity of existing BIA equations
40 
 
For males, the METS was the only BIA equation that met all the validity criteria (Table 
2.2).  All five BIA equations met the criteria for the correlation coefficient.  The 
Pearson’s r ranged from 0.81 to 0.87.  The best fit line for the modified Segal fatness-
specific and the Sun et al. equations had slopes significantly different from 1 (p-values < 
0.01).  The best fit line for the modified Segal fatness-specific equation also has an 
intercept significantly different from 0 (p-value < 0.01).  The Zillikens equation met the 
criteria for the t test and the Kolmogorov-Smirnoff test.  The Luke equation did not 
meet any additional criteria.   
The Bland-Altman plots in Figure 2.1 compare the predictive equations to the 
Bod Pod measurements.  The y-axis is the difference in the Bod Pod and the prediction 
equation, and the x-axis is the mean of the two values.  There is no apparent bias when 
the plotted points are centered around zero and evenly distributed above and below the 
centerline throughout the mean value for FFM.  For females, the Bland-Altman plots of 
the modified Segal fatness-specific equation and the Sun et al. equation are not centered 
around zero on the y-axis (Figure 2.1a, b).  In both the modified Segal fatness-specific 
and the Luke equations, the distribution of the observations around the y-axis increased 
as FFM increased (Figure 2.1a, e).  The METS and Zillikens equations plots did not have 
any apparent bias (Figure 2.1c, d).   
For males, the Bland-Altman plot of the modified Segal fatness-specific equation 
shows the difference in the values increase as FFM increases (Figure 2.1f).  The plot of 
the Luke equation is not centered around zero (Figure 2.1j).   There is no apparent bias 
in the Bland-Altman plots of the Sun et al., METS, and Zillikins equations (Figure 2.1g-
i).  
  
41 
 
Figure 2. 1: Bland-Altman plots of ADP and BIA  
Bland-Altman plots comparing FFM measured by ADP and calculated by 
existing BIA equations.  a-e are female participants.  f- j are male 
participants.  a- Comparison of FFM from Bod Pod and the modified Segal 
fatness-specific equation in females. b- Comparison of FFM from Bod Pod 
and the Sun et al. equation in females. c - Comparison of FFM from Bod 
Pod and the METS equation in females. d - Comparison of FFM from Bod 
Pod and the Zillikens equation in females. e - Comparison of FFM from 
Bod Pod and the Luke equation in females. f- Comparison of FFM from 
Bod Pod and the modified Segal fatness-specific equation in males. g- 
Comparison of FFM from Bod Pod and the Sun et al. equation in males. h - 
Comparison of FFM from Bod Pod and the METS equation in males. i - 
Comparison of FFM from Bod Pod and the Zillikens equation in males. j- 
Comparison of FFM from Bod Pod and the Luke equation in males. 
a: Females – Segal    b: Females – Sun     
  
  
42 
 
 
Figure 2.1 (continued) 
c: Females – METS    d: Females – Zillikens 
 
e: Females – Luke 
 
f: Males – Segal     g: Males – Sun  
  
 
  
43 
 
Figure 2.1 (continued) 
h: Males – METS     i: Males - Zillikens 
 
j: Males – Luke 
 
 
  
44 
 
None of the six SFT equations met the criteria to be considered valid measures of 
FFM (Table 2.3).  For females, the Durnin and Wormsley and the Jackson, Pollack, and 
Ward equations were substantially correlated with the Bod Pod (Pearson’s r 0.87 and 
0.82, respectively).  The Sloan equation was not substantially correlated (Pearson’s r = 
0.79).  None of the equations met the Student’s t test, SEE, TE, slope, or the 
Kolmogorov-Smirnoff test criteria.  Only the Jackson, Pollack, and Ward equation met 
the criteria for the intercept (p-value = 0.08).  All three males’ SFT equations were 
substantially correlated with the Bod Pod (Pearson’s r ranged from 0.82-0.89).  None of 
the equations met any additional criteria. 
The Bland-Altman plots in Figure 2.2 show bias in all the SFT equations for both 
females and males.  None of the six equations were centered at zero on the y-axis. In 
addition, the difference in the body density decreased as the mean body density 
increased in all the plots, except in the plot for the Jackson, Pollack, and Ward equation.  
Overall, the existing skinfold thickness equations overestimated body density, which 
resulted in an underestimation of body fat. Therefore, the existing SFT equations for 
males and females are not valid for estimating body fat in Ethiopian adults.
45 
 
 
Table 2. 3: Evaluation of 
the validity of existing SFT 
equations 
46 
 
a: Females- Durnin and Wormsley  b: Females -Sloan     
  
c: Females -Jackson, Pollack, and Ward d: Males - Durnin and Wormsley 
   
e: Males – Sloan    f: Males- Jackson and Pollack 
    
 
Figure 2. 2: Bland-Altman plots of ADP and SFT 
Bland-Altman plots comparing body density measured by ADP and 
calculated by existing SFT equations.  a-c are female participants.  d- f are 
male participants.  a- Comparison of body density from Bod Pod and the 
47 
 
Durnin and Wormsley equation in females. b- Comparison of body 
density from Bod Pod and the Sloan equation in females. c - Comparison 
of body density from Bod Pod and the Jackson, Pollack, and Ward 
equation in females. d - Comparison of body density from Bod Pod and 
the Durnin and Wormsley equation in males. e - Comparison of body 
density from Bod Pod and the Sloan equation in males. f- Comparison of 
body density from Bod Pod and the Jackson and Pollack equation in 
males.  
Equation 1 is the new BIA equation for adult Ethiopian females.  All four 
variables, height, weight, resistance, and height squared divided by resistance, were 
significantly associated with body density as measured by the Bod Pod when regressed 
individually.  The final model had two variables, weight and height squared divided by 
resistance, and one component.   
(1) 
!
y = 11.34 + 0.57 !" + 0.09 × wt  
#
Where:  ht = height 
R = resistance 
wt = weight 
  
48 
 
Equation 2 is the new SFT equation for adult Ethiopian females.  All variables 
except height met the screening cutoff to be considered for the final model.  The final 
equation had two variables, suprailiac crest and subscapular SFT, and one component.  
(2) 
y = 1.105 − 0.019 × ln(α$) − 0.0147 × ln(α%) 
Where α$= suprailiac crest skinfold thickness measurement 
α% = supscapular skinfold thickness measurement 
Equation 3 is the new SFT equation for adult Ethiopian males.  When evaluated 
individually, physical activity and height did not meet the cutoff for further analysis.  
The final equation for males had two variables, triceps SFT and waist circumference, 
and two components. 
(3) 
y = 1.214 − 0.019 × ln(α$) − 0.0014 × α% 
Where α$= triceps skinfold thickness measurement 
   α% = waist	circumference 
The three new equations performed better in the validation criteria than the 
existing equations.  As seen in Table 2.4, the new females’ BIA equation and the males’ 
new SFT equation met most of Lohman’s cross-validation criteria and the KS test.  The 
Bland-Altman plots are presented in Figure 2.3.  There was no apparent bias in the 
females’ new BIA equation (Figure 2.3a).  The plot for the males’ new SFT equation 
shows that the difference in body density between the equation and the Bod Pod 
decreases as body density increases (Figure 2.3c).  For the females’ new SFT equation, 
the SEE and the total error were greater at 3.8 kg than the cutoff of 2.8 kg.  Additionally, 
49 
 
the Bland-Altman plot of the females’ new SFT equation is centered above zero (Figure 
2.3b).   
The females’ new BIA equation met all the criteria for estimating percent body 
fat in adult Ethiopians.  The females’ new SFT equation and the males’ new SFT 
equation performed much better than the existing equations but did not meet all 
criteria. 
50 
 
 
Table 2. 4: Evaluation 
of new BIA and 
SFT equations 
51 
 
 
a: Females – BIA    b: Females - SFT     
  
c: Males - SFT 
 
Figure 2. 3: Bland-Altman plots of ADP and new equations 
Bland-Altman plots comparing body composition measured by ADP and 
calculated by new prediction equations. a- Comparison of FFM from Bod 
Pod and the new BIA equation in females. b- Comparison of body density 
from Bod Pod and the new SFT equation in females. c - Comparison of 
body density from Bod Pod and the new SFT equation in males. 
 
 
  
52 
 
Discussion 
Using a population residing in Jimma City, Ethiopia, this study found that the 
METS equation is a valid BIA prediction equation for estimating FFM and percent body 
fat in Ethiopian adult males. However, none of the existing BIA equations tested for 
adult Ethiopian females were valid for estimating FFM.  Therefore, a new BIA equation 
was proposed. In addition, for both males and females, none of the existing SFT 
equations tested met the criteria to be a valid method for estimating percent body fat in 
adult Ethiopian males or females. Therefore, we proposed new, sex-specific equations 
for estimating percent body fat in Ethiopian adults using SFT.  The new SFT equations 
met more of Lohman’s cross-validation criteria than the existing ones but did not meet 
all of the criteria.   
This study has several strengths.  These are the first predictive equations for 
measuring FM and FFM in adult Ethiopians using SFT and BIA to the best of our 
knowledge. The study used a range of ages between 18 and 45, BMI, and ethnic groups 
within Ethiopia. Additionally, when validating the existing prediction equations, we 
used Lohman’s cross-validation criteria along with the Kolmogorov-Smirnov test.  
Using these procedures follows a precedent set forth by previous, similar studies.  
Finally, in developing our equation, we used a dimension reduction technique, partial 
least squares regression.  We could test potentially collinear variables in the same model 
using this technique, allowing for an objective variable selection method.  Also, using 
PLS provided more robust coefficients and more reliable equations.   
A significant limitation of this paper is the small sample size.  While the 
population size was within the existing precedence set forth by prior studies, it was still 
relatively small.  The sample size was large enough to demonstrate that the existing 
equations were not valid for this population.  However, only the females’ BIA equation 
53 
 
met all the criteria to be valid.  The new SFT equations likely lacked power, and their 
validity would improve if we used a larger sample size.  However, they still perform 
much better than the existing equations creating an improved tool for researchers and 
clinicians.  The small sample size also limited the representativeness of the study 
population.  Many ethnic groups within Ethiopia were not represented or 
underrepresented in this study sample.   
This population had a larger BMI than the urban average in Ethiopia, limiting its 
accuracy in people with lower BMIs.  Using additional participants with a more 
nationally representative distribution of BMI would strengthen this analysis.  Self-
selection bias contributed to the comparatively high BMI of the study participation, 
especially among female participants.  Studies using the Bod Pod are among the few 
opportunities available in Jimma City to measure percent body fat, motivating those 
interested in this information to participate in these studies.  This resulted in the 
recruitment of several older people concerned about their health and younger males 
interested in fitness.  The two populations balanced each other out within our male 
participant population, leading to an average BMI slightly higher than the urban 
national average.  The same was not true of the female study participants, resulting in a 
population with a higher BMI.     
This study adds to the body of evidence demonstrating that a broader range of 
tools is needed to measure body composition in different ethnic populations.  The 
skinfold thickness equations evaluated in this paper were created using populations of 
European descent which is likely why they failed most of Lohman’s cross-validation 
criteria.  However, despite using an African population, the SFT equations proposed in 
the paper are likely not valid in other regions in Africa.  African populations have 
greater genetic variability than non-African populations [71], making it less likely that a 
54 
 
few equations would be broadly applicable to the entire continent.  The potential 
limited applicability of equations across African populations was demonstrated during 
the evaluation of the existing BIA equations.  The existing BIA prediction equations 
were selected because they were either created using or validated in populations of 
African origins.  However, most were still not valid in Ethiopian adults.   
Despite the significant financial and logistical problems that limit access to 
criterion methodologies, this work should be repeated in other African contexts.  
Overweight and obesity are increasing, and BMI is insufficient to screen those most at 
risk of adverse health and economic impacts. Thus, developing a broader range of valid 
equations to assess percent fat mass is essential for addressing malnutrition throughout 
the continent.    
In Ethiopia, the new equations proposed by this study will improve researchers’ 
and clinicians’ ability to screen their participants and patients for insufficient or 
excessive adiposity compared to BMI alone and previously existing equations. In 
addition, as the prevalence of overweight and obesity continue to increase in Ethiopia, 
these equations will make it easier to identify those at risk for adverse health outcomes 
of overnutrition.  
  
55 
 
REFERENCES 
1.  Popkin, B.M.; Corvalan, C.; Grummer-Strawn, L.M. Dynamics of the Double 
Burden of Malnutrition and the Changing Nutrition Reality. Lancet 2020, 395, 65–
74, doi:10.1016/S0140-6736(19)32497-3. 
2.  Gómez, M.I.; Barrett, C.B.; Raney, T.; Pinstrup-Andersen, P.; Meerman, J.; 
Croppenstedt, A.; Carisma, B.; Thompson, B. Post-Green Revolution Food Systems 
and the Triple Burden of Malnutrition. Food Policy 2013, 42, 129–138, 
doi:10.1016/j.foodpol.2013.06.009. 
3.  Ruel, M.T.; Alderman, H. Nutrition-Sensitive Interventions and Programmes: How 
Can They Help to Accelerate Progress in Improving Maternal and Child Nutrition? 
Lancet 2013, 382, 536–551, doi:10.1016/S0140-6736(13)60843-0. 
4.  Popkin, B.M.; Gordon-Larsen, P. The Nutrition Transition: Worldwide Obesity 
Dynamics and Their Determinants. Int J Obes 2004, 28, S2–S9, 
doi:10.1038/sj.ijo.0802804. 
5.  Moges, B.; Amare, B.; Fantahun, B.; Kassu, A. High Prevalence of Overweight, 
Obesity, and Hypertension with Increased Risk to Cardiovascular Disorders 
among Adults in Northwest Ethiopia: A Cross Sectional Study. BMC Cardiovasc 
Disord 2014, 14, 155, doi:10.1186/1471-2261-14-155. 
6.  Wells, J.C.; Sawaya, A.L.; Wibaek, R.; Mwangome, M.; Poullas, M.S.; Yajnik, C.S.; 
Demaio, A. The Double Burden of Malnutrition: Aetiological Pathways and 
Consequences for Health. Lancet 2020, 395, 75–88, doi:10.1016/S0140-
6736(19)32472-9. 
7.  Nelms, M. Nutrition Therapy and Pathophysiology; Fourth edition.; Cengage: 
Boston, MA, 2020; ISBN 0-357-04171-2. 
8.  Nugent, R.; Levin, C.; Hale, J.; Hutchinson, B. Economic Effects of the Double 
Burden of Malnutrition. Lancet 2020, 395, 156–164, doi:10.1016/S0140-
6736(19)32473-0. 
9.  Victora, C.G.; Adair, L.; Fall, C.; Hallal, P.C.; Martorell, R.; Richter, L.; Sachdev, 
H.S.; Maternal and Child Undernutrition Study Group Maternal and Child 
Undernutrition: Consequences for Adult Health and Human Capital. Lancet 2008, 
371, 340–357, doi:10.1016/S0140-6736(07)61692-4. 
56 
 
10.  Alderman, H.; Behrman, J.R.; Hoddinott, J. Economic and Nutritional Analyses 
Offer Substantial Synergies for Understanding Human Nutrition. J Nutr 2007, 137, 
537–544, doi:10.1093/jn/137.3.537. 
11.  Hoddinott, J.; Behrman, J.R.; Maluccio, J.A.; Melgar, P.; Quisumbing, A.R.; 
Ramirez-Zea, M.; Stein, A.D.; Yount, K.M.; Martorell, R. Adult Consequences of 
Growth Failure in Early Childhood. Am J Clin Nutr 2013, 98, 1170–1178, 
doi:10.3945/ajcn.113.064584. 
12.  Gallagher, D.; Visser, M.; Sepúlveda, D.; Pierson, R.N.; Harris, T.; Heymsfield, S.B. 
How Useful Is Body Mass Index for Comparison of Body Fatness across Age, Sex, 
and Ethnic Groups? Am J Epidemiol 1996, 143, 228–239, 
doi:10.1093/oxfordjournals.aje.a008733. 
13.  WHO Expert Consultation Appropriate Body-Mass Index for Asian Populations 
and Its Implications for Policy and Intervention Strategies. Lancet 2004, 363, 157–
163, doi:10.1016/S0140-6736(03)15268-3. 
14.  Jean, N.; Somers, V.K.; Sochor, O.; Medina-Inojosa, J.; Llano, E.M.; Lopez-Jimenez, 
F. Normal-Weight Obesity: Implications for Cardiovascular Health. Curr 
Atheroscler Rep 2014, 16, 464, doi:10.1007/s11883-014-0464-7. 
15.  Liu, X.; Wu, Q.; Yan, G.; Duan, J.; Chen, Z.; Yang, P.; Bragazzi, N.L.; Lu, Y.; Yuan, 
H. Cardiometabolic Index: A New Tool for Screening the Metabolically Obese 
Normal Weight Phenotype. J Endocrinol Invest 2021, 44, 1253–1261, 
doi:10.1007/s40618-020-01417-z. 
16.  Marques-Vidal, P.; Pécoud, A.; Hayoz, D.; Paccaud, F.; Mooser, V.; Waeber, G.; 
Vollenweider, P. Normal Weight Obesity: Relationship with Lipids, Glycaemic 
Status, Liver Enzymes and Inflammation. Nutr Metab Cardiovasc Dis 2010, 20, 
669–675, doi:10.1016/j.numecd.2009.06.001. 
17.  Blüher, M. Metabolically Healthy Obesity. Endocrine Reviews 2020, 41, 405–420, 
doi:10.1210/endrev/bnaa004. 
18.  Deurenberg, P.; Yap, M.; Staveren, W.A. van. Body Mass Index and Percent Body 
Fat: A Meta-Analysis among Different Ethnic Groups. Int J Obes Relat Metab 
Disord 1998, 22, 1164–1171. 
19.  Sinaga, M.; Worku, M.; Yemane, T.; Tegene, E.; Wakayo, T.; Girma, T.; Lindstrom, 
D.; Belachew, T. Optimal Cut-off for Obesity and Markers of Metabolic Syndrome 
57 
 
for Ethiopian Adults. Nutrition Journal 2018, 17, 109, doi:10.1186/s12937-018-0416-
0. 
20.  Sinaga, M.; Yemane, T.; Tegene, E.; Lidstrom, D.; Belachew, T. Performance of 
Newly Developed Body Mass Index Cut-off for Diagnosing Obesity among 
Ethiopian Adults. J Physiol Anthropol 2019, 38, 14, doi:10.1186/s40101-019-0205-2. 
21.  Heymsfield, Steven B.; Lohman, Timothy G.; d; Going, Scott B. Human Body 
Composition; 2nd ed.; Human Kinetics: Champaign, IL, 2005; ISBN 0736046550. 
22.  Mora, S.; Pessin, J.E. An Adipocentric View of Signaling and Intracellular 
Trafficking. Diabetes Metab Res Rev 2002, 18, 345–356, doi:10.1002/dmrr.321. 
23.  Trayhurn, P. Endocrine and Signalling Role of Adipose Tissue: New Perspectives 
on Fat. Acta Physiol Scand 2005, 184, 285–293, doi:10.1111/j.1365-
201X.2005.01468.x. 
24.  Bouillanne, O.; Dupont-Belmont, C.; Hay, P.; Hamon-Vilcot, B.; Cynober, L.; 
Aussel, C. Fat Mass Protects Hospitalized Elderly Persons against Morbidity and 
Mortality. Am J Clin Nutr 2009, 90, 505–510, doi:10.3945/ajcn.2009.27819. 
25.  Trayhurn, P.; Beattie, J.H. Physiological Role of Adipose Tissue: White Adipose 
Tissue as an Endocrine and Secretory Organ. Proceedings of the Nutrition Society 
2001, 60, 329–339, doi:10.1079/PNS200194. 
26.  Sedlmeier, A.M.; Baumeister, S.E.; Weber, A.; Fischer, B.; Thorand, B.; Ittermann, 
T.; Dörr, M.; Felix, S.B.; Völzke, H.; Peters, A.; et al. Relation of Body Fat Mass and 
Fat-Free Mass to Total Mortality: Results from 7 Prospective Cohort Studies. Am J 
Clin Nutr 2021, 113, 639–646, doi:10.1093/ajcn/nqaa339. 
27.  Szymańska, E.; Bouwman, J.; Strassburg, K.; Vervoort, J.; Kangas, A.J.; Soininen, P.; 
Ala-Korpela, M.; Westerhuis, J.; van Duynhoven, J.P.M.; Mela, D.J.; et al. Gender-
Dependent Associations of Metabolite Profiles and Body Fat Distribution in a 
Healthy Population with Central Obesity: Towards Metabolomics Diagnostics. 
OMICS: A Journal of Integrative Biology 2012, 16, 652–667, 
doi:10.1089/omi.2012.0062. 
28.  Rosen, E.D.; Spiegelman, B.M. What We Talk about When We Talk about Fat. Cell 
2014, 156, 20–44, doi:10.1016/j.cell.2013.12.012. 
58 
 
29.  Long, A.E.; Prewitt, T.E.; Kaufman, J.S.; Rotimi, C.N.; Cooper, R.S.; McGee, D.L. 
Weight–Height Relationships among Eight Populations of West African Origin: 
The Case against Constant BMI Standards. Int J Obes 1998, 22, 842–846, 
doi:10.1038/sj.ijo.0800670. 
30.  Lee, S.Y.; Gallagher, D. Assessment Methods in Human Body Composition. Curr 
Opin Clin Nutr Metab Care 2008, 11, 566–572, doi:10.1097/MCO.0b013e32830b5f23. 
31.  Wagner, D.R.; Heyward, V.H. Measures of Body Composition in Blacks and 
Whites: A Comparative Review. Am J Clin Nutr 2000, 71, 1392–1402, 
doi:10.1093/ajcn/71.6.1392. 
32.  Dioum, A.; Gartner, A.; Cissé, A.S.; Delpeuch, F.; Maire, B.; Wade, S.; Schutz, Y. 
Validity of Impedance-Based Equations for the Prediction of Total Body Water as 
Measured by Deuterium Dilution in African Women. Am J Clin Nutr 2005, 81, 597–
604, doi:10.1093/ajcn/81.3.597. 
33.  Stolarczyk, L.M.; Heyward, V.H.; Van Loan, M.D.; Hicks, V.L.; Wilson, W.L.; 
Reano, L.M. The Fatness-Specific Bioelectrical Impedance Analysis Equations of 
Segal et al: Are They Generalizable and Practical? Am J Clin Nutr 1997, 66, 8–17, 
doi:10.1093/ajcn/66.1.8. 
34.  McArdle, W.D. Exercise Physiology: Nutrition, Energy, and Human Performance; 
Eighth edition.; Wolters Kluwer Health/Lippincott Williams & Wilkins: Baltimore, 
2015; ISBN 978-1-4511-9155-4. 
35.  Sloan, A.W. Estimation of Body Fat in Young Men. J Appl Physiol 1967, 23, 311–
315, doi:10.1152/jappl.1967.23.3.311. 
36.  Jackson, A.S.; Pollock, M.L. Generalized Equations for Predicting Body Density of 
Men. Br J Nutr 1978, 40, 497–504, doi:10.1079/BJN19780152. 
37.  Jackson, A.S.; Pollock, M.L.; Ward, A. Generalized Equations for Predicting Body 
Density of Women. Med Sci Sports 1980, 12, 175–182. 
38.  Durnin, J.V.G.A.; Womersley, J. Body Fat Assessed from Total Body Density and 
Its Estimation from Skinfold Thickness: Measurements on 481 Men and Women 
Aged from 16 to 72 Years. Br J Nutr 1974, 32, 77–97, doi:10.1079/BJN19740060. 
59 
 
39.  Zillikens, M.C.; Conway, J.M. Anthropometry in Blacks: Applicability of 
Generalized Skinfold Equations and Differences in Fat Patterning between Blacks 
and Whites. Am J Clin Nutr 1990, 52, 45–51, doi:10.1093/ajcn/52.1.45. 
40.  Brandon, L.J. Comparison of Existing Skinfold Equations for Estimating Body Fat 
in African American and White Women. Am J Clin Nutr 1998, 67, 1155–1161, 
doi:10.1093/ajcn/67.6.1155. 
41.  Wong, W.W.; Stuff, J.E.; Butte, N.F.; Smith, E.O.; Ellis, K.J. Estimating Body Fat in 
African American and White Adolescent Girls: A Comparison of Skinfold-
Thickness Equations with a 4-Compartment Criterion Model. Am J Clin Nutr 2000, 
72, 348–354, doi:10.1093/ajcn/72.2.348. 
42.  Sun, S.S.; Chumlea, W.C.; Heymsfield, S.B.; Lukaski, H.C.; Schoeller, D.; Friedl, K.; 
Kuczmarski, R.J.; Flegal, K.M.; Johnson, C.L.; Hubbard, V.S. Development of 
Bioelectrical Impedance Analysis Prediction Equations for Body Composition with 
the Use of a Multicomponent Model for Use in Epidemiologic Surveys. Am J Clin 
Nutr 2003, 77, 331–340, doi:10.1093/ajcn/77.2.331. 
43.  Wagner, D.R.; Heyward, V.H.; Kocina, P.S.; Stolarczyk, L.M.; Wilson, W.L. 
Predictive Accuracy of BIA Equations for Estimating Fat-Free Mass of Black Men. 
Med Sci Sports 1997, 29, 969–974. 
44.  Luke, A.; Bovet, P.; Forrester, T.E.; Lambert, E.V.; Plange-Rhule, J.; Dugas, L.R.; 
Durazo-Arvizu, R.A.; Kroff, J.; Richie, W.N.; Schoeller, D.A. Prediction of Fat-Free 
Mass Using Bioelectrical Impedance Analysis in Young Adults from Five 
Populations of African Origin. Eur J Clin Nutr 2013, 67, 956–960, 
doi:10.1038/ejcn.2013.123. 
45.  Central Statistical Agency (CSA) Ethiopia and ICF Ethiopia Demographic and 
Health Survey 2016. 
46.  Admassu, B.; Wells, J.C.K.; Girma, T.; Andersen, G.S.; Owino, V.; Belachew, T.; 
Michaelsen, K.F.; Abera, M.; Wibaek, R.; Friis, H.; et al. Body Composition at Birth 
and Height at 2 Years: A Prospective Cohort Study among Children in Jimma, 
Ethiopia. Pediatr Res 2017, 82, 209–214, doi:10.1038/pr.2017.59. 
47.  Andersen, G.S.; Girma, T.; Wells, J.C.K.; Kæstel, P.; Michaelsen, K.F.; Friis, H. Fat 
and Fat-Free Mass at Birth: Air Displacement Plethysmography Measurements on 
350 Ethiopian Newborns. Pediatr Res 2011, 70, 501–506, 
doi:10.1203/PDR.0b013e31822d7470. 
60 
 
48.  Andersen, G.S.; Girma, T.; Wells, J.C.; Kæstel, P.; Leventi, M.; Hother, A.-L.; 
Michaelsen, K.F.; Friis, H. Body Composition from Birth to 6 Mo of Age in 
Ethiopian Infants: Reference Data Obtained by Air-Displacement 
Plethysmography. Am J Clin Nutr 2013, 98, 885–894, doi:10.3945/ajcn.113.063032. 
49.  Hegelund, M.H.; Wells, J.C.; Girma, T.; Faurholt-Jepsen, D.; Zerfu, D.; Christensen, 
D.L.; Friis, H.; Olsen, M.F. Validation of Bioelectrical Impedance Analysis in 
Ethiopian Adults with HIV. J Nutr Sci 2017, 6, e62, doi:10.1017/jns.2017.67. 
50.  Deurenberg, P.; Wolde-Gebriel, Z.; Schouten, F.J.M. Validity of Predicted Total 
Body Water and Extracellular Water Using Multifrequency Bioelectrical Impedance 
in an Ethiopian Population. Ann Nutr Metab 1995, 39, 234–241, 
doi:10.1159/000177868. 
51.  Ribeiro, S.M.L.; Kehayias, J.J. Sarcopenia and the Analysis of Body Composition. 
Adv Nutr 2014, 5, 260–267, doi:10.3945/an.113.005256. 
52.  Morley, J.E. Anorexia, Body Composition, and Ageing. Current Opinion in Clinical 
Nutrition & Metabolic Care 2001, 4, 9–13. 
53.  Bosaeus, M.; Andersson-Hall, U.; Andersson, L.; Karlsson, T.; Ellegård, L.; 
Holmäng, A. Body Composition during Pregnancy: Longitudinal Changes and 
Method Comparisons. Reprod. Sci. 2020, 27, 1477–1489, doi:10.1007/s43032-020-
00141-6. 
54.  Butte, N.; Hopkinson, J.; Ellis, K.; Wong, W.; Smith, E. Changes in Fat-Free Mass 
and Fat Mass in Postpartum Women: A Comparison of Body Composition Models. 
Int J Obes 1997, 21, 874–880, doi:10.1038/sj.ijo.0800482. 
55.  Leiter, L.A.; Lukaski, H.C.; Kenny, D.J.; Barnie, A.; Camelon, K.; Ferguson, R.S.; 
MacLean, S.; Simkins, S.; Zinman, B.; Cleary, P.A. The Use of Bioelectrical 
Impedance Analysis (BIA) to Estimate Body Composition in the Diabetes Control 
and Complications Trial (DCCT). Int J Obes 1994, 18, 829–835. 
56.  Taing, K.Y.; Farkouh, M.E.; Moineddin, R.; Tu, J.V.; Jha, P. Comparative 
Associations between Anthropometric and Bioelectric Impedance Analysis Derived 
Adiposity Measures with Blood Pressure and Hypertension in India: A Cross-
Sectional Analysis. BMC Obes 2017, 4, 37, doi:10.1186/s40608-017-0173-1. 
57.  World Health Organization WHO STEPS Surveillance Manual: The WHO 
STEPwise Approach to Noncommunicable Disease Risk Factor Surveillance. 
61 
 
58.  Bia 450 Bioimpedance Analyzer User’s Guide Third Edition; 3rd ed.; Biodynamics 
Corporation: Seattle, Washington. 
59.  NHANES III (1988-1994) - Anthropometric Procedure Videos Available online: 
https://wwwn.cdc.gov/nchs/nhanes/nhanes3/anthropometricvideos.aspx 
(accessed on 26 August 2021). 
60.  Segal, K.R.; Van Loan, M.; Fitzgerald, P.I.; Hodgdon, J.A.; Van Itallie, T.B. Lean 
Body Mass Estimation by Bioelectrical Impedance Analysis: A Four-Site Cross-
Validation Study. The American Journal of Clinical Nutrition 1988, 47, 7–14, 
doi:10.1093/ajcn/47.1.7. 
61.  Courville, A.B.; Yang, S.B.; Andrus, S.; Hayat, N.; Kuemmerle, A.; Leahy, E.; Briker, 
S.; Zambell, K.; Chung, S.; Sumner, A.E. Body Adiposity Measured by Bioelectrical 
Impedance Is an Alternative to Dual-Energy X-Ray Absorptiometry in Black 
Africans: The Africans in America Study. Nutrition 2020, 74, 110733, 
doi:10.1016/j.nut.2020.110733. 
62.  Luke, A.; Durazo-Arvizu, R.; Rotimi, C.; Prewitt, T.E.; Forrester, T.; Wilks, R.; 
Ogunbiyi, O.J.; Schoeller, D.A.; McGee, D.; Cooper, R.S. Relation between Body 
Mass Index and Body Fat in Black Population Samples from Nigeria, Jamaica, and 
the United States. Am J Epidemiol 1997, 145, 620–628, 
doi:10.1093/oxfordjournals.aje.a009159. 
63.  Zillikens, M.C.; Conway, J.M. Estimation of Total Body Water by Bioelectrical 
Impedance Analysis in Blacks. Am J Hum Biol 1991, 3, 25–32, 
doi:10.1002/ajhb.1310030106. 
64.  Sloan, A.W.; Burt, J.J.; Blyth, C.S. Estimation of Body Fat in Young Women. Journal 
of Applied Physiology 1962, 17, 967–970, doi:10.1152/jappl.1962.17.6.967. 
65.  Schutte, J.E.; Townsend, E.J.; Hugg, J.; Shoup, R.F.; Malina, R.M.; Blomqvist, C.G. 
Density of Lean Body Mass Is Greater in Blacks than in Whites. J Appl Physiol 
Respir Environ Exerc Physiol 1984, 56, 1647–1649, doi:10.1152/jappl.1984.56.6.1647. 
66.  Lohman, T.G. Advances in Body Composition Assessment; Human Kinetics 
Publishers: Champaign, Ill., 1992; ISBN 0-87322-327-6. 
67.  Bland, J.M.; Altman, D.G. Statistical Methods for Assessing Agreement between 
Two Methods of Clinical Measurement. Lancet 1986, 1, 307–310. 
62 
 
68.  Mevik, B.-H.; Wehrens, R. The Pls Package: Principal Component and Partial Least 
Squares Regression in R. Journal of Statistical Software 2007, 18, 1–23, 
doi:10.18637/jss.v018.i02. 
69.  Mehmood, T.; Sæbø, S.; Liland, K.H. Comparison of Variable Selection Methods in 
Partial Least Squares Regression. J Chemom 2020, 34, e3226, doi:10.1002/cem.3226. 
70.  Mevik, B.-H. Introduction to the Pls Package. 24. 
71.  Campbell, M.C.; Tishkoff, S.A. AFRICAN GENETIC DIVERSITY: Implications for 
Human Demographic History, Modern Human Origins, and Complex Disease 
Mapping. Annu Rev Genomics Hum Genet 2008, 9, 403–433, 
doi:10.1146/annurev.genom.9.081307.164258. 
63 
 
CHAPTER 3 
VALIDITY OF THE GLOBAL PHYSICAL ACTIVITY QUESTIONAIRE (GPAQ) AND 
24-HOUR RECALL OF TIME USE AND PERCEIVED EXERTION IN FEMALES IN 
RURAL TIGRAY, ETHIOPIA 
Abstract   
Nutritional status is the result of energy intake and physical activity.  There are no 
validated survey tools for estimating physical activity in rural, highland Ethiopia, 
making it difficult to create adequate programs to address undernutrition.  This study 
used accelerometry to validate two questionnaires, the global physical activity 
questionnaire (GPAQ) and the 24-hour recall of time use and perceived physical activity 
questionnaire in 180 females living in rural Tigray, Ethiopia.  Participants were females 
between 18 and 45 years who had previously participated in an impact evaluation of a 
public works safety net.  They wore an accelerometer for eight days and responded to 
perceived exertion questionnaires twice.  Data was collected on 89 females during the 
short rainy period and 91 females during the main rainy season.  A survey method was 
considered valid if the proportion of time spent in moderate and vigorous levels of 
physical activity (MVPA) was acceptably correlated (r > 0.40) to the proportion of time 
spent in MVPA recorded by accelerometry. The GPAQ had high reliability, but the 
overall validity was poorer than accelerometry.  The proportion of time spent in MVPA 
according to the accelerometer was associated with discordance between GPAQ and 
accelerometry.  Moderate and vigorous physical activity, as measured by the 24-hour 
recall, had a fair agreement with accelerometry.  The agreement increased to 
moderate/acceptable when adjusted for season and body mass index (BMI).  We 
conclude that the 24-hour recall of perceived energy expenditure is a valid tool to 
64 
 
estimate physical activity in females living in rural highland Ethiopia. It can be used in 
future research to understand the physical activity demands of living in rural highland 
Ethiopia, enabling more targeted programs to address undernutrition. 
 
  
65 
 
Introduction 
Given the multiple challenges, addressing nutritional issues in low-income 
countries is complex. According to data from the 2016 Demographic and Health Survey 
(DHS), in Ethiopia, the focus of this paper, 22% of adult females are underweight [1].  
Underweight females and their children are at a greater risk of adverse health outcomes 
and death.  A low pre-pregnancy BMI is associated with fetal growth restrictions, the 
cause of 12% of child deaths worldwide; stunting in childhood; obesity and non-
communicable diseases in adulthood; and developmental deficits that can negatively 
affect school performance and economic productivity [2–5].  
Efforts to target undernutrition are primarily focused on increasing energy 
intake, improving dietary diversity, and addressing poverty [10].  One safety net 
program that addresses poverty in many low-and middle-income countries offers 
individuals food or cash payments for labor on public works projects. An example of 
this type of safety net is Ethiopia’s Productive Safety Net Programme (PSNP) [11]. 
However, these interventions require strenuous physical effort.  Even if household 
members are not directly participating in the program, they may need to compensate 
for other household members’ efforts. While the impacts of these public works 
programs on food security and caloric acquisition are increasingly well-characterized 
[11,12], their impact on overall energy intake is not because there is a lack of 
understanding about physical activity [13].   
The efforts to understand and improve nutritional status primarily focus on 
dietary intake because the survey tools are readily available to researchers.  Examples of 
survey tools include food frequency questionnaires, dietary recalls, and dietary records.  
Comparatively, similar tools for measuring physical activity are lacking.  Two examples 
of physical activity assessment tools include the Global Physical Activity Questionnaire 
66 
 
(GPAQ) [14] and the International Physical Activity Questionnaire [15] which are 
commonly used to measure sedentary behavior in populations with high rates of 
obesity [16,17] and are more accurate in European and Asian settings than in physically 
active populations in sub-Saharan Africa [18].  These questionnaires are the physical 
activity equivalent to seven-day dietary recall.  There has been less effort to validate 
questionnaires for physical activity with more detail and over shorter periods that 
would be the equivalent of a 24-hour recall. 
Physical activity survey tools are challenging to validate.  Criterion methods, 
such as indirect and direct calorimetry, are expensive and have limited applicability in 
resource-constrained settings [19,20].  Indirect calorimetry either uses bulky machinery, 
which disrupts the activities of free-living adults, or doubly labeled water, which 
requires access to a laboratory and trained technician  [21,22]. Direct calorimetry is 
confined to a laboratory and cannot measure energy expenditure in free-living persons.   
Fortunately, significant insight can still be gained by quantifying the proportion 
of time spent at an increased level of physical activity provides valuable information to 
enhance research and nutrition programs.  Advances in technology have resulted in 
better methods to objectively measure physical activity exertion levels [23]. For 
example, triaxial accelerometers record acceleration within the vertical, anteroposterior, 
and mediolateral planes in free-living persons [19,24,25] using unobtrusive devices that 
do not hamper physical activity.  
This study tested the reliability and validity of two survey methods, the GPAQ 
and a 24-hour recall of time use and perceived exertion, to estimate time spent in 
sedentary activities, moderate exertion activities, vigorous exertion activities, and 
moderate and vigorous physical activity (MVPA) using accelerometry in adult 
Ethiopian females living in rural highland regions of Ethiopia.   
67 
 
Material and Methods 
Participants and study setting 
This study took place in Ethiopia, a low-income country in East Africa. Ethiopia 
has mountainous highland regions in the center and north of the country and flat 
lowland regions east and south. Within Ethiopia, data were collected in the northern 
region of Tigray, which is landlocked and mountainous. Livelihoods are based around 
rainfed agriculture. A short rainy season occurs between March and May, and the main 
rainy season occurs between June and October.  During the short rainy season, the 
PSNP is operational.  During the main rainy season, more than 90 percent of total crops 
in the country are produced and then harvested between October and December [26]. 
As a result, food security increases following the harvest during the short rainy season 
[27]. 
Study participants were females between the ages of 18 and 45 with a child 
between the ages of 6 and 23 months during the first round of data collection during the 
short rainy season.  They were included in the study if they had previously participated 
in an impact evaluation of the PSNP, were not pregnant, and were not planning on 
leaving the area during data collection. In addition, the impact evaluation recruited 
households, which were eligible if the household members included a child (index 
child) aged 6-23 months and its mother (mother of the index child) [26].  The impact 
evaluation had one round of data collection during the short rainy season and the 
second round of data collection during the main rainy season. 
Sample size 
Participants in this study were the mothers of the index children. Data collection 
occurred only in the Central, Eastern, and South zones, but not the North Western zone, 
68 
 
because of logistical and safety constraints.  Within each zone, four counties where the 
earlier impact evaluation took place were randomly selected. Thus, twelve out of 
twenty-one potential counties were included in the study.  Within each county, villages 
were screened to ensure that they contained at least two households that received 
benefits from the public works component of the PSNP (beneficiaries) and two 
households that received no benefits (non-beneficiaries); villages that did not meet this 
criterion were excluded.  From the remaining villages, two were randomly selected per 
county.  Within villages, two PSNP and two non-PSNP households were randomly 
selected; within these, the mother of the index child was screened for eligibility to have 
two beneficiaries and two non-beneficiaries per village.   
When unable to recruit the intended number of beneficiaries due to pregnancy, 
travel, or participants declining to participate, additional non-beneficiaries were 
recruited.  There were four participants per village, yielding a target sample of 96 
participants (96 = 4 participants per village x 2 villages per county x 4 counties per zone 
x 3 zones).  A study size of 96 was selected based on precedents set by earlier studies 
which had 45 participants [16], 78 participants [17], and several hundred participants 
[19], the need to account for loss to follow up, and the logistics associated with fielding 
an intensive data collection effort across multiple remote locations.  
Data collection 
Data collection took place over two four-week periods.  The first round (PSNP 
round), April-May 2019, coincided with the small rainy season when minimal 
agricultural activities took place, and the PSNP was operational. The second round 
(agricultural round), September-October 2019, took place during the main rainy season 
when households were engaged in farming activities such as preparing the land, 
planting, and weeding, and the PSNP was non-operational.  Efforts were made to 
69 
 
include the same females in both rounds of data collection.  However, this was not 
always feasible due to females being away from their homes during the agricultural 
round. Therefore, additional females from the village were recruited to maintain the 
desired results sample size.  Seventy-three participants participated in both the PSNP 
and the agricultural round of data collection, twenty-two participants participated in 
only the PSNP round, and twenty-three participants participated in only the 
agricultural round. 
On the first day of each data-collection period, prospective participants were 
screened for eligibility, enrolled in the study, and given an accelerometer and heart rate 
monitor to wear.   However, the heart rate monitors were too large for the participants 
and did not record sufficient data. Therefore, the analysis did not include heart rate 
data; physical activity was calculated using only accelerometry data.  The participants 
wore the accelerometer wore around their waist for eight successive days, which 
permitted the collection of seven full days of data on physical activity.  They were 
instructed to always wear the accelerometer except when bathing or swimming.  They 
were encouraged to wear the devices while sleeping but were told to remove them if 
uncomfortable.   
During the week they wore the accelerometers, participants were visited six 
times.  At each visit, the participants were checked to ensure they were wearing the 
accelerometers properly, and the battery life of the accelerometers was checked.  On the 
second visit, participants responded to the GPAQ and 24-hour recall for the first time.  
They also provided information on their age, education, and whether their partner 
currently resided in the same household.  Their weight was measured twice, the 
average of the two weights was used for analysis (Seca 874, Seca, Hamburg, Germany).  
Additional demographic, village information, and height were recorded during the 
70 
 
initial impact evaluation. Participants responded to the GPAQ and the 24-hour recall 
survey questions a second time on the final day of data collection when the 
accelerometers were collected from the participants.   
Physical activity 
Physical exertion is divided into four intensity levels: sedentary, light, moderate, 
and vigorous [20].  Sedentary activity includes activities that require no increase in 
effort and include activities such as sitting and reclining.  Light-intensity activities are 
more active than sedentary activities but do not require increases in breathing or heart 
rate.  Moderate-intensity activities are activities that require moderate effort and cause 
small increases in breathing and heart rate. Finally, vigorous-intensity activities require 
significant effort and cause large increases in breathing and heart rate [28].  The primary 
outcome of interest was the percent of time spent in moderate and vigorous physical 
activity (MVPA) as measured by accelerometry, the GPAQ, and the 24-hour recall of 
time use and perceived energy exertion.  Additional outcomes included the percent of 
time spent in sedentary/light physical exertion, moderate exertion alone, and vigorous 
exertion alone.   
The reference measure used for validating physical activity was time spent in 
different levels of physical exertion as measured by triaxial accelerometers.  
Accelerometers collect data on movement using counts, which determine the level of 
physical activity of the participant.  This study used the Actigraph wGT3X-BT 
accelerometer (Actigraph Corp, Florida USA).  The Actigraph wGT3X-BT is a reliable 
measure of physical activity [26], and they have been used in various populations in 
various settings, including validating the GPAQ in the United Kingdom [30], France 
[31], and India [32].  
71 
 
The GPAQ measures physical activity “in a typical week,” as recalled by survey 
participants in a face-to-face interview. The World Health Organization (WHO) 
developed it as the global standard for monitoring physical activity [28] and is 
frequently used to assess the proportion of a population that meets the minimum 
recommended time spent in moderate or vigorous activity in an average week.  The 
validity of GPAQ to measure physical activity has been assessed in the United Kingdom 
[30], Bangladesh [33],  and New York City [34].  Previously, pedometers were used in a 
validation study of the GPAQ in Ethiopia [14]. The GPAQ is broader in scope than the 
24-hour recall of physical exertion.  It collects information on whether (yes/no) a 
participant spent time in either moderate or vigorous activity across three domains: 
work and home, travel to and from places, and recreational activities.  The respondents 
then report the frequency (days/week) and duration (minutes/day) of time spent at 
each level of exertion in work and home and recreational activities. Time spent 
traveling to and from places is assumed to be moderate-intensity [30].     
The 24-hour recall measures time use in minutes and perceived energy exertion 
from the previous day as recalled by survey participants in a face-to-face interview.  
Perceived energy exertion was measured in minutes on a 4-point scale that included 
sedentary, light, moderate, and vigorous.  The 24-hour recall questionnaire was adapted 
from the time allocation module of the Women’s Empowerment in Agriculture Index 
[33], which was validated and used in Ethiopia and other countries, including India, 
Bangladesh, and Malawi. [36–38].  The 24-hour recall questionnaire was used to collect 
information on the participants’ primary and secondary activities and their perceived 
energy exertion during 30-minute epochs starting an hour before sunrise and 
concluding an hour after sunset.  Participants listed all the activities they participated in 
during the previous day, the activities were then placed in chronological order, and the 
72 
 
participants then estimated the amount of time that each activity took.  When possible, 
participants validated the time of an activity using scheduled or other notable events 
such as meals or religious activities.  For each activity, participants reported activity 
intensity as sedentary, light, moderate, or vigorous.  The 24-hour recall captured more 
detail over a shorter time frame than the GPAQ.  When participants recorded both a 
primary and secondary activity, each activity and corresponding exertion level were 
allocated 15 minutes.  If a participant reported only one activity during a 30-minute 
epoch, that activity was allocated all 30 minutes.   Previous studies used the 
Compendium of Physical Activity to determine the metabolic equivalents (METs) 
associated with specific work  [39,40].  However, self-reported exertion level was 
determined to be preferable due to the nature of manual labor in Ethiopia. Other 
studies used the approach in Australia [41], the United States [42], and China [43]. 
Data processing 
Accelerometers recorded data in 1-second epochs with sampling frequencies at 
30 Hz [44].  Data were processed, cleaned, and analyzed using ActiLife 6 software 
(Actigraph, Florida, USA).  During processing, the algorithm by Choi et al. was used to 
define non-wear time during waking hours [45].  The raw data were reintegrated into 
10-second epochs for the refined Crouter algorithm  [29,46–49].  Next, the data were 
converted into summary scores for each hour of the day, including minutes spent at 
different exertion levels.  Data were checked for outliers and valid observations.  A day 
was considered valid if it had 12 hours of recorded wear time.  A week was valid if it 
had five days of valid wear time. For use in analysis of the GPAQ, the participant 
needed a valid week of accelerometry wear time.  For the 24-hour recall data analysis, 
the accelerometry data were restricted to the same time frame as the 24-hour recall and 
considered valid if there were at least 12 hours of accelerometry data for the day of 
73 
 
recall.  The outcome of interest was the proportion of time spent at each level of 
exertion.  The numerator was the number of minutes spent at each level of exertion.  
The denominator was the number of minutes recorded during either the week or the 
day. 
Stata version 14.2 (StataCorp, College Station, Texas) was used for the remainder 
of the data analysis, except the breakpoint analysis, which used R 4.0.4 (The R 
Foundation for Statistical Computing Platform) and RStudio V1.44.1106 (RStudio, PBC).   
The GPAQ was analyzed according to the analysis guide [28].  Observations 
were excluded if a participant reported implausible values, such as spending more than 
16 hours a day or more than seven days a week in an activity.  No observations had 
implausible values.  The outcomes of interest for the reliability analysis were the 
dichotomous responses about whether a participant spent time in an activity and the 
total minutes reported spent in each activity.  The outcome of interest for the validity 
analysis was the proportion of time spent at each level of exertion.  The numerator was 
the number of minutes spent at each level of exertion throughout the week.  The 
denominator was the number of minutes in the week. 
Data from the 24-hour recall were checked for implausible amounts of time spent 
in high levels of exertion. Then, the minutes spent in each level of physical exertion 
were summed for that day.  The outcome of interest was the proportion of time spent at 
each level of exertion.  The numerator was the number of minutes spent at each level of 
exertion, and the denominator was the number of minutes captured by the 24-hour 
recall, 14 hours.   
GPAQ analysis 
74 
 
The reliability of the GPAQ was tested using the k statistic and percent 
agreement for dichotomous responses, whether a participant spent time in an activity, 
[50] and the Spearman Rank Correlation Coefficient (r) for continuous responses, how 
much time a participant spent in an activity [51].  For the GPAQ to be reliable, the 
agreement for both dichotomous and continuous responses needed to be 
moderate/acceptable (r > 0.41) [14].  The reliability of the GPAQ was tested for both 
rounds combined and stratified by round.  Reliability was not calculated for 24-hour 
recall because each response corresponded to a different 24-hour period. Thus, a lack of 
agreement in a participant’s responses may reflect her being engaged in different 
activities during the two recall days rather than her not reliably recalling what she did 
the previous day.   
The validity of the GPAQ was determined by comparing the proportion of time 
spent in each of the four areas of interest, light/sedentary activity, moderate activity, 
vigorous activity, and MVPA, to the values measured by accelerometry.  Values were 
compared using the Pearson correlation coefficient and Spearman Rank correlation 
coefficient.  The cutoffs for the Spearman correlation for both reliability and validity 
were as follows: a correlation coefficient (r) between 0 to 0.2 was poor agreement, 0.21 
to 0.40 was fair agreement, 0.41 to 0.60 was moderate/acceptable agreement, 0.61 to 0.80 
was substantial agreement, and 0.81 to 1.0 was near perfect agreement [14].   
Linear regression prediction models were used to identify participants’ 
characteristics as predictors of discordance between the GPAQ and accelerometer.  
Discordancy was calculated by subtracting the amount of time spent in MVPA 
according to the GPAQ from the amount spent in MVPA according to accelerometry.  
The average of the GPAQ values from the two visits per round was used.  The 
75 
 
dependent variable in the analysis was the difference between the two values, 
calculated by the proportion of the time spent at MVPA from the GPAQ minus the 
proportion of the time at MVPA from accelerometry.  Participant characteristics were 
the independent variables.  Each variable was regressed individually to determine if 
there was a statistically significant association.  Variables were retained for inclusion in 
the full model if their p-value was less than 0.20. Backward stepwise regression was 
used to build a prediction model for discordancy; variables were retained in the model 
if their p-value was less than 0.05.  Prediction models were considered for the two 
rounds combined and each round individually. Interaction terms for the round were 
considered in the combined model.   
Variables considered potential discordance predictors included age, education, 
BMI, whether their partner lived in the household, round, geography, and food 
security. The relationship between participant characteristics and self-reported physical 
activity is not well understood. BMI and marital status were selected based on previous 
literature [52].  The additional variables were selected based on their relationships with 
other health outcomes [53–55].  Median village altitude was used as an indicator of 
topography.  The topography of a village could be essential to understand why 
participants over or underestimated their physical exertion. For example, participants 
that live at the top of a mountain may perceive that they are exerting themselves more 
due to the rugged terrain, or they may underestimate their physical exertion because 
they are habituated to the effort of walking up steep hills. Food security was measured 
using the food gap, a subjective indicator used widely in Ethiopia [26].  Scatter plots 
were used to evaluate the relationship between the variables and determine if 
transforming the variables was necessary.   
24-hour recall of perceived exertion analysis 
76 
 
Validity of the 24-hour recall of physical activity was determined by comparing 
the proportion of time spent in each of the four areas of interest, light/sedentary 
activity, moderate activity, vigorous activity, and MVPA, to the values measured by 
accelerometry.  Values were compared using both Pearson correlation coefficients and 
Spearman Rank correlation coefficients.  Variables checked for validity were the 
proportion of time spent at sedentary/light physical activity, moderate physical 
activity, vigorous physical activity, and MVPA.   
Multivariable regression modeling was used to create a calibrated recall value.  
24-hour data was split into training (80%) and testing (20%) samples.  The training 
sample was used to create the regression model to calibrate the 24-hour recall data with 
the results of accelerometry as the dependent variable and the results of the recall data 
and participant characteristics as independent variables.  The model was fit using 
backward-stepwise regression that used Bayesian information criterion (BIC) to select 
the best fit model.  Random effect and fixed effect models were considered to address 
potential independence issues resulting from participants who appeared twice in the 
dataset during the PSNP round and again during the agricultural round.  In addition, 
breakpoint analysis was considered using the strucchange package in R software 
(version 4) [56–58].  The model was used on the testing data to calculate a new, 
calibrated value for the proportion of the time at MVPA.  The new, calibrated value was 
compared to the accelerometry results using Pearson and Spearman Rank correlation. 
Ethical approval 
This study received ethical approval from the IRB at Cornell University 
(1902008596) and the International Food Policy Research Institute (DSG-19-0203P). 
Results 
77 
 
Study population 
Demographic characteristics of the study population appear in Table 3.1.  
Accelerometry was collected on 95 females during the PSNP round and 96 during the 
agricultural round.  There was valid accelerometry data from 89 females from the PSNP 
round and 91 from the agricultural round.  Sixty-four females participated in both 
rounds.  Twenty-five females only participated in the PSNP round, and 27 females only 
participated in the agricultural round.  There were few differences in participant 
characteristics between rounds. 
The average age of the study participants was 30 years.  Forty-six percent had no 
education, and 86.7% of females had partners living in their households.  The average 
BMI of the index mothers was 19.4 kg/ 2m , and approximately 33% of females were 
underweight (BMI < 18.5 kg/ 2m ).  Dietary diversity was measured using the Minimum 
Dietary Diversity-Women (MDD-W) [59].  Overall, dietary diversity was low; only 4.4% 
of females met minimum dietary diversity by consuming food from at least five of the 
ten recommended food groups.  On average, females only consumed foods from three 
food groups.  The main food groups consumed were starchy staples, beans and peas, 
and non-vitamin A-rich vegetables.  Most of the mothers’ time was spent on childcare, 
cooking/eating, domestic activities, and personal care during the small rainy season. 
During the main rainy season, most of their time was spent on childcare, 
cooking/eating, and food production activities [26]. 
  
78 
 
 
Table 3. 1: Summary of study participants' characteristics 
Characteristic PSNP Round Agricultural 
Round 
N 89 91 
Age 29 (7) 30 (7) 
Any education 47.8% 44.4% 
Partner lives in household 87.8% 85.5% 
BMI (kg/ 2m ) 19.4 (2.1) 19.4 (2.0) 
Underweight (BMI < 18.5 kg/ 2m ) 33.3% 33.0% 
Minimum dietary diversity 4.4% 4.4% 
Household size 5.5 (1.8) 5.7 (1.7) 
Male headed households 90.0% 85.6% 
Households were smallholder farmers 84.4% 83.3% 
Benefits from public works safety net 48.3% 45.5% 
Food insecure during past 6 months 32.2% 28.9% 
Food groups consumed 3.0 (0.5) 3.2 (0.8) 
Distance to nearest city with 50,000 
inhabitants (km) 67.3 (28.0) 66.5 (27.9) 
Travel time to market (minutes) 44.9 (34.5) 44.6 (34.8) 
Time to fetch water and return (minutes) 44.0 (39.0) 42.8 (31.4) 
Houses with no or open windows 62.2% 61.1% 
Houses with dirt floors 67.8% 67.8% 
Houses with corrugated metal roofs 53.3 61.1% 
Access to electricity 43.3% 44.4% 
Owned mobile phone 67.8% 66.7% 
Values are presented as means with standard deviation in parentheses or as a percent 
of the study population. 
 
 The participants lived in households with an average of 5.6 occupants.  Males 
headed more households during the PSNP round than the agricultural round (90.0% vs. 
85.6%).  Typical of the region, 83.9% of households were actively engaged in farming.  
The average area of land worked by the households was 8.9 hectares.  Less than half, 
46.7%, benefitted from the public works component of the PSNP.  Food security was 
measured by the food gap, which was the number of months the household could not 
satisfy its food need in the twelve previous months [26]. For this survey, the reporting 
period was shortened to six months because of the overlap in reporting period.  Slightly 
79 
 
more households reported being food insecure in the past six months during the PSNP 
round than the agricultural round. 
The households were located in rural areas of Tigray, where the average distance 
to the nearest city with at least 50,000 inhabitants was 67 km.  The average travel time to 
the market was 45 minutes.  Most houses had no windows or open windows, dirt 
floors, and corrugated metal roofs.  Only 7% of participants had water piped into their 
dwelling.  For the remaining participants, the average time to fetch water and return 
was approximately 44 minutes.  Forty-four percent of households had access to 
electricity, and two-thirds (67.2%) owned a mobile phone. 
GPAQ 
Reliability 
The reliability of the responses to the dichotomous questions, whether a 
participant spent time in a domain at either moderate or vigorous exertion, from the 
GPAQ is found in Table 3.2.   When the rounds were combined, there was a substantial 
agreement regarding whether participants spent time at moderate- or vigorous-
intensity levels while working.  However, the responses to the same questions had an 
acceptable agreement when the PSNP round was evaluated independently.  The 
agreement for whether the participants participated in moderate or vigorous work 
during the agricultural round was substantial.  For travel, the agreement was fair (k = 
0.33) for the PSNP round and near perfect for the agricultural round (k = 0.82).  In the 
rounds combined, there was substantial agreement. However, there was a poor 
agreement for vigorous recreation and poor to a fair agreement for moderate recreation.  
For each visit of each survey round, less than 2% of participants said they engaged in 
moderate or vigorous recreational activities.   
80 
 
Table 3. 2: Reliability of the GPAQ for dichotomous responses  
All PSNP Round Agricultural 
Domain Exertion level Round 
N  185 93 92 
Work     
 Vigorous 0.63*** 0.57*** 0.69*** 
(85.41%) (81.72%) (89.13%) 
 Moderate 0.63*** 0.46*** 0.80*** 
(83.78%) (76.34%) (91.30%) 
Travel  0.69*** 0.33*** 0.82*** 
(87.03%) (82.80%) (91.30%) 
Recreation 1     
 Vigorous  -0.02 -0.03 -0.01 
(95.68%) (94.62%) (96.74%) 
 Moderate  0.39*** 0.49*** 0.00 
(98.38%) (97.85%) (98.91%) 
The k for each response is shown with the percent of agreement below.  *** Indicates 
statistical significance <0.01 1 Less than five participants said they participated in 
vigorous or moderate recreation during each visit of each round. 
 
Agreement for the number of minutes respondents reported spending at 
vigorous work was fair during the PSNP round (Spearman’s r = 0.45) and substantial 
during the agricultural round (Spearman’s r = 0.78) (Table 3.3).  Agreement for the 
number of minutes respondents reported spending at moderate work was substantial 
during the PSNP round, and there was a near-perfect agreement during the agricultural 
round (Spearman’s r = 0.81). Agreement for the number of minutes respondents 
reported traveling was substantial during the PSNP round and near-perfect during the 
agricultural round (Spearman’s r = 0.84).  Agreement for total minutes respondents 
reported spending on recreation activities ranged from moderate/acceptable (r = 0.44) 
to substantial (r = 0.75). 
  
81 
 
 
Table 3. 3: Reliability of the GPAQ for continuous responses – Spearman’s r 
All PSNP Round Agricultural 
Domain Exertion level Round 
N  185 93 92 
Work     
 Vigorous 0.64 *** 0.45 ** 0.78 *** 
 Moderate 0.74 *** 0.64 *** 0.81 **** 
Travel  0.69 *** 0.50 ** 0.84 **** 
 
Recreation 1     
 Vigorous  -0.02 -0.03 -0.02 
 Moderate  -0.02 -0.03 -0.02 
Total minutes     
 Vigorous 0.61 *** 0.44 ** 0.72 *** 
 Moderate 0.69 *** 0.60 ** 0.75 *** 
 Sedentary/light 0.67 *** 0.69 *** 0.62 *** 
* Indicates fair agreement, ** indicates moderate/acceptable agreement, *** indicates 
substantial agreement, **** indicated near-perfect agreement. 
 
Validity 
The validity of the GPAQ was determined by comparing the proportion of 
minutes spent at each level of exertion as reported by the participant to the proportion 
of minutes measured by the accelerometry with the rounds combined and 
disaggregated by round (Table 3.4).  Overall, there was a lack of agreement between the 
proportion of minutes spent in each level of exertion according to the GPAQ and 
accelerometry.  When the two rounds are combined, the average proportion of time 
spent at sedentary/light and moderate exertion were similar when measured by the 
GPAQ (88.9%) and accelerometry (90.2%).  Despite the similarities in the average 
proportion of time spent at each exertion level, there was a poor agreement with 
Pearson’s r never exceeding 0.20 and Spearman’s r never exceeding 0.18. 
  
82 
 
Table 3. 4: Validity of GPAQ 
Pearson Spearman’s 
Exertion level GPAQ Accelerometry  r rank r 
Total     
N  180 180   
Proportion of time at 
sedentary/light exertion 88.9% 90.2% -0.06 -0.02 
Proportion of time at moderate 
exertion  9.5% 9.3% -0.12 -0.06 
Proportion of time at vigorous 
exertion 1.6% 0.4% 0.06 0.16 
Proportion of time at MVPA 11.1% 9.8% -0.06 -0.02 
PSNP Round     
N 89 89   
Proportion of time at 
sedentary/light exertion 91.8% 90.6% -0.11 0.10 
Proportion of time at moderate 
exertion 6.7% 9.6% -0.01 0.05 
Proportion of time at vigorous 
exertion 1.4% 0.4% 0.20 0.18 
Proportion of time at MVPA 8.2% 10.0% 0.07 0.10 
Agricultural Round     
N 91 91   
Proportion of time at 
sedentary/light exertion 86.0% 90.4% -0.11 -0.09 
Proportion of time at moderate 
exertion 12.1% 9.0% -0.15 -0.13 
Proportion of time at vigorous 
exertion 1.8% 0.5% 0.01 0.15 
Proportion of time at MVPA 14.0% 9.6% -0.11 -0.09 
* Indicates fair agreement, ** indicates moderate/acceptable agreement, *** indicates 
substantial agreement, **** indicated near-perfect agreement. 
 
According to the accelerometry data, the proportion of time spent at different 
exertion levels was very similar during both rounds.  However, according to the GPAQ, 
participants spent less time at sedentary/light exertion and more time at moderate 
exertion during the agricultural round than the PSNP round. Participants reported 
spending 91.8% of their time in sedentary/light exertion during the PSNP round and 
86% during the agricultural round. Meaning, participants reported spending 5.8% more 
83 
 
of their time at MVPA during the agricultural round compared to the PSNP round, 
approximately 50 minutes per day.   
Discordancy 
During the PSNP round, the mean difference between the percent of the time in 
MVPA was 0.02 with a range of -0.19 to 0.15.  On average, participants underestimated 
the proportion of time spent at MVPA by 2%.  During the agricultural round, the mean 
difference between the percent of the time in MVPA was -0.04 with a range of -0.32 to 
0.18.  On average, participants overestimated the proportion of time spent at MVPA by 
4%.   
In a univariate analysis, when both rounds are combined round, the proportion 
of time spent at MVPA according to the accelerometer, BMI, a partner in household, 
and food security were all significant predictors of discordancy (Table 3.5).  However, 
when fit into a multivariate model, only round and the proportion of time spent at 
MVPA remained statistically significant.  During the agricultural round compared to 
the PSNP round, participants overestimated the percent of time spent at MVPA by 6% 
when controlling for the proportion of time spent at MVPA.  When controlling for 
round, as the proportion of time spent at MVPA increased, the participants 
underestimated the proportion of time spent at MVPA.   
In a univariate analysis for the PSNP round alone, the proportion of time spent at 
MVPA, having a partner residing in the household, and altitude were the only 
significant predictors of discordancy that met the cutoff to be considered in the 
multivariate model.  A partner residing in the household and altitude were no longer 
significant when included in a model with the proportion of time spent at MVPA.  For 
the agricultural round, the proportion of time spent at MVPA, BMI, a partner in the  
84 
 
 
Table 3. 5: GPAQ Discordancy 
85 
 
household, and food insecurity were predictors of discordancy. However, when fit into 
a multivariate analysis, only the proportion of time at MVPA remained significant.  The 
discordancy between rounds was evaluated using Spearman’s Rank Correlation.  The 
Spearman’s r was 0.38, which is a fair agreement. 
24-hour recall 
Validity 
The validity of the 24-hour recall to the accelerometry measurements were 
compared with the rounds combined and each round on its own, Table 3.6.  When the 
rounds were combined, participants overestimated the percentage of moderate or 
vigorous exertion time and underestimated the time spent in sedentary/light exertion.  
Participants estimated that they spent 23% of their time in MVPA, whereas 
accelerometry measured 13% of their time spent in MVPA.  Despite the large difference 
in values, there was a fair agreement between the two, Spearman’s r =0.25. 
According to accelerometry, participants spent more time in MVPA during the 
agricultural round than during the PSNP round, 15% vs. 11%.  There was a fair 
agreement between the reported and the measured proportion of the time at MVPA 
during the PSNP round, Spearman’s r = 0.27.   
Participants reported spending a greater proportion of time in MVPA during the 
agricultural round than the PSNP round, 28% vs. 17%.  There was poor agreement 
between the participants’ self-reported proportion of time spent at MVPA compared to 
accelerometry, Spearman’s r = 0.16. 
  
86 
 
Table 3. 6: Validity of 24-hour recall 
24-hour Accelerometry Pearson’s Spearman’s 
Exertion level recall r rank r 
Total     
 N 180 180   
Proportion of time at 
 sedentary/light exertion 
77% 87% 0.26* 0.25* 
Proportion of time at 
 moderate exertion  
20% 12% 0.19 0.20 
Proportion of time at 
vigorous exertion 3% 1% 0.15 0.24*  
 Proportion of time at MVPA 23% 13% 0.26* 0.25* 
PSNP Round     
 N 89 89   
Proportion of time at 
sedentary/light exertion 83% 89% 0.25* 0.26*  
Proportion of time at 
 moderate exertion  
16% 11% 0.20 0.23* 
Proportion of time at 
 vigorous exertion 
1% 0% 0.01 0.18 
 Proportion of time at MVPA 17% 11% 0.25* 0.27* 
Agricultural Round     
 N 91 91   
Proportion of time at 
 sedentary/light exertion 
72% 85% 0.19 0.16 
Proportion of time at 
 moderate exertion  
23% 14% 0.13 0.10 
Proportion of time at 
 vigorous exertion 
5% 1% 0.09 0.22* 
 Proportion of time at MVPA 28% 15% 0.19 0.16 
* Indicates fair agreement, ** indicates moderate/acceptable agreement, *** indicates 
substantial agreement, **** indicated near-perfect agreement. 
 
Calibration 
The calibrated model of physical activity was created with both round combined, 
Table 3.7.  Breakpoint analysis indicated a breakpoint of 0.00 and was not used.  The 
BMI was significant in a multivariate model that regressed the proportion of time spent 
at MVPA according to 24-hour recall against the proportion spent at MVPA according 
to accelerometry.  Random effects had no impact on the coefficients or p- 
87 
 
 
88 
 
Table 3. 7: Calibration of the 24-hour recall model 
values of the covariates in the model.  When applied to the testing set of participants, 
calibration of the model using BMI increased the Spearman r to 0.53, increasing the 
agreement to moderate/acceptable.  The improvement in correlation between the raw 
values and the calibrated values is evident in when the values are plotted, Figure 3.1.  
 
 
Figure 3. 1: Scatter plot comparing the proportion of time spent at MVPA  
Scatter plot comparing the calibrated and raw 24-hour recall of proportion 
of the time spent in MVPA for the testing group against the proportion of 
time spent in MVPA from the accelerometer 
 
 
  
89 
 
Discussion 
The regular activities of life in rural highland Ethiopia are physically demanding, 
complicating efforts to improve vulnerable people's nutritional status.  While increased 
energy and micronutrients are important for health and economic productivity [60], it is 
challenging to understand the caloric needs of individuals without insight into their 
physical activity. Therefore, this study aimed to validate two different survey tools 
using accelerometers, the GPAQ, which the WHO had developed, and a 24-hour recall 
of time use and perceived energy exertion.   
Using the proportion of time spent at MVPA as the outcome of interest, the 
GPAQ had high reliability but poor validity compared to the proportion of time sent at 
MVPA as measured by accelerometry.  Overall, the GPAQ is not a valid method of 
measuring the proportion of time spent at different levels of physical exertion.  
Although the GPAQ lacked agreement with accelerometry, it still generated 
important information regarding characteristics associated with how a participant 
perceived or recalled their rate of exertion.  Those who spent the lowest proportion of 
time at MVPA tended to overestimate their proportion of time spent at MVPA.  Those 
who spent a greater proportion of their time at MVPA tended to underestimate it.  The 
underlying mechanism is not known; however, it may be a training effect.  Spending 
more time at MVPA decreases the rate of perceived exertion [61,62] when a person is 
not approaching their ventilatory threshold [63].  Unfortunately, we do not have heart 
rate data to determine increases in heart rate accompanied the increases in activity level.  
Participants overestimated the proportion of their time spent at MVPA more in the 
agricultural round when they were more likely to be engaged in food production 
activities than childcare or personal care activities [26].  More research is needed to 
determine if participants overestimated their physical activity because of the type of 
90 
 
activities they were engaged in or an underlying seasonal effect such as food security, 
caloric intake, [27] or diet composition [64].   
The calibrated 24-hour recall was a valid tool to measure the proportion of time 
spent in MVPA in this population.  While the reliability of the raw values only had a 
fair agreement, the agreement was increased to substantial when calibrated using BMI. 
BMI being significant in the multivariate analysis suggests that it is important in how 
the participants perceived physical exertion.  Some research exists that shows an 
association between higher rates of perceived exertion and adiposity in overweight or 
obese persons [65,66]. However, only three participants would be classified as 
overweight using the standard BMI cutoffs, and none had a BMI greater than 26 kg/ 2m .  
Previous research has suggested that Ethiopian adults have higher levels of excess 
adiposity at lower BMIs.  However, those studies consisted of an urban population with 
a more sedentary lifestyle [67].  The percent body fat of this population is not known.  
Future research into this relationship should consider participant characteristics, 
including body composition, physical function, anemia [68], and overall health.  
This study has strengths. Previous attempts to validate the GPAQ in Ethiopia 
had only used pedometers as the reference measure, whereas this study used triaxial 
accelerometers, enabling us to record movement in all directions, not just steps.  
Additionally, this study identified a significant association between GPAQ discordancy, 
the proportion of time spent at MVPA according to accelerometry, and season.   
This study identified a valid survey module, the 24-hour recall of time use and 
energy exertion, to estimate the proportion of the time in MVPA.  The 24-hour recall 
had many characteristics that likely contributed to it being valid when the GPAQ was 
not, including a shorter recall time, prompts that aided recall, and more detailed 
questions regarding their energy exertion. In addition, the 24-hour recall only required 
91 
 
the participants to recall their activity from the previous day instead of asking about an 
average week. Finally, the additional prompts encouraged participants to think about 
their time and activities before responding to their perceived exertion.    
This study has limitations.  Ethiopia is a large and diverse country; this study 
took only one region, Tigray. Accelerometry’s accuracy is improved when combined 
with heart-rate monitoring [69], which we were unable to use.  Additionally, while the 
sample size was adequate, additional participants might have allowed more precise 
estimates of associations.  The sample had some different participants in each round 
despite efforts to use the same participants (Table A2.1).  There were no differences in 
participant characteristics between the samples for each round.  However, we 
compared the results from the different samples in a sensitivity analysis looking at the 
validity of the GPAQ and 24-hour recall.  The GPAQ was still not valid (Table A2.2).  
The agreement between the proportion of time at MVPA as reported to the GPAQ and 
recorded by accelerometry was poor for both rounds, both in the groups that appeared 
in both rounds of data collection and those that only appeared in one round of data 
collection.  The 24-recall had a lower agreement when disaggregated by rounds and by 
those that appeared in both rounds compared to only one (Table A2.3).  
Methods to assess physical activity use a rationale similar to those used to assess 
dietary intake [70].  A single 24-hour recall per participant can be used to characterize 
the average of a group, provided the sample represents the underlying population of 
interest and variations throughout the week and timeframe of interest [71–74].  A seven-
day recall may better account for variation throughout the week, but the longer recall 
period can be difficult to remember accurately [75].  As such, 24-hour recall and seven-
day recall for dietary intake or physical activity have overlapping limitations, including 
being open to systematic bias and random error.  However, both have been validated 
92 
 
and used extensively for collecting data on dietary intake in a wide variety of scenarios 
[76–80].   
The agreement between the proportion of the time at MVPA using the calibrated 
24-hour recall and accelerometry had an acceptable agreement. However, it is 
important not to overstate the ability of this model to determine exact levels of physical 
activity or energy expenditure.  The agreement is sufficient to compare time spent at 
MVPA between groups and seasons in the rural highland regions of Ethiopia, but 
caution should be used for analysis beyond that.  
This study validated a new survey tool, the 24-hour recall, to estimate physical 
activity in females living in rural highland Ethiopia.  This tool can be used in future 
research to understand the demands of life in the region. Understanding physical 
activity levels is a critical component in addressing undernutrition in this region and 
will allow researchers to look beyond consumption patterns and into expenditure. In 
addition, the 24-hour recall of time use and perceived exertion will allow comparisons 
between groups of people to determine if there is an overall difference in the physical 
activity in females that participate in manual labor, such as that required by the PSNP. 
 
  
93 
 
REFERENCES 
1.  Central Statistical Agency (CSA) [Ethiopia] and ICF Ethiopia Demographic and 
Health Survey; CSA and ICF: Addis Ababa, Ethiopia and Rockville, Maryland, 
USA, 2016; ISBN 2511115530. 
2.  Black, R.E.; Victora, C.G.; Walker, S.P.; Bhutta, Z.A.; Christian, P.; de Onis, M.; 
Ezzati, M.; Grantham-McGregor, S.M.; Katz, J.; Martorell, R.; et al. Maternal and 
Child Undernutrition and Overweight in Low-Income and Middle-Income 
Countries. The Lancet 2013, 382, 427–451. 
3.  Strauss, J.; Thomas, D. Health, Nutrition, and Economic Development. Journal of 
Economic Literature 1998, 36, 766–817. 
4.  Black, R.E.; Allen, L.H.; Bhutta, Z.A.; Caulfield, L.E.; de Onis, M.; Ezzati, M.; 
Mathers, C.; Rivera, J. Maternal and Child Undernutrition: Global and Regional 
Exposures and Health Consequences. The Lancet 2008, 371, 243–260. 
5.  Victora, C.G.; Adair, L.; Fall, C.; Hallal, P.C.; Martorell, R.; Richter, L.; Sachdev, 
H.S. Maternal and Child Undernutrition: Consequences for Adult Health and 
Human Capital. Lancet 2008, 371, 340–357, doi:10.1016/S0140-6736(07)61692-4. 
6.  Barouki, R.; Gluckman, P.D.; Grandjean, P.; Hanson, M.; Heindel, J.J. 
Developmental Origins of Non-Communicable Disease: Implications for Research 
and Public Health. Environmental Health: A Global Access Science Source 2012, 11, 
1–9, doi:10.1186/1476-069X-11-42. 
7.  Goldstein, R.F.; Abell, S.K.; Ranasinha, S.; Misso, M.; Boyle, J.A.; Black, M.H.; Li, N.; 
Hu, G.; Corrado, F.; Rode, L.; et al. Association of Gestational Weight Gain with 
Maternal and Infant Outcomes: A Systematic Review and Meta-Analysis. JAMA - 
Journal of the American Medical Association 2017, 317, 2207–2225, 
doi:10.1001/jama.2017.3635. 
8.  Ay, L.; Kruithof, C.J.; Bakker, R.; Steegers, E.A.P.; Witteman, J.C.M.; Moll, H.A.; 
Hofman, A.; MacKenbach, J.P.; Hokken-Koelega, A.C.S.; Jaddoe, V.W.V. Maternal 
Anthropometrics Are Associated with Fetal Size in Different Periods of Pregnancy 
and at Birth. the Generation R Study. BJOG: An International Journal of Obstetrics 
and Gynaecology 2009, 116, 953–963, doi:10.1111/j.1471-0528.2009.02143.x. 
9.  Flick, A.A.; Brookfield, K.F.; De La Torre, L.; Tudela, C.M.; Duthely, L.; González-
Quintero, V.H. Excessive Weight Gain among Obese Women and Pregnancy 
Outcomes. American Journal of Perinatology 2010, 27, 333–338, doi:10.1055/s-0029-
1243304. 
10.  Svedberg, P.; Svedberg, P. of D.E.I. for I.E.S.P. Poverty and Undernutrition: Theory, 
Measurement, and Policy; Oxford University Press, 2000; ISBN 978-0-19-829268-5. 
11.  World Bank The State of Social Safety Nets 2018; World Bank: Washington DC, 
2018; ISBN 9781464812545. 
94 
 
12.  Report, D.; Berhane, G.; Golan, J.; Hirvonen, K.; Hoddinott, J.; Kim, S.; Taffesse, 
A.S.; Abate, Y.; Abay, K.; Assefa, T.; et al. Evaluation of the Nutrition Sensitive 
Features of the Productive Safety Nets Programme IV: Baseline Survey Report. 
2018. 
13.  Nelms, M.; Sucher, K.P. Nutrition Therapy & Pathophysiology; Fourth.; Cengage: 
Boston, 2020. 
14.  Bull, F.C.; Maslin, T.S.; Armstrong, T. Global Physical Activity Questionnaire 
(GPAQ): Nine Country Reliability and Validity Study. Journal of Physical Activity 
and Health 2009, 6, 790–804, doi:10.1123/jpah.6.6.790. 
15.  Craig, C.L.; Marshall, A.L.; Sjöström, M.; Bauman, A.E.; Booth, M.L.; Ainsworth, 
B.E.; Pratt, M.; Ekelund, U.; Yngve, A.; Sallis, J.F.; et al. International Physical 
Activity Questionnaire: 12-Country Reliability and Validity. Medicine and Science 
in Sports and Exercise 2003, 35, 1381–1395, 
doi:10.1249/01.MSS.0000078924.61453.FB. 
16.  Laeremans, M.; Dons, E.; Avila-Palencia, I.; Carrasco-Turigas, G.; Orjuela, J.P.; 
Anaya, E.; Brand, C.; Cole-Hunter, T.; De Nazelle, A.; Götschi, T.; et al. Physical 
Activity and Sedentary Behaviour in Daily Life: A Comparative Analysis of the 
Global Physical Activity Questionnaire (GPAQ) and the SenseWear Armband. 
PLoS ONE 2017, 12, 1–15, doi:10.1371/journal.pone.0177765. 
17.  Pedisic, Z.; Shrestha, N.; Loprinzi, P.D.; Mehata, S.; Mishra, S.R. Prevalence, 
Patterns, and Correlates of Physical Activity in Nepal: Findings from a Nationally 
Representative Study Using the Global Physical Activity Questionnaire (GPAQ). 
BMC Public Health 2019, 19, 1–8, doi:10.1186/s12889-019-7215-1. 
18.  Bull, F.C.; Maslin, T.S.; Armstrong, T. Global Physical Activity Questionnaire 
(GPAQ): Nine Country Reliability and Validity Study. Journal of Physical Activity 
and Health 2009, 6, 790–804. 
19.  Laurent, S.; Boutouyrie, P.; Asmar, R.; Gautier, I.; Laloux, B.; Guize, L.; 
Ducimetiere, P.; Benetos, A.; Jonville, S.; Jutand, L.; et al. Assessment of Moderate 
Intensity Physical Activity in the Field. Respiratory physiology & neurobiology 
2007, 8, 442–449, doi:10.1249/MSS.0b013e31826beae5. 
20.  Leonard, W.R. Laboratory and Field Methods for Measuring Human Energy 
Expenditure. American Journal of Human Biology 2012, 24, 372–384, 
doi:10.1002/ajhb.22260. 
21.  Prince, S.A.; Adamo, K.B.; Hamel, M.E.; Hardt, J.; Gorber, S.C.; Tremblay, M. A 
Comparison of Direct versus Self-Report Measures for Assessing Physical Activity 
in Adults: A Systematic Review. International Journal of Behavioral Nutrition and 
Physical Activity 2008, 5, doi:10.1186/1479-5868-5-56. 
95 
 
22.  Ainslie, P.N.; Reilly, T.; Westerterp, K.R. Estimating Human Energy Expenditure 
Doubly Labelled Water. Sports Medicine 2003, 33, 683–698, doi:10.2165/00007256-
200333090-00004. 
23.  Chen, K.Y.; Bassett, D.R. The Technology of Accelerometry-Based Activity 
Monitors: Current and Future. Medicine and Science in Sports and Exercise 2005, 
37, doi:10.1249/01.mss.0000185571.49104.82. 
24.  Sasaki, J.E.; John, D.; Freedson, P.S. Validation and Comparison of ActiGraph 
Activity Monitors. Journal of Science and Medicine in Sport 2011, 14, 411–416, 
doi:10.1016/j.jsams.2011.04.003. 
25.  Ndahimana, D.; Kim, E.-K. Measurement Methods for Physical Activity and 
Energy Expenditure: A Review. Clinical nutrition research 2017, 6, 68–80, 
doi:10.7762/cnr.2017.6.2.68. 
26.  Berhane, G.; Golan, J.; Hirvonen, K.; Hoddinott, J.F.; Kim, Sunny; Seyoum Tafesse, 
Alemayehu; Abay, Kibrewossen; Assefa, Thomas; Habte, Yetmwork Evaluation of 
the Nutrition-Sensitive Features of the Fourth Phase of Ethiopia’s Productive 
Safety Net Programme; 0 ed.; International Food Policy Research Institute: 
Washington, DC, 2020. 
27.  Hirvonen, K.; Taffesse, A.S.; Worku Hassen, I. Seasonality and Household Diets in 
Ethiopia. Public Health Nutr. 2016, 19, 1723–1730, doi:10.1017/S1368980015003237. 
28.  World Health Organization Global Physical Activity Questionnaire (GPAQ) 
Analysis Guide; World Health Organization: Geneva, Switzerland, 2005. 
29.  Aadland, E.; Ylvisåker, E. Reliability of the Actigraph GT3X+ Accelerometer in 
Adults under Free-Living Conditions. PLoS ONE 2015, 10, 1–10, 
doi:10.1371/journal.pone.0134606. 
30.  Cleland, C.L.; Hunter, R.F.; Kee, F.; Cupples, M.E.; Sallis, J.F.; Tully, M.A. Validity 
of the Global Physical Activity Questionnaire (GPAQ) in Assessing Levels and 
Change in Moderate-Vigorous Physical Activity and Sedentary Behaviour. BMC 
Public Health 2014, 14, 1255, doi:10.1186/1471-2458-14-1255. 
31.  Rivière, F.; Widad, F.Z.; Speyer, E.; Erpelding, M.L.; Escalon, H.; Vuillemin, A. 
Reliability and Validity of the French Version of the Global Physical Activity 
Questionnaire. Journal of Sport and Health Science 2018, 7, 339–345, 
doi:10.1016/j.jshs.2016.08.004. 
32.  Mathews, E.; Salvo, D.; Sarma, P.S.; Thankappan, K.R.; Pratt, M. Adapting and 
Validating the Global Physical Activity Questionnaire (GPAQ) for Trivandrum, 
India, 2013. Preventing chronic disease 2016, 13, E53, doi:10.5888/pcd13.150528. 
33.  Mumu, S.J.; Ali, L.; Barnett, A.; Merom, D. Validity of the Global Physical Activity 
Questionnaire (GPAQ) in Bangladesh. BMC Public Health 2017, 17, 1–10, 
doi:10.1186/s12889-017-4666-0. 
96 
 
34.  Lim, S.; Wyker, B.; Bartley, K.; Eisenhower, D. Measurement Error of Self-Reported 
Physical Activity Levels in New York City: Assessment and Correction. American 
Journal of Epidemiology 2015, 181, 648–655, doi:10.1093/aje/kwu470. 
35.  Alkire, S.; Meinzen-Dick, R.; Peterman, A.; Quisumbing, A.R.; Seymour, G.; Vaz, A. 
The Women’s Empowerment in Agriculture Index. IFPRI Discussion Paper 2012, 
19, 346, doi:10.1016/j.worlddev.2013.06.007. 
36.  Quisumbing, A.; Meinzen-Dick, R.; Malapit, H. Women’s Empowerment and 
Gender Equality in South Asian Agriculture: Measuring Progress Using the 
Project-Level Women’s Empowerment in Agriculture Index (pro-WEAI) in 
Bangladesh and India. World Development 2021, 105396, 
doi:10.1016/j.worlddev.2021.105396. 
37.  Mponela, P.; Villamor, G.B.; Snapp, S.; Tamene, L.; Le, Q.B.; Borgemeister, C. The 
Role of Women Empowerment and Labour Dependency on Adoption of Integrated 
Soil Fertility Management in Malawi. Soil Use and Management 2021, 37, 390–402, 
doi:10.1111/sum.12627. 
38.  Crookston, B.T.; West, J.H.; Davis, S.F.; Hall, P.C.; Seymour, G.; Gray, B.L. 
Understanding Female and Male Empowerment in Burkina Faso Using the Project-
Level Women’s Empowerment in Agriculture Index (pro-WEAI): A Longitudinal 
Study. BMC Women’s Health 2021, 21, 1–11, doi:10.1186/s12905-021-01371-9. 
39.  Tudor-Locke, C.; Washington, T.L.; Ainsworth, B.E.; Troiano, R.P. Linking the 
American Time Use Survey (ATUS) and the Compendium of Physical Activities: 
Methods and Rationale. Journal of Physical Activity and Health 2009, 6, 347–353. 
40.  Tudor-Locke, C.; Leonardi, C.; Johnson, W.D.; Katzmarzyk, P.T. Time Spent in 
Physical Activity and Sedentary Behaviors on the Working Day: The American 
Time Use Survey. Journal of Occupational and Environmental Medicine 2011, 53, 
1382–1387, doi:10.1097/JOM.0b013e31823c1402. 
41.  van der Ploeg, H.P.; Merom, D.; Chau, J.Y.; Bittman, M.; Trost, S.G.; Bauman, A.E. 
Advances in Population Surveillance for Physical Activity and Sedentary Behavior: 
Reliability and Validity of Time Use Surveys. American Journal of Epidemiology 
2010, 172, 1199–1206, doi:10.1093/aje/kwq265. 
42.  Tudor-Locke, C.; Ainsworth, B.E.; Washington, T.L.; Troiano, R. Assigning 
Metabolic Equivalent Values to the 2002 Census Occupational Classification 
System. Journal of Physical Activity and Health 2011, 8, 581–586. 
43.  Ng, S.W.; Popkin, B.M. Time Use and Physical Activity: A Shift Away from 
Movement across the Globe. Obesity reviews : an official journal of the 
International Association for the Study of Obesity 2012, 13, 659–680, 
doi:10.1111/j.1467-789X.2011.00982.x. 
97 
 
44.  Brønd, J.C.; Arvidsson, D. Sampling Frequency Affects the Processing of Actigraph 
Raw Acceleration Data to Activity Counts. Journal of Applied Physiology 2015, 
120, 362–369, doi:10.1152/japplphysiol.00628.2015. 
45.  Choi, L.; Liu, Z.; Matthews, C.E.; Buchowski, M.S. Validation of Accelerometer 
Wear and Nonwear Time Classification Algorithm. Medicine and Science in Sports 
and Exercise 2011, 43, 357–364, doi:10.1249/MSS.0b013e3181ed61a3. 
46.  Crouter, S.E.; Kuffel, E.; Haas, J.G.; Frongllo, E.A.; Bassett, D.R. A Refined 2-
Regression Model for the Actigraph Accelerometer. Medicine & Science in Sports & 
Exercise 2010, 42, 1029–1037, doi:10.1249/MSS.0b013e3181c37458.A. 
47.  Aguilar-Farias, N.; Peeters, G.M.E.E. (Geeske); Brychta, R.J.; Chen, K.Y.; Brown, 
W.J. Comparing ActiGraph Equations for Estimating Energy Expenditure in Older 
Adults. Journal of Sports Sciences 2019, 37, 188–195, 
doi:10.1080/02640414.2018.1488437. 
48.  Blakstad, M.M.; Nevins, J.E.H.; Venkatramanan, S.; Przybyszewski, E.M.; Haas, J.D. 
Iron Status Is Associated with Worker Productivity, Independent of Physical Effort 
in Indian Tea Estate Workers. Applied Physiology, Nutrition and Metabolism 2020, 
45, 1360–1367, doi:10.1139/apnm-2020-0001. 
49.  Metzger, J.S.; Catellier, D.J.; Evenson, K.R.; Treuth, M.S.; Rosamond, W.D.; Siega-
Riz, A.M. Patterns of Objectively Measured Physical Activity in the United States. 
Medicine and Science in Sports and Exercise 2008, 40, 630–638, 
doi:10.1249/MSS.0b013e3181620ebc. 
50.  McHugh, M.L. Lessons in Biostatistics Interrater Reliability : The Kappa Statistic. 
Biochemia Medica 2012, 22, 276–282, doi:10.11613/BM.2012.031. 
51.  Gisev, N.; Bell, J.S.; Chen, T.F. Interrater Agreement and Interrater Reliability: Key 
Concepts, Approaches, and Applications. Research in Social and Administrative 
Pharmacy 2013, 9, 330–338, doi:10.1016/j.sapharm.2012.04.004. 
52.  Gorzelitz, J.; Peppard, P.E.; Malecki, K.; Gennuso, K.; Nieto, F.J.; Cadmus-Bertram, 
L. Predictors of Discordance in Self-Report versus Device-Measured Physical 
Activity Measurement. Annals of Epidemiology 2018, 28, 427–431, 
doi:10.1016/j.annepidem.2018.03.016. 
53.  Robards, J.; Evandrou, M.; Falkingham, J.; Vlachantoni, A. Marital Status, Health 
and Mortality. Maturitas 2012, 73, 295–299, doi:10.1016/j.maturitas.2012.08.007. 
54.  Burtscher, M. Effects of Living at Higher Altitudes on Mortality: A Narrative 
Review. Aging and Disease 2014, 5, 274–280, doi:10.14336/AD.2014.0500274. 
55.  Cutler, D.M.; Lleras-Muney, A. Education and Health: Evaluating Theories and 
Evidence. National Bureau of Economic Research Working Paper 2006, 
doi:10.1590/s0100-40422000000600014. 
98 
 
56.  Zeileis, A.; Leisch, F.; Hornik, K.; Kleiber, C.; Hansen, B.; Merkle, E.C. Testing, 
Monitoring, and Dating Structural Changes Description - Package ‘ Strucchange .’ 
2019. 
57.  Zeileis, A.; Leisch, F.; Homik, K.; Kleiber, C. Strucchange: An R Package for Testing 
for Structural Change. Journal of Statistical Software 2002, 7, 1–38. 
58.  Metcalf, K.M.; Baquero, B.I.; Coronado Garcia, M.L.; Francis, S.L.; Janz, K.F.; 
Laroche, H.H.; Sewell, D.K. Calibration of the Global Physical Activity 
Questionnaire to Accelerometry Measured Physical Activity and Sedentary 
Behavior. BMC Public Health 2018, 18, 1–10, doi:10.1186/s12889-018-5310-3. 
59.  Minimum Dietary Diversity for Women (MDD-W) | Food and Nutrition Technical 
Assistance III Project (FANTA) Available online: 
https://www.fantaproject.org/monitoring-and-evaluation/minimum-dietary-
diversity-women-indicator-mddw. 
60.  Black, R.E.; Victora, C.G.; Walker, S.P.; Bhutta, Z.A.; Christian, P.; Onis, M. de; 
Ezzati, M.; Grantham-McGregor, S.; Katz, J.; Martorell, R.; et al. Maternal and Child 
Undernutrition and Overweight in Low-Income and Middle-Income Countries. 
The Lancet 2013, 382, 427–451, doi:10.1016/S0140-6736(13)60937-X. 
61.  Williams, J.G.; Eston, R.G. Determination of the Intensity Dimension in Vigorous 
Exercise Programmes with Particular Reference to the Use of the Rating of 
Perceived Exertion. Sports Med 1989, 8, 177–189, doi:10.2165/00007256-198908030-
00004. 
62.  Testa, M.; Noakes, T.D.; Desgorces, F.-D. Training State Improves the Relationship 
between Rating of Perceived Exertion and Relative Exercise Volume during 
Resistance Exercises. J Strength Cond Res 2012, 26, 2990–2996, 
doi:10.1519/JSC.0b013e31824301d1. 
63.  Grego, F.; Collardeau, M.; Vallier, J.M.; Delignieres, D.; Brisswalter, J. Effect of 
Long Duration Exercise on the Ratings of Perceived Exertion and Perceived 
Difficulty of Walking and Running at the Ventilatory Threshold. J Sports Med Phys 
Fitness 2004, 44, 368–374. 
64.  Kim, S.S.; Nguyen, P.H.; Tran, L.M.; Abebe, Y.; Asrat, Y.; Tharaney, M.; Menon, P. 
Maternal Behavioural Determinants and Livestock Ownership Are Associated with 
Animal Source Food Consumption among Young Children during Fasting in Rural 
Ethiopia. Matern Child Nutr 2018, 15, e12695, doi:10.1111/mcn.12695. 
65.  Ekkekakis, P.; Lind, E. Exercise Does Not Feel the Same When You Are 
Overweight: The Impact of Self-Selected and Imposed Intensity on Affect and 
Exertion. Int J Obes 2006, 30, 652–660, doi:10.1038/sj.ijo.0803052. 
66.  Martinez-Amezcua, P.; Simonsick, E.M.; Wanigatunga, A.A.; Urbanek, J.K.; Shaffer, 
N.C.; Ferrucci, L.; Schrack, J.A. The Association Between Adiposity and Perceived 
99 
 
Physical Fatigability in Mid-to-Late Life. Obesity (Silver Spring) 2019, 27, 1177–
1183, doi:10.1002/oby.22500. 
67.  Sinaga, M.; Yemane, T.; Tegene, E.; Lidstrom, D.; Belachew, T. Performance of 
Newly Developed Body Mass Index Cut-off for Diagnosing Obesity among 
Ethiopian Adults. J Physiol Anthropol 2019, 38, 14, doi:10.1186/s40101-019-0205-2. 
68.  Simonsick, E.M.; Patel, K.V.; Schrack, J.A.; Ferrucci, L. Fatigability as a Predictor of 
Subclinical and Clinical Anemia in Well-Functioning Older Adults. Journal of the 
American Geriatrics Society 2020, 68, 2297–2302, doi:10.1111/jgs.16657. 
69.  Ndahimana, D.; Kim, E.-K. Measurement Methods for Physical Activity and 
Energy Expenditure: A Review. Clin Nutr Res 2017, 6, 68–80, 
doi:10.7762/cnr.2017.6.2.68. 
70.  McClung, H.L.; Ptomey, L.T.; Shook, R.P.; Aggarwal, A.; Gorczyca, A.M.; Sazonov, 
E.S.; Becofsky, K.; Weiss, R.; Das, S.K. Dietary Intake and Physical Activity 
Assessment: Current Tools, Techniques, and Technologies for Use in Adult 
Populations. American Journal of Preventive Medicine 2018, 55, e93–e104, 
doi:10.1016/j.amepre.2018.06.011. 
71.  Gibson, R.S.; Ruth Charrondiere, U.; Bell, W. Measurement Errors in Dietary 
Assessment Using Self-Reported 24-Hour Recalls in Low-Income Countries and 
Strategies for Their Prevention. Advances in Nutrition 2017, 8, 980–991, 
doi:10.3945/an.117.016980. 
72.  Souverein, O.W.; Dekkers, A.L.; Geelen, A.; Haubrock, J.; de Vries, J.H.; Ocké, M.C.; 
Harttig, U.; Boeing, H.; van ‘t Veer, P. Comparing Four Methods to Estimate Usual 
Intake Distributions. European Journal of Clinical Nutrition 2011, 65, S92–S101, 
doi:10.1038/ejcn.2011.93. 
73.  Kroes, R.; Müller, D.; Lambe, J.; Löwik, M.R.H.; Van Klaveren, J.; Kleiner, J.; 
Massey, R.; Mayer, S.; Urieta, I.; Verger, P.; et al. Assessment of Intake from the 
Diet. Food and Chemical Toxicology 2002, 40, 327–385, doi:10.1016/S0278-
6915(01)00113-2. 
74.  Biro, G.; Hulshof, K.; Ovesen, L.; Amorim Cruz, J.A. Selection of Methodology to 
Assess Food Intake. European Journal of Clinical Nutrition 2002, 56, S25–S32, 
doi:10.1038/sj.ejcn.1601426. 
75.  Hebert, J.R.; Ockene, I.S.; Hurley, T.G.; Luippold, R.; Well, A.D.; Harmatz, M.G.; 
Rider, L.; Saperia, G.; Mullins, B.; Phillips, J.C.C.; et al. Development and Testing of 
a Seven-Day Dietary Recall. Journal of Clinical Epidemiology 1997, 50, 925–937, 
doi:10.1016/S0895-4356(97)00098-X. 
76.  Gibson, R.S.; Ferguson, E.L. An Interactive 24-Hour Recall for Assessing the 
Adequacy of Iron and Zinc Intakes in Developing Countries; ILSI Press: 
Washington DC, 1999; ISBN 9780981817613. 
100 
 
77.  Kubota, Y.; Maruyama, K.; Sato, S.; Ishikawa, Y.; Shimamoto, T.; Inagawa, M.; 
Ohshima, M.; Murai, S.; Iso, H. Reproducibility of 24-Hour Dietary Recall for 
Vitamin Intakes by Middle-Aged Japanese Men and Women. Journal of Nutrition, 
Health and Aging 2010, 14, 196–200, doi:10.1007/s12603-010-0048-3. 
78.  Engebretsen, I.M.S.; Wamani, H.; Karamagi, C.; Semiyaga, N.; Tumwine, J.; 
Tylleskär, T. Low Adherence to Exclusive Breastfeeding in Eastern Uganda: A 
Community-Based Cross-Sectional Study Comparing Dietary Recall since Birth 
with 24-Hour Recall. BMC Pediatrics 2007, 7, 1–12, doi:10.1186/1471-2431-7-10. 
79.  St. George, S.M.; Van Horn, M.L.; Lawman, H.G.; Wilson, D.K. Reliability of 24-
Hour Dietary Recalls as a Measure of Diet in African American Youth. J Acad Nutr 
Diet 201AD, 11, 1551–1559, doi:10.1016/j.jand.2016.05.011.Reliability. 
80.  Imaeda, N.; Kuriki, K.; Fujiwara, N.; Goto, C.; Tokudome, Y.; Tokudome, S. Usual 
Dietary Intakes of Selected Trace Elements (Zn, Cu, Mn, I, Se, Cr, and Mo) and 
Biotin Revealed by a Survey of Four-Season 7-Consecutive Day Weighed Dietary 
Records in Middle-Aged Japanese Dietitians. Journal of Nutritional Science and 
Vitaminology 2013, 59, 281–288, doi:10.3177/jnsv.59.281. 
  
101 
 
CHAPTER 4 
RELIABILITY AND VALIDITY OF PHYSICAL FUNCTION TESTS AND ADL SURVEY 
QUESTIONS IN FEMALES LIVING IN RURAL HIGHLAND ETHIOPIA 
Abstract 
In rural, highland Ethiopia, physical function, which is the physical ability to 
fulfill one’s daily roles and responsibilities, may be compromised by a lack of access to 
nutrition, healthcare, and sanitation.  Decreased physical function would be detrimental 
to health and income-generating activities.  Unfortunately, there is a lack of validated 
methods to measure physical function in adult females in this region.  This study aims 
to test the feasibility and reliability of physical tests, including the sit-to-stand (STS) and 
usual gait speed (UGS).  These physical tests will validate context-appropriate activities 
of daily living (ADL) questions for use in future surveys.  Study participants consisted 
of 316 females between the age of 18 and 45 years living in rural Tigray, Ethiopia that 
had previously participated in an impact evaluation of a safety net program.  Over a 
one-week period, participants completed the STS and UGS tests and responded to the 
ADL questionnaires three times.  Feasibility was ascertained qualitatively.  Reliability 
was assessed by comparing the results of the tests and questions between each visit 
using either Cohen’s k or Pearson’s r.  Validity was assessed by regressing the 
responses to the ADL questions against the results of the STS test, controlling for 
relevant participant characteristics.  STS was determined to be a feasible, reliable, and 
valid physical function test in rural, highland Ethiopia.  UGS lacked feasibility and 
reliability.  The validity of the ADLs was inconclusive.  
102 
 
Introduction 
A person’s ability to functionally, independently, and physically fulfill their 
activities of daily living (ADL) is known as physical function [1,2].  It is both a cause 
and consequence of body size and composition and physical activity.  Intrinsically, 
physical function is essential for overall health and quality of life [3,4].  Instrumentally, 
physical function can identify people at risk for further declines in health [5–9], 
requiring additional care [4], or having future health expenditures [10].  Additionally, 
studies have used it as an outcome measure to determine the impact of an intervention 
[11–13].  Outside of health measures, physical function impacts social networks [14,15] 
and income-generating activities [16–18].  Thus, measuring physical function provides 
vital insight into various aspects of health and wellbeing. 
Physical function is measured through physical tests or validated survey 
questions focused on the respondents’ ability to complete tasks required for personal 
care and independent living [19].  Two examples of physical tests include usual gait 
speed (UGS) and sit-to-stand tests (STS).  Gait speed strongly relates to 
pathophysiological mechanisms [20] and is considered a global marker of wellbeing 
[21].  The STS requires a combination of muscle power and balance [22].  It uses upper 
and lower-body musculature, requires continuous balance adjustments, and employs a 
motor pattern commonly used during ADLs [23].  Both tests collect continuous outcome 
measures, which collect data on a gradient at the upper end of the functional spectrum, 
making them more useful in young and active populations [5,24,25].  Additionally, they 
do not require specialized training or resources, making them useful in various settings.   
Survey questions require less time and training to administer than direct tests, 
making them suitable for population studies.  ADL surveys are one example of the 
validated survey questions used to measure physical function [26].  In addition, they 
103 
 
use population-specific questions to determine how well participants complete their 
ADLs [27,28].  For example, surveys for elderly or infirmed populations often ask 
questions such as the participant’s ability to bath, dress, and feed themselves 
[1,19,26,29].   
Medical practitioners and researchers in high-income countries have extensively 
measured physical function in a variety of populations, such as older adults with and 
without dementia [5,27,29–32] and people with chronic illness or pain (e.g., arthritis) 
[33–35].  However, measuring physical function has been underused in low- and 
middle-income countries (LMIC) populations, with few exceptions [27,36–38].   
Measuring physical function in LMIC is essential because daily life is physically 
demanding.  According to the 2016 Ethiopia Demographic and Health Survey (DHS), of 
the employed females living in rural Ethiopia, approximately 65% were employed in 
skilled manual labor, unskilled manual labor, or agriculture.  Additionally, 53% of 
females in rural households spent more than 30 minutes fetching drinking water daily 
[39].  Furthermore, populations living in rural Ethiopia are exposed to many risk factors 
that would negatively impact their physical function, including access to nutrition [40–
42], healthcare, and sanitation [43–45].  According to the 2016 DHS report, 25% of 
females living in rural Ethiopia were underweight (BMI < 18.5 kg/ 2m ) and 25% were 
anemic [39].  Thus, declines in physical function would significantly impact both 
necessary income-generating activities and activities needed for daily life.   
Physical function is not well understood in rural Ethiopia because the physical 
function assessment tools have not been studied in this setting.  Therefore, the 
objectives of this study were to determine the feasibility, reliability, and validity of the 
UGS, STS, and ADLs survey to measure physical function in females living in rural 
104 
 
highland Ethiopia.  By understanding physical function in the region, future research 
could identify those needing additional healthcare and factors influencing health. 
Materials and methods 
Study setting 
The setting for this study is Ethiopia, a low-income country located in east 
Africa.  Ethiopia is characterized by mountainous highland regions in the center and 
north of the country and flat lowland regions in the east and south.  Data were collected 
in the northern region of Tigray, which is landlocked and mountainous.  Most people in 
rural highland regions have livelihoods based on rainfed agriculture.  A short rainy 
season occurs between March and May, and the main rainy season occurs between June 
and October.  During the main rainy season, more than 90 percent of total crop 
production in the country takes place, with harvesting season taking place between 
October and December [46].  In addition, during the small rainy season, the poorest 
households in the regions receive additional support from a government social safety 
net called the Productive Safety Net Program (PSNP).  The PSNP provides cash or food 
payments for work on community projects such as terracing land, building roads, water 
reclamation, and reforestation efforts [46,47]. 
Study participants had previously responded to an extensive impact evaluation 
of the PSNP [46].  These individuals live in remote areas; the average distance to the 
nearest city was 226 km and 2.5 km to the nearest health post.  The average travel time 
to the market was 85 minutes by foot.  Most houses had no or open windows, dirt 
floors, and corrugated metal roofs.  The average time for participants to fetch water and 
return was 45 minutes.  Only 1% of participants had water piped into their dwelling.  
105 
 
Forty-two percent of households had access to electricity, and a little more than half 
(54.1%) owned a mobile phone [46]. 
Food security was high as measured by a food gap, with the average food gap 
being less than two months in the previous six months [46].  Approximately 90% of the 
households were smallholder farmers, working plots of 1.2 hectares.  Only 12% of 
households had a homestead garden in the previous 12 months.  Thirteen percent of 
households were engaged in additional business activities, and 21% had a household 
member participating in wage employment [46]. 
The initial impact evaluation collected additional data on one child (index child) 
aged 6-23m and their mother (mother of the index child).  The average BMI of the 
mothers was 19.5 kg/ 2m , and approximately 39% of females were underweight (BMI < 
18.5 kg/ 2m ) [46].  Overall, as measured by the minimum dietary diversity for females 
(MDD-W) [48], diet diversity was low, with the average score being 3.0 out of 10.0 food 
groups.  The main food groups consumed were starchy staples, beans and peas, and 
non-vitamin A-rich vegetables.  Mothers of the index children spent most of their time 
conducting childcare, cooking/ eating, domestic activities, and personal care during the 
dry season.  During the main rainy season, they spent most of their time on childcare, 
cooking/ eating, and food production activities [46]. 
Study subjects 
Participants in this study were the mother of index children whose households 
had previously taken part in an impact evaluation on the PSNP.  They were recruited 
based on the villages and counties where they lived.  Data collection took place in the 
Central, Eastern, and South zones for logistical and safety constraints.  Within these 
zones, counties were excluded if the PSNP was inactive.  Within each county, villages 
were excluded if they did not have the population needed for a concurrently running 
106 
 
study on physical activity.  This additional criterion was that they had two households 
that received benefits from the public works program and two that did not.  All 
participants living in the selected villages were screened for eligibility.  Inclusion 
criteria included females between the ages of 18-45 years, who had participated in the 
impact evaluation, were not pregnant, and were not planning on leaving the area 
during data collection.  One-hundred and seventy-eight females participated in the first 
round of data collection during the small rainy season, the PSNP round, and 138 
females participated in the second round of data collection during the main rainy 
season, the agricultural round.  One hundred and twenty-seven females participated in 
data collection during both rounds; 51 only participated during the PSNP round and 11 
during the agricultural round. 
Data collection 
Data collection took place over two four-week periods.  The first four-week 
period, April-May 2019, coincided with the small rainy season when minimal 
agricultural activities took place and the PSNP was operational.  The second four-week 
period, September-October 2019, occurred during the main rainy season when 
households were engaged in farming activities such as preparing the land, planting, 
and weeding.  The PSNP was non-operational.   
Over eight days, participants were visited up to six times.  Participants answered 
survey questions during three visits, including the ADL questions, and completed the 
physical function tests.  In addition, the participants answered questions about their 
age, education, and whether their partner resided in the same household during the 
first visit.  Relevant anthropometric information, including weight (Seca, Model 874 dr, 
Hamburg, Germany) and leg length, were also measured.  Females wore light clothing 
while being weighed.  Leg length was measured from the greater trochanter to the 
107 
 
lateral malleolus using a standard measuring tape.  All anthropometric measurements 
were measured twice, and the average of the two was used for analysis.  Additional 
demographic and village information was collected during the large impact evaluation 
and subsequently used in this study.  
Physical function instruments 
Sit-to-stand 
A repeated sit-to-stand test measures the time it takes a participant to rise from a 
seated position without using their hands.  This task requires a combination of muscle 
power and balance [22].  It uses upper and lower-body musculature, requires 
continuous balance adjustments, and employs a motor pattern commonly used in ADLs 
[23].  There is no standardized protocol for this test, and several variations exist.  For 
this study, we adapted the protocol used by Takai et al. (6), modifying the seated 
position and the number of tests to make them appropriate for this setting.  The subjects 
started in a standing position, then sank into a low seated squat before standing again.  
They squatted seven times as quickly as possible.  The bottoms of their feet remained in 
contact with the ground throughout the test.  The low seated squat is a common 
position for adults in developing countries to adopt when sitting or resting.  
Participants were able to complete the number of repetitions but reported that it was 
challenging and had started to fatigue.  Participants crossed their arms at the wrists and 
held them against their chest to ensure they were unaided in rising. 
A stopwatch recorded the time to the nearest 10th of a second.  The sit-to-stand 
test was performed twice with an interval of 1 minute between the trials.  The shorter of 
the two times was used for data analysis.  A power index (Psts) was calculated using 
108 
 
Equation 1.  Previous studies have found that the power index is correlated with the 
maximum voluntary isometric knee extension force [49]. 
(1) 
L × body mass × g × 7
Psts = ln T 	 sts
 
Where: Psts is the power index from the test 
L is the leg length measured from the greater trochanter to the lateral 
malleolus 
Body mass is measured in kg to the nearest tenth 
G is the acceleration of gravity (9.8 m/s2) 
Tsts is the time it took to complete the test 
Usual gait speed 
During the UGS, participants were timed as they walked 20 meters at their usual 
pace.  Enumerators laid out pre-measured lengths of rope on a flat surface, free of 
obstacles, on stable ground.  The participants started three steps behind the start of the 
distance over which they were recorded walking and continued walking until they 
were three steps beyond the finishing point.  The test was repeated three times, and the 
two closest times (out of three) were averaged and used for data analysis. 
Activities of daily living survey 
ADL surveys have been used to assess physical function since they were 
developed in 1963 [50].  Multiple versions of the survey exist.  Each version asks the 
study participants if they experienced difficulty completing routine activities such as 
getting in/out of bed, showering, and preparing food.  The routine activities selected 
109 
 
for this version of an ADL survey were based on pilot data from a “24-hour recall of 
time use” and “perceived energy exertion” survey conducted by the study team.  Three 
ADLs were selected that almost all participants had engaged in and were not expected 
to vary much in difficulty throughout the year.  The ADLs survey questions collected 
information on participants’ perceived difficulty (i) traveling to and from the market, 
(ii) preparing food, and (iii) cleaning and housework due to the physical nature of the 
activity.   
The participants ranked the level of difficulty encountered on an ordinal scale 1) 
experienced no difficulty, 2) experienced some difficult, 3) needed assistance to 
complete the task, 4) avoided completing the task unless necessary, 5) not responsible 
for the task.  The responses to the ADL survey questions were collapsed into binary 
values whether the participant experienced no difficulty or any level of difficulty.  
Participants who were not responsible for a task was “missing.” Additionally, a binary 
variable indicated whether a participant responded positively to experiencing 
difficulties with any of the ADLs, and another variable captured the number of ADLs a 
participant reported difficulties. 
Feasibility, reliability, and validity 
Feasibility was assessed based on the training and tools needed to administer the 
tests and qualitative reports from the enumerators on their ability to administer the 
tests.  The reliability of each physical function test was evaluated using the Pearson 
correlation coefficient (r).  The physical function tests from each visit were compared to 
the other visits with data.  A coefficient between 0 to 0.20 was poor agreement, 0.21 to 
0.40 was fair agreement, 0.41 to 0.60 was moderate/ acceptable agreement, 0.61 to 0.80 
was substantial agreement, and 0.81 to 1.0 was near perfect agreement [51].  Reliability 
110 
 
for each ADL question was evaluated using Cohen’s kappa statistic (k).  The same 
cutoffs were used to assess agreement for the correlation coefficient  [52].  Due to the 
low agreement between UGS and Psts and low agreement of UGS between visits, only 
Psts was used to assess the validity of the ADL survey questions.  
One visit per participant per round was randomly selected for analysis for the 
validity of the ADL survey.  Each ADL question was regressed Psts variable to look for a 
significant association and several participant characteristics.  The variables were 
regressed with the rounds collapsed and by round.  In rounds where the p-value for an 
ADL question was less than 0.20, the ADL question was fit into a multivariate regress 
with Psts as the dependent variable, and significant participant characteristics from that 
round were considered in the model.  Backward stepwise regression was used to 
determine if the ADL question was significantly related to Psts when controlling for 
relevant participant characteristics.  In the PSNP round, three ADL variables; reported 
difficulty with cleaning and housework, difficulty preparing food, and difficulty with 
any ADL; were fit into models that considered age, BMI, underweight, and self-
reported long-term illness or injury as controlling variables.  In the agricultural round, 
reported difficulty in traveling to the market was placed into a model that considered 
age, BMI, underweight, food security, and self-reported long-term illness or injury as 
controlling variables. 
Ethical approval 
This study received ethical approval from the IRB at Cornell University 
(1902008596) and the International Food Policy Research Institute (DSG-19-0203P). 
111 
 
Results 
Participants were comparable between rounds in age, education, BMI, marital 
status, public work beneficiary status, and their self-reported long-term and short-term 
illnesses and injuries, Table 4.1.  The mean age of the participants was 30 years, and the 
range was 18 to 45.  The average BMI was 19.4 kg/ 2m .  However, more participants were 
underweight in the PSNP round than the agricultural (35.4% versus 31.6%), though this 
was not statistically significant (p-value = 0.6).  In addition, there was a seasonal 
difference in food security, with more participants reporting a food gap in the past six 
months during the PSNP round (33.1%) than in the agricultural round (26.8%).   
 
Table 4. 1: Study participants characteristics 
Characteristic PSNP round Agricultural round 
N 178 138 
Age (years) 30 (6.6) 30 (6.5) 
Any education 50.0% 51.0% 
BMI (kg/ 2m ) 19.4 (2.1) 19.5 (2.3) 
Underweight (BMI < 18.5 kg/ 2m ) 35.4% 31.6% 
Partner lives in household 85.4% 83.3% 
Food insecure during past 6 months 33.1% 26.8% 
Long term illness or injury 15.7% 14.5% 
Short term illness or injury 24.7% 26.8% 
Benefits from public works safety net 47.2% 44.9% 
Psts  9.9 (0.3) 9.9 (0.3) 
UGS 8.7 (1.3) 8.5 (1.4) 
Reported difficulty traveling to and from 14.0% 21.0% 
market  
Reported difficulty cleaning and 12.4% 20.3% 
housework 
Reported difficulty preparing food 12.4% 21.0% 
Reported difficulty with any ADL 17.4% 26.1% 
Number of ADLs reported difficulties   
 Zero 82.6% 73.9% 
 One 5.1% 6.5% 
 Two 3.4% 2.9% 
 Three 9.0% 16.7% 
Values are presented as means with standard deviation in parentheses or as a percent 
of the study population. 
 
112 
 
The two physical function tests were similar between rounds, Psts and UGS.  
However, more participants reported difficulty with Any ADL and each ADL during 
the agricultural round compared to the PSNP round.  During the PSNP round, 82.6% of 
participants reported experiencing no difficulties with the ADLs; this decreased to 
73.9% in the agricultural round.  In the PSNP round, 9.0% of participants reported 
difficulties with all three ADLs. This increased to 16.7% of participants in the 
agricultural round. 
The enumerators reported that the STS was feasible to administer to the study 
population.  Respondents were challenged by the STS but were able to complete it.  
There were more reported difficulties in administering the UGS, particularly finding a 
sufficient area to administer the tests.  The UGS was not feasible in some households, 
making it an unsuitable test for use in rural Ethiopia.   
The Psts had a near-perfect agreement for each visit within a round, with a 
correlation coefficient greater than 0.80 (Table 4.2) except visit 1 compared to visit 3 in 
the PSNP round.  There was substantial agreement between the values.  Overall, UGS 
had a lower agreement, with most comparisons having moderate agreement.  There 
was substantial agreement between visits 2 and 3 in the agricultural round (correlation 
coefficient = 0.62).  There was nearly always poor agreement between the UGS and Psts 
between each visit for both rounds. 
During the PSNP round, there was a near-perfect agreement to the responses to 
the ADL questions between the visits (k ranged from 0.78 to 0.97), Table 4.3.  During the 
agricultural round, only the agreement to the cleaning and housework ADL question 
decreased to substantial when comparing visit 1 to visit 2 (k = 0.77) and visit 1 to visit 3 
(k = 0.70).  Comparison of reported difficulties with any ADL between all visits during 
113 
 
the agricultural round had a substantial agreement with Cohen’s k, ranging from 0.70 to 
0.76.  There was substantial agreement between visits for the number of ADLs that the 
participants reported having difficulties for both rounds except between visits 1 and 2 
in the PSNP round, where there was near-perfect agreement. 
 
Table 4. 2: Reliability of physical function tests 
  PSNP round Agricultural round 
Visit 
Test N Pearson’s r N Pearson’s r 
numbers 
Psts      
 1 vs 2 176 0.81**** 126 0.88**** 
 1 vs 3 147 0.78*** 122 0.82**** 
 2 vs 3 147 0.81**** 129 0.84**** 
UGS      
 1 vs 2 176 0.54** 124 0.59** 
 1 vs 3 151 0.40* 117 0.40* 
 2 vs 3 152 0.47** 121 0.62** 
Psts vs UGS      
 Visit 1 175 -0.22* 127 -0.12 
 Visit 2 176 -0.10 132 -0.05 
 Visit 3 147 -0.18 124 -0.05 
* Indicates fair agreement, ** indicates moderate/ acceptable agreement, *** indicates 
substantial agreement, **** indicated near-perfect agreement. 
 
114 
 
 
 
Table 4. 3: Correlation of ADL questions 
  PSNP round Agricultural round 
ADL question Visit N Cohen’s k N Cohen’s k 
numbers 
Reported difficulty      
traveling to and 1 vs 2 174 0.89**** 128 0.81**** 
from the market 1 vs 3 150 0.81**** 124 0.83**** 
 2 vs 3 150 0.89**** 131 0.80*** 
Reported difficulty      
cleaning and 1 vs 2 175 0.97**** 128 0.77*** 
housework 1 vs 3 149 0.85**** 124 0.70*** 
 2 vs 3 149 0.88**** 131 0.84**** 
Reported difficulty      
preparing food 1 vs 2 176 0.97**** 128 0.84**** 
1 vs 3 150 0.85**** 124 0.86**** 
 2 vs 3 150 0.88**** 131 0.88**** 
Reported difficulty      
with any ADL 1 vs 2 176 0.91**** 128 0.76*** 
 1 vs 3 151 0.78*** 124 0.70*** 
 2 vs 3 151 0.84**** 131 0.71*** 
Number of ADLs      
reported difficulty 1 vs 2 177 0.91**** 137 0.74*** 
 1 vs 3 177 0.75*** 137 0.65*** 
 2 vs 3 177 0.79*** 137 0.71*** 
* Indicates fair agreement, ** indicates moderate/ acceptable agreement, *** indicates 
substantial agreement, **** indicated near-perfect agreement. 
 
In a univariate analysis with the rounds combined and stratified by round, age, 
BMI, underweight, and self-reported long-term illness or injury were significantly 
correlated to Psts. (p-value < 0.05), Table 4.4.  For every year increase in age, physical 
function as measured by the Psts decreased.  An increase in BMI was associated with 
improvements in physical function as measured by Psts, while those underweight had 
lower physical function as measured by Psts.  Those that self-reported a long-term illness 
or injury had a lower physical function than those that did not.  Among the ADL 
questions, none were significant when the rounds were assessed together.  During the 
115 
 
PSNP round, the ADL questions about cleaning and housework and preparing food, 
reporting difficulties with any ADL, and the number of ADLs one reported having 
difficulty with was significant at the a = 0.10.  Only the ADL question about traveling to 
and from the market was significant at the a = 0.10 level during the agricultural round. 
 
 
Table 4. 4: Validity of ADLs in relationship to Psts 
 All PSNP round Agricultural round 
Model 
parameter Coefficient p-value Coefficient p-value Coefficient p-value 
N 316 178 138 
Age -0.01 0.01** -0.01 0.06* -0.01 0.04** 
Any education -0.02 0.46 -0.01 0.88 -0.04 0.30 
BMI (kg/ 2m ) 0.04 <0.01*** 0.05 <0.01*** 0.03 <0.01*** 
Underweight 
(BMI < 18.5 -0.15 <0.01*** -0.18 <0.01*** -0.11 0.02** 
kg/ 2m ) 
Partner lives in 
household 0.05 0.29 0.05 0.41 0.04 0.50 
Food insecure -0.05 0.13* -0.01 0.78 -0.11 0.02** 
Long term 
illness -0.13 0.04** -0.13 0.04** -0.14 0.02** 
Short term 
illness 0.01 0.78 0.02 0.74 -0.04 0.37 
Safety net 
beneficiary 0.04 0.17* 0.03 0.51 0.06 0.16* 
Travel -0.01 0.82 0.08 0.21 -0.10 0.07* 
Cleaning 0.02 0.59 0.12 0.09* -0.06 0.29 
Preparing food 0.03 0.52 0.12 0.08* -0.05 0.33 
Any ADL 0.03 0.53 0.12 0.08* -0.05 0.33 
Number of 
ADLs 0.01 0.74 0.04 0.09* -0.03 0.17 
The mean value of ln(Psts) for each round was 9.92 with a standard deviation of 0.30 
for round 1 and 0.25 for round 2.   * Indicates statistical significance less than 0.20, 
qualifying for consideration in the multivariate model, ** Indicates statistical 
significance at <0.05, *** Indicates statistical significance <0.01 
 
  
116 
 
Multivariate models were constructed for each ADL question where the response was 
significant at a = 0.10, Table 4.5.  In the PSNP round, the ADL question for cleaning and 
housework and difficulty with any ADL were significant after controlling for BMI. The 
ADL question for preparing food and the number of ADLs a participant reported 
having difficulty with were significant after controlling for BMI and self-reported long-
term illness or injury.  For the agricultural round, the ADL question for traveling was 
significant after controlling for BMI.  In the PSNP round, all significant ADL variables 
had a positive coefficient, indicating that they were positively associated with Psts and 
had improved physical function compared to those that responded that they had no 
difficulties.  During the agricultural round, difficulty traveling to and from the market 
had a negative coefficient, indicating that those who reported difficulty had lower Psts 
and lowered physical function than those who responded that they had no difficulties.  
A summation of the tests’ feasibility, reliability, and validity is in Appendix 3. 
  
117 
 
 
Table 4. 5: Validity of ADLs in relationship to Psts 
 ADL BMI Long-term illness or 
injury 
ADL question Coefficient p-value Coefficient p-value Coefficient p-value 
PSNP round    
 Clean 0.13 0.04** 0.05 0.02**   
 Food 0.15 0.02** 0.05 <0.01*** -0.12 0.04** 
 Any ADL 0.12 0.03** 0.05 <0.01***   
 Number of 0.05 0.03** 0.05 <0.01*** -0.12 0.05** 
ADL 
Agricultural       
round 
 Travel -0.10 0.04** 0.03 <0.01***   
The mean value of Psts for each round was 9.92 with a standard deviation of 0.30 for 
round 1 and 0.25 for round 2.  * Indicates statistical significance less than 0.20, ** 
Indicates statistical significance at <0.05, *** Indicates statistical significance <0.01 
 
Discussion 
 This research found that Psts is a feasible, reliable, and valid method to measure 
physical function in adult females living in rural highland Ethiopia.  Measuring Psts 
required very little training or materials, making it a low-cost, feasible option for 
research conducted in resource-constrained settings.  Additionally, enumerators 
reported that the test was easy to administer and acceptable to the study participants.  
118 
 
Psts had high reliability with a near-perfect agreement between visits in each round.  
Additionally, Psts was associated with age, BMI, and self-reported long-term illness or 
injury, indicating it is a valid measure of physical function.  Overall, Psts is a promising 
tool to measure physical function in similar populations.  UGS was neither feasible nor 
reliable, and it is not likely a good measure of physical function in adult females living 
in rural highland Ethiopia.  The ADL survey was reliable, but it had inconclusive 
validity.  The high reliability of the subjective scoring for each ADL among the rounds 
indicates that the participants understood the question and interpreted it consistently.  
More work is needed for it to be used to measure physical function in rural Ethiopia. 
 Strengths of this research include the careful attention to the development of the 
methodologies and the administration of the STS and UGS in a novel setting.  Using STS 
as a measure of physical function is a strength of this paper.  Previous work on physical 
function in LMICs used grip strength [36,38,53].  To the best of the authors’ knowledge, 
STS has not previously been used in rural regions of LMIC.  STS is a better measure of 
physical function than grip strength because it requires muscle strength, mobility, 
balance, and coordination.  In addition, STS is more likely to reflect the necessary 
activities to fulfill a study participants’ daily roles and responsibilities.  
 There were limitations in this paper, particularly around the development of the 
ADLs.  The three activities were selected because they appeared in most responses in 
the time use pilot survey, and there was minimal seasonal variation in the frequency or 
difficulty of completing these activities.  Unfortunately, logistical and time constraints 
prevented the piloting of these questions prior to the study.  Questions added to the 
agricultural round of data collection found that these activities were the ones that were 
most likely to be altered when a participant was unwell.  However, this was not 
119 
 
reflected in that data.  There is a lack of information to help us understand why the 
ADL questionnaire had limited validity or to aid in creating a new survey.  
More research is needed to explore survey questions and activities that 
accurately characterize this population’s physical function.  Additionally, research is 
needed to understand the cause of the reported difficulties in completing ADLs.  While 
the ADL questions may not be detecting a decrease in physical function, the responses 
should not be ignored or dismissed.     
Future research that includes Psts should examine several additional factors that 
impact physical function.  This research’s nutritional status was limited to BMI, but it 
should include body composition.  Physical function is both a cause and a consequence 
of lean body mass, making it an important consideration when evaluating physical 
function.  Additionally, more detailed dietary intake and biomarkers such as 
hemoglobin should be measured.  Diet quality, especially protein and micronutrient 
intake, impact physical function [40–42].   Physical activity, including type, intensity, 
and duration, should be included in future research.  Medical history was limited to 
self-reported long or short-term illness or injury in this research, but more in-depth 
medical history would be valuable to future research, including the type, timing, 
treatment, and health-seeking behaviors around illnesses and injuries.  Finally, research 
regarding physical function needs more frequent data points over longer periods.  
There are seasonal variations in many factors that influence physical function, including 
diet [54], physical activity [46], and exposure to diseases [55].  More data points will 
help determine how quickly physical function changes in females in rural Ethiopia in 
response to different factors and if there are long-term changes.       
The STS was piloted among the enumerators prior to the start of the study.  
There was not a consistent chair height among the households of the study participants, 
120 
 
eliminating the possibility of using a chair for the seated position.  Like other cultures, 
people in rural Ethiopia will frequently sit in a deep squatted position while resting or 
working near the ground (e.g., farming activities such as planting and weeding or 
domestic work such as cooking or laundry).  As such, the study team determined that it 
would be an appropriate option for the seated position of the STS, especially since the 
Psts variable took leg length into account.  The enumerators determined that seven was 
the ideal number of repetitions.  It was challenging enough to reflect variations in 
physical function, but participants were able to complete it.   
 Measuring physical function and identifying those experiencing difficulties in 
completing their ADLS is essential in all vulnerable populations, including females 
living in rural Ethiopia.  ADLs in rural highland Ethiopia are physically demanding.  
Taking care of oneself, family, and household; and many income-generating activities 
require significant manual labor and traveling significant distances on foot.  Therefore, 
decreases in physical function could affect the wellbeing of one’s family by impacting 
household income, diet quality, and health through water, sanitation, and hygiene.   
A feasible and reliable measure of physical function is vital in addressing poverty and 
malnutrition.  It would allow healthcare providers to identify those requiring additional 
medical care and researchers to identify factors that influence physical function, 
including nutritional, behavioral, and environmental factors.  For example, public 
works safety net programs require nutritionally vulnerable people to engage in 
demanding manual labor [46,47].  The impact of this additional labor on physical 
function is not understood, but a negative relationship could reduce the program’s 
impact.  This research has demonstrated that a timed seven repetition STS is a reliable 
measure of physical function in females living in rural, highland Ethiopia. 
121 
 
  
REFERENCES 
1.  Spector, W.D.; Katz, S.; Murphy, J.B.; Fulton, J.P. The Hierarchical Relationship 
between Activities of Daily Living and Instrumental Activities of Daily Living. J. 
Chronic Dis. 1987, 40, 481–489, doi:10.1016/0021-9681(87)90004-X. 
2.  Beaton, D.E.; Schemitsch, E. Measures of Health-Related Quality of Life and 
Physical Function. Clin. Orthop. Relat. Res. 2003, 413, 90–105, 
doi:10.1097/01.blo.0000079772.06654.c8. 
3.  Tomey, K.M.; Sowers, M.R. Assessment of Physical Functioning: A Conceptual 
Model Encompassing Environmental Factors and Individual Compensation 
Strategies. Phys. Ther. 2009, 89, 705–714, doi:10.2522/ptj.20080213. 
4.  O’Neill, D.; Forman, D.E. The Importance of Physical Function as a Clinical 
Outcome: Assessment and Enhancement. Clin. Cardiol. 2020, 43, 108–117, 
doi:10.1002/clc.23311. 
5.  Seeman, T.E.; Charpentier, P.A.; Berkman, L.F.; Tinetti, M.E.; Guralnik, J.M.; Albert, 
M.; Blazer, D.; Rowe, J.W. Predicting Changes in Physical Performance in a High-
Functioning Elderly Cohort: MacArthur Studies of Successful Aging. J. Gerontol. 
1994, 49, M97–M108, doi:10.1093/geronj/49.3.M97. 
6.  Wolinsky, F.D.; Miller, D.K.; Andresen, E.M.; Malmstrom, T.K.; Miller, J.P.; Miller, 
T.R. Effect of Subclinical Status in Functional Limitation and Disability on Adverse 
Health Outcomes 3 Years Later. J. Gerontol. A. Biol. Sci. Med. Sci. 2007, 62, 101–106, 
doi:10.1093/gerona/62.1.101. 
7.  Viccaro, L.J.; Perera, S.; Studenski, S.A. Is Timed Up and Go Better Than Gait Speed 
in Predicting Health, Function, and Falls in Older Adults? J. Am. Geriatr. Soc. 2011, 
59, 887–892, doi:10.1111/j.1532-5415.2011.03336.x. 
8.  Shukla, B.; Bassement, J.; Vijay, V.; Yadav, S.; Hewson, D. Instrumented Analysis of 
the Sit-to-Stand Movement for Geriatric Screening: A Systematic Review. 
Bioengineering 2020, 7, 139, doi:10.3390/bioengineering7040139. 
9.  Gleeson, M.; Sherrington, C.; Keay, L. Exercise and Physical Training Improve 
Physical Function in Older Adults with Visual Impairments but Their Effect on 
Falls Is Unclear: A Systematic Review. J. Physiother. 2014, 60, 130–135, 
doi:10.1016/j.jphys.2014.06.010. 
10.  Fried, T.R.; Bradley, E.H.; Williams, C.S.; Tinetti, M.E. Functional Disability and 
Health Care Expenditures for Older Persons. Arch. Intern. Med. 2001, 161, 2602–
2607, doi:10.1001/archinte.161.21.2602. 
11.  Morey, M.C.; Sloane, R.; Pieper, C.F.; Peterson, M.J.; Pearson, M.P.; Ekelund, C.C.; 
Crowley, G.M.; Demark-Wahnefried, W.; Snyder, D.C.; Clipp, E.C.; et al. Effect of 
122 
 
Physical Activity Guidelines on Physical Function in Older Adults. J. Am. Geriatr. 
Soc. 2008, 56, 1873–1878, doi:10.1111/j.1532-5415.2008.01937.x. 
12.  Beswick, A.D.; Rees, K.; Dieppe, P.; Ayis, S.; Gooberman-Hill, R.; Horwood, J.; 
Ebrahim, S. Complex Interventions to Improve Physical Function and Maintain 
Independent Living in Elderly People: A Systematic Review and Meta-Analysis. 
The Lancet 2008, 371, 725–735, doi:10.1016/S0140-6736(08)60342-6. 
13.  Calvo-Ayala, E.; Khan, B.A.; Farber, M.O.; Ely, E.W.; Boustani, M.A. Interventions 
to Improve the Physical Function of ICU Survivors: A Systematic Review. Chest 
2013, 144, 1469–1480, doi:10.1378/chest.13-0779. 
14.  Imamura, K.; Kamide, N.; Sakamoto, M.; Sato, H.; Shiba, Y.; Matsunaga, A. 
Association Between Social Network and Physical Function in Community-
Dwelling Older Adults in Japan. Phys. Ther. Res. 2020, 23, 153–159, 
doi:10.1298/ptr.E10028. 
15.  Rogers, S.N.; Lowe, D.; Yueh, B.; Weymuller Jr, E.A. The Physical Function and 
Social-Emotional Function Subscales of the University of Washington Quality of 
Life Questionnaire. Arch. Otolaryngol. Neck Surg. 2010, 136, 352–357, 
doi:10.1001/archoto.2010.32. 
16.  Noppert, G.A.; Brown, C.S.; Chanti-Ketterl, M.; Hall, K.S.; Newby, L.K.; Cohen, 
H.J.; Morey, M.C. The Impact of Multiple Dimensions of Socioeconomic Status on 
Physical Functioning Across the Life Course. Gerontol. Geriatr. Med. 2018, 4, 
2333721418794021, doi:10.1177/2333721418794021. 
17.  Nilsen, C.; Agahi, N.; Kåreholt, I. Work Stressors in Late Midlife and Physical 
Functioning in Old Age. J. Aging Health 2017, 29, 893–911, 
doi:10.1177/0898264316654673. 
18.  Palumbo, A.; De Roos, A.; Cannuscio, C.; Robinson, L.; Mossey, J.; Weitlauf, J.; 
Garcia, L.; Wallace, R.; Michael, Y. Work Characteristics Associated with Physical 
Functioning in Women. Int. J. Environ. Res. Public. Health 2017, 14, 424, 
doi:10.3390/ijerph14040424. 
19.  Wiener, J.M.; Hanley, R.J.; Clark, R.; Van Nostrand, J.F. Measuring the Activities of 
Daily Living: Comparisons Across National Surveys. J. Gerontol. 1990, 45, S229–
S237, doi:10.1093/geronj/45.6.S229. 
20.  Studenski, S.; Perera, S.; Wallace, D.; Chandler, J.M.; Duncan, P.W.; Rooney, E.; 
Fox, M.; Guralnik, J.M. Physical Performance Measures in the Clinical Setting. J. 
Am. Geriatr. Soc. 2003, 51, 314–322, doi:10.1046/j.1532-5415.2003.51104.x. 
21.  Cesari, M.; Scarlata, S. What the 4-Metre Gait Speed Measures and Why It Cannot 
Replace Functional Capacity Tests. Eur. Respir. J. 2014, 43, 1819–1820, 
doi:10.1183/09031936.00207913. 
123 
 
22.  Lindemann, U.; Claus, H.; Stuber, M.; Augat, P.; Muche, R.; Nikolaus, T.; Becker, C. 
Measuring Power during the Sit-to-Stand Transfer. Eur. J. Appl. Physiol. 2003, 89, 
466–470, doi:10.1007/s00421-003-0837-z. 
23.  Smith, W.N.; Rossi, G.D.; Adams, J.B.; Abderlarahman, K.; Asfour, S.A.; Roos, B.A.; 
Signorile, J.F. Simple Equations to Predict Concentric Lower-Body Muscle Power in 
Older Adults Using the 30-Second Chair-Rise Test: A Pilot Study. Clin. Interv. 
Aging 8. 
24.  Guralnik, J.M.; Seeman, T.E.; Tinetti, M.E.; Nevitt, M.C.; Berkman, L.F. Validation 
and Use of Performance Measures of Functioning in a Non-Disabled Older 
Population: MacArthur Studies of Successful Aging. Aging Clin. Exp. Res. 1994, 6, 
410–419, doi:10.1007/BF03324272. 
25.  Curb, J.D.; Ceria-Ulep, C.D.; Rodriguez, B.L.; Grove, J.; Guralnik, J.; Willcox, B.J.; 
Donlon, T.A.; Masaki, K.H.; Chen, R. Performance-Based Measures of Physical 
Function for High-Function Populations. J. Am. Geriatr. Soc. 2006, 54, 737–742, 
doi:10.1111/j.1532-5415.2006.00700.x. 
26.  Katz, S. Assessing Self-Maintenance: Activities of Daily Living, Mobility, and 
Instrumental Activities of Daily Living. J. Am. Geriatr. Soc. 1983, 31, 721–727, 
doi:10.1111/j.1532-5415.1983.tb03391.x. 
27.  Paddick, S.-M.; Gray, W.K.; Ogunjimi, L.; lwezuala, B.; Olakehinde, O.; Kisoli, A.; 
Kissima, J.; Mbowe, G.; Mkenda, S.; Dotchin, C.L.; et al. Validation of the 
Identification and Intervention for Dementia in Elderly Africans (IDEA) Cognitive 
Screen in Nigeria and Tanzania. BMC Geriatr. 2015, 15, 53, doi:10.1186/s12877-015-
0040-1. 
28.  Algarni, A.D.; Alrabai, H.M.; Al-Ahaideb, A.; Kachanathu, S.J.; AlShammari, S.A. 
Arabic Translation, Cultural Adaptation, and Validation Study of Knee Outcome 
Survey: Activities of Daily Living Scale (KOS-ADLS). Rheumatol. Int. 2017, 37, 
1585–1589, doi:10.1007/s00296-017-3776-4. 
29.  Bucks, R.S.; Ashworth, D.L.; Wilcock, G.K.; Siegfried, K. Assessment of Activities of 
Daily Living in Dementia: Development of the Bristol Activities of Daily Living 
Scale. Age Ageing 1996, 25, 113–120, doi:10.1093/ageing/25.2.113. 
30.  Fitzpatrick, A.L.; Buchanan, C.K.; Nahin, R.L.; DeKosky, S.T.; Atkinson, H.H.; 
Carlson, M.C.; Williamson, J.D.; for the Ginkgo Evaluation of Memory (GEM) 
Study Investigators Associations of Gait Speed and Other Measures of Physical 
Function with Cognition in a Healthy Cohort of Elderly Persons. J. Gerontol. Ser. A 
2007, 62, 1244–1251, doi:10.1093/gerona/62.11.1244. 
31.  Wang, L.; Larson, E.B.; Bowen, J.D.; van Belle, G. Performance-Based Physical 
Function and Future Dementia in Older People. Arch. Intern. Med. 2006, 166, 1115–
1120, doi:10.1001/archinte.166.10.1115. 
124 
 
32.  Busch, T. de A.; Duarte, Y.A.; Pires Nunes, D.; Lebrão, M.L.; Satya Naslavsky, M.; 
dos Santos Rodrigues, A.; Amaro, E. Factors Associated with Lower Gait Speed 
among the Elderly Living in a Developing Country: A Cross-Sectional Population-
Based Study. BMC Geriatr. 2015, 15, 35, doi:10.1186/s12877-015-0031-2. 
33.  Jones, S.E.; Kon, S.S.C.; Canavan, J.L.; Patel, M.S.; Clark, A.L.; Nolan, C.M.; Polkey, 
M.I.; Man, W.D.-C. The Five-Repetition Sit-to-Stand Test as a Functional Outcome 
Measure in COPD. Thorax 2013, 68, 1015–1020, doi:10.1136/thoraxjnl-2013-203576. 
34.  White, D.K.; Master, H. Patient Reported Measures of Physical Function in Knee 
Osteoarthritis. Rheum. Dis. Clin. North Am. 2016, 42, 239–252, 
doi:10.1016/j.rdc.2016.01.005. 
35.  Urits, I.; Smoots, D.; Franscioni, H.; Patel, A.; Fackler, N.; Wiley, S.; Berger, A.A.; 
Kassem, H.; Urman, R.D.; Manchikanti, L.; et al. Injection Techniques for Common 
Chronic Pain Conditions of the Foot: A Comprehensive Review. Pain Ther. 2020, 9, 
145–160, doi:10.1007/s40122-020-00157-5. 
36.  Bernard, C.; Dabis, F.; de Rekeneire, N. Physical Function, Grip Strength and 
Frailty in People Living with HIV in Sub-Saharan Africa: Systematic Review. Trop. 
Med. Int. Health TM IH 2017, 22, 516–525, doi:10.1111/tmi.12852. 
37.  Akosile, C.O.; Mgbeojedo, U.G.; Maruf, F.A.; Okoye, E.C.; Umeonwuka, I.C.; 
Ogunniyi, A. Depression, Functional Disability and Quality of Life among 
Nigerian Older Adults: Prevalences and Relationships. Arch. Gerontol. Geriatr. 
2018, 74, 39–43, doi:10.1016/j.archger.2017.08.011. 
38.  Pieterse, S.; Manandhar, M.; Ismail, S. The Association between Nutritional Status 
and Handgrip Strength in Older Rwandan Refugees. Eur. J. Clin. Nutr. 2002, 56, 
933–939, doi:10.1038/sj.ejcn.1601443. 
39.  Central Statistical Agency (CSA) Ethiopia and ICF Ethiopia Demographic and 
Health Survey; CSA and ICF: Addis Ababa, Ethiopia and Rockville, Maryland, 
USA, 2016; 
40.  McLoughlin, D.M.; Spargo, E.; Wassif, W.S.; Newham, D.J.; Peters, T.J.; Lantos, 
P.L.; Russell, G.F.M. Structural and Functional Changes in Skeletal Muscle in 
Anorexia Nervosa. Acta Neuropathol. (Berl.) 1998, 95, 632–540, 
doi:10.1007/s004010050850. 
41.  Cimolin, V.; Galli, M.; Vismara, L.; Vimercati, S.L.; Precilios, H.; Cattani, L.; De 
Souza, S.F.; Petroni, M.L.; Capodaglio, P. Gait Analysis in Anorexia and Bulimia 
Nervosa. J. Appl. Biomater. Funct. Mater. 2013, 11, 122–128, 
doi:10.5301/JABFM.5000152. 
42.  Asamane, E.A.; Greig, C.A.; Thompson, J.L. The Association between Nutrient 
Intake, Nutritional Status and Physical Function of Community-Dwelling 
Ethnically Diverse Older Adults. BMC Nutr. 2020, 6, 36, doi:10.1186/s40795-020-
00363-6. 
125 
 
43.  Cesari, M.; Penninx, B.W.J.H.; Pahor, M.; Lauretani, F.; Corsi, A.M.; Williams, G.R.; 
Guralnik, J.M.; Ferrucci, L. Inflammatory Markers and Physical Performance in 
Older Persons: The InCHIANTI Study. J. Gerontol. Ser. A 2004, 59, M242–M248, 
doi:10.1093/gerona/59.3.M242. 
44.  Brinkley, T.E.; Leng, X.; Miller, M.E.; Kitzman, D.W.; Pahor, M.; Berry, M.J.; Marsh, 
A.P.; Kritchevsky, S.B.; Nicklas, B.J. Chronic Inflammation Is Associated with Low 
Physical Function in Older Adults Across Multiple Comorbidities. J. Gerontol. A. 
Biol. Sci. Med. Sci. 2009, 64A, 455–461, doi:10.1093/gerona/gln038. 
45.  Aiello, A.E.; Haan, M.N.; Pierce, C.M.; Simanek, A.M.; Liang, J. Persistent Infection, 
Inflammation, and Functional Impairment in Older Latinos. J. Gerontol. Ser. A 
2008, 63, 610–618, doi:10.1093/gerona/63.6.610. 
46.  Berhane, G.; Golan, J.; Hirvonen, K.; Hoddinott, J.F.; Kim, Sunny; Seyoum Tafesse, 
Alemayehu; Abay, Kibrewossen; Assefa, Thomas; Habte, Yetmwork Evaluation of 
the Nutrition-Sensitive Features of the Fourth Phase of Ethiopia’s Productive 
Safety Net Programme; 0 ed.; International Food Policy Research Institute: 
Washington, DC, 2020; 
47.  World Bank The State of Social Safety Nets 2018; World Bank: Washington, DC, 
2018; 
48.  FAO and FHI 360 Minimum Dietary Diversity for Women- A Guide to 
Measurement; FAO: Rome, 2016; 
49.  Takai, Y.; Ohta, M.; Akagi, R.; Kanehisa, H.; Kawakami, Y.; Fukunaga, T. Sit-to-
Stand Test to Evaluate Knee Extensor Muscle Size and Strength in the Elderly: A 
Novel Approach. J. Physiol. Anthropol. 2009, 28, 123–128, doi:10.2114/jpa2.28.123. 
50.  Gurland, B.J.; Maurer, M.S. Life and Works of Sidney Katz, MD: A Life Marked by 
Fundamental Discovery. J. Am. Med. Dir. Assoc. 2012, 13, 764–765, 
doi:10.1016/j.jamda.2012.09.003. 
51.  Bull, F.C.; Maslin, T.S.; Armstrong, T. Global Physical Activity Questionnaire 
(GPAQ): Nine Country Reliability and Validity Study. J. Phys. Act. Health 2009, 6, 
790–804, doi:10.1123/jpah.6.6.790. 
52.  Watson, P.F.; Petrie, A. Method Agreement Analysis: A Review of Correct 
Methodology. Theriogenology 2010, 73, 1167–1179, 
doi:10.1016/j.theriogenology.2010.01.003. 
53.  Hoddinott, J.; Behrman, J.R.; Maluccio, J.A.; Melgar, P.; Quisumbing, A.R.; 
Ramirez-Zea, M.; Stein, A.D.; Yount, K.M.; Martorell, R. Adult Consequences of 
Growth Failure in Early Childhood. Am. J. Clin. Nutr. 2013, 98, 1170–1178, 
doi:10.3945/ajcn.113.064584. 
54.  Baye, K.; Hirvonen, K. Seasonality: A Missing Link in Preventing Undernutrition. 
Lancet Child Adolesc. Health 2020, 4, e3, doi:10.1016/S2352-4642(19)30343-8. 
126 
 
55.  Altizer, S.; Dobson, A.; Hosseini, P.; Hudson, P.; Pascual, M.; Rohani, P. Seasonality 
and the Dynamics of Infectious Diseases. Ecol. Lett. 2006, 9, 467–484, 
doi:10.1111/j.1461-0248.2005.00879.x. 
 
 
  
127 
 
CHAPTER 5 
CONCLUSION 
 This dissertation validated three novel techniques to evaluate nutritional status 
in Ethiopia.  Chapter 1 provided background information on why these tools are 
needed.  Chapter 2 assessed the validity of bioelectrical impedance analysis (BIA) and 
skinfold thickness (SFT) prediction equations to calculate the fat mass (FM), fat-free 
mass (FFM), and percent body fat in Ethiopian adults.  Chapter 3 evaluated two tools to 
quantify the proportion of time spent and moderate or vigorous physical activity 
(MVPA) in females living in rural, highland Ethiopia.  It created a calibrated model of 
the 24-hour recall of time use and perceived exertion.  Chapter 4 evaluated three 
methods to measure physical function in females in rural, highland Ethiopia. 
Summary of results 
 In Chapter 2, existing BIA and SFT prediction equations to measure body 
composition in Ethiopian adults were validated.  Lohman’s criteria for cross-validation 
and the Kolmogorov-Smirnov test were used to determine validity.  The existing 
equations needed to meet all the criteria to be valid.  Only one existing prediction 
equation for BIA was valid to measure FM and FFM in adult males.  This was the 
Modeling Epidemiologic Transition Study (METS).  None of the five equations 
evaluated for BIA were valid for measuring body composition in adult females.  
Additionally, none of the three SFT prediction equations were valid for measuring body 
density in males or females.  In general, the existing prediction equations 
underestimated percent body fat.   
Subsequently, a new BIA prediction equation for females and new gender-
specific SFT prediction equations were created.  The new BIA prediction equation met 
128 
 
all the criteria and was valid.  Unfortunately, the females' new SFT prediction equations 
did not meet the standard estimate of error (SEE) or total error (TE) criteria.  In 
addition, the Bland-Altman plots of the new SFT prediction equations revealed some 
bias.  However, the new equations performed significantly better on all the validation 
criteria than the existing ones, making them valuable tools for researchers, 
policymakers, and healthcare workers. 
 In Chapter 3, the GPAQ and the 24-hour recall of time use and perceived 
exertion were evaluated for reliability and validity using triaxial accelerometry.  The 
surveys were considered valid if the proportion of time spent at MVPA had a 
moderate/acceptable agreement (r ≥ 0.41) with the proportion of time at MVPA 
recorded by accelerometry.  The Global Physical Activity Questionnaire (GPAQ) had 
high reliability but was not valid because the agreement with accelerometry was poor.  
A subsequent analysis of discordancy in responses found that participants who spent 
less time in MVPA tended to overestimate their time spent at MVPA.  In contrast, those 
that spent the greatest proportion of their time in MVPA underestimated their efforts. 
 The 24-hour recall was not evaluated for reliability because the recall period for 
the survey was different between the two visits.  The proportion of time spent at MVPA 
according to the 24-hour recall only had a fair agreement with the proportion of time 
spent at MVPA recorded by accelerometry.  However, when calibrated by the 
respondent’s BMI, the agreement between the 24-hour recall and the accelerometry 
increased to moderate/acceptable (r = 0.53).  Therefore, the calibrated 24-hour recall is a 
valid tool to estimate the proportion of time spent in MVPA in females in rural 
highland Ethiopia. 
129 
 
 Chapter 4 assessed the feasibility, reliability, and validity of various tools to 
evaluate physical function in females in rural highland Ethiopia.  There were two direct 
measures of physical function, the STS and the usual gait speed (UGS), and an activities 
of daily living (ADL) questionnaire.  The ADL questionnaire asked if participants had 
experienced physical difficulty doing housework, preparing meals, and traveling to and 
from the market.  The time from the STS was used to calculate Psts, which accounted for 
anthropometric measurements.  Psts had substantial reliability (r ³ 0.78) and was 
significantly associated with other covariates related to physical function, including age, 
BMI, and self-reported long-term illness/injury.  UGS has a fair to moderate/acceptable 
agreement between the visits in each round.  However, UGS and Psts had a poor to fair 
agreement.  Therefore, Psts was a more reliable and valid measure of physical function 
than UGS.  The ADL questionnaire had a substantial agreement between visits in each 
round.  However, the validity of the ADL was inconclusive because of mixed results in 
its association with Psts.  During the PSNP round, participants that reported difficulties 
with cleaning, preparing food, any ADL, and the number of ADL had improved 
physical function compared to those that did not.  During the agricultural round, those 
that reported difficulties traveling to and from the market had decreased physical 
function compared to those that did not.  These associations were only strengthened 
when controlling for BMI or BMI and self-reported long-term illness/injury. 
Strengths and limitations 
 The main strength of this research is its innovation of applying existing 
nutritional science research methodologies to a population in Ethiopia.  Research 
methods and tools have progressed in other disciplines of nutritional science, but the 
methods have not been applied in LMIC.  This research demonstrates that using these 
130 
 
methodologies in LMIC such as Ethiopia is feasible.  The inclusion of these 
methodologies will provide a holistic view of malnutrition.  Demonstrating the 
feasibility, reliability, and validity of these methods in LMIC is only a first step.     
 There were many limitations in conducting this work.  For all three chapters, the 
study population was at the lower end of the sample size range than similar work due 
to logistic and financial constraints.  The physical activity findings would be more 
substantial if accelerometry were combined with heart rate monitoring.  The physical 
function study would have been improved with more time for scoping and piloting the 
ADL questionnaire.   
These newly validated tools were used in isolation but are interconnected.  Body 
composition equations would be more accurate with the inclusion of physical activity 
levels.  Physical activity and physical function would be better understood if adjusted 
for percent body fat rather than just BMI.  Additionally, the analyses did not include 
dietary intake.  There was no nutritional data available for the body composition 
analysis.  Only diet diversity and food security were available for the physical activity 
and physical function analyses.  Dietary intake, including energy and micronutrients 
intake, contributes to a person's ability to undertake physical activity and physical 
function.   
Finally, data for chapters 3 and 4 were only collected at two time points within a 
single year.  With only this amount of data, it is difficult to distinguish between 
seasonal variation, long-term trends, and unique changes in physical activity and 
physical function.  More frequent data points over several years are needed to 
characterize the relationship between data points. 
131 
 
Contribution to research and future directions 
 This research will provide researchers with better tools to characterize and 
address the double burden of malnutrition in rural highland Ethiopia.  The results can 
be used to identify those that are malnourished based upon their percent body fat.  
Body composition is a better measure of nutritional status than BMI because it measures 
FM, FFM, and percent body fat.  Additionally, it will allow researchers to assess 
physical activity and function, both causes and consequences of malnutrition.  Physical 
activity is an essential component in determining dietary needs.  Physical function will 
allow researchers to understand the functional significance of malnutrition.   
 The new SFT prediction equations for males and females would be improved 
with additional participants.  In addition, the equations would be more generalizable 
within Ethiopia if the additional participants' average BMI was nationally 
representative and included persons residing in rural regions of the country.   
The new equations are more useful for researchers than existing equations or 
BMI alone.  In one study, BMI miscategorized almost half of the participants as normal 
weight when they had excess adipose tissue [1].  As a result, it is a poor indicator of 
malnutrition in Ethiopia, resulting in researchers not adequately identifying people that 
are malnourished.  Using percent body fat will better identify those with insufficient or 
excessive adipose stores and in need of nutrition interventions.   
Body composition is a better outcome measure for nutrition research within 
Ethiopia.  Existing nutrition research focuses on increasing BMI but knowing whether a 
study participant is putting on additional fat mass versus fat-free mass is significantly 
more informative about the overall impact on a person’s well-being.  The goal of 
nutrition research and nutrition interventions for most people within Ethiopia is to have 
132 
 
a healthy proportion of fat mass and fat-free mass, which is impacted by dietary 
composition and physical activity.     
Impact evaluations of the PSNP have found that despite its impact on economic 
resiliency and food security, there has not been a corresponding improvement in 
nutritional status as measured by BMI.  However, there may be other improvements in 
nutritional status that are not captured by the percent of the participants categorized as 
normal by their BMI.  Inclusion of body composition in future impact evaluations may 
determine that a larger percentage of the population has sufficient adipose tissue at 
baseline, and increases in their BMI are unnecessary.  Alternatively, it may determine 
that the PSNP improved the participants' percent body fat and that the impact on 
nutritional status is not reflected in BMI.  BMI is only a rough measure of nutritional 
status, and body composition adds much-needed detail characterizing malnutrition in 
Ethiopia. 
Additionally, the new prediction equations can be used by healthcare workers to 
screen people for diseases associated with excess adiposity, which is crucial because the 
prevalence of cardiovascular disease and diabetes in Ethiopia is increasing [2,3].  This 
research also adds to the literature regarding the inadequacy of existing prediction 
equations in measuring body composition in sub-Saharan Africa [4,5].  This research 
should be replicated in other countries in Africa to understand and address 
malnutrition better. 
 The validation of a calibrated 24-hour recall of time use and perceived exertion 
will enable researchers in Ethiopia to quantify the proportion of time spent at MVPA, 
identifying who is most active and which activities require the most effort.  
Subsequently, they will be able to assess dietary needs better.  Additionally, this 
133 
 
calibrated model will determine the impact of the Productive Safety Net Programme 
(PSNP) and agricultural practices on physical activity.   
The PSNP and agricultural work requires significant manual labor.  If this 
increases in the proportion of time spent at MVPA among a population who is already 
food insecure and impoverished, the program may worsen malnutrition in adults living 
in rural highland Ethiopia.  This calibrated measure will determine if there is a 
significant difference in the proportion of time spent at MVPA between PSNP 
beneficiaries and non-beneficiaries across seasons.  This difference in MVPA may 
contribute to the PSNP not having its intended impact on BMI. 
Additionally, this research demonstrated that it is feasible to validate a survey 
tool to quantify the proportion of time spent at MVPA.  Future research should validate 
the calibrated 24-hour recall in other, similar contexts.  If this calibrated model from this 
study is not valid, context-specific calibrated models should be created.  Physical 
activity is an essential contributing factor to nutritional status.  Therefore, it should not 
be overlooked in LMIC. 
 This research demonstrated that Psts is a feasible, reliable, and valid method to 
measure physical function in active, young adults in LMIC.  Unfortunately, there is 
minimal research into physical function in LMIC.  The few existing studies focused on 
aging populations.  However, life in rural parts of LMIC is very physically demanding.  
Therefore, decreases in physical function can have significant repercussions for the 
individual and other household members due to the inability to fulfill the ADLs and 
participate in income-generating activities.   
 The inclusion of the STS in the PSNP impact evaluation may provide crucial 
information on the functional impacts on participation.  As stated before, the PSNP 
required strenuous work on labor-intensive projects.  Participating in those projects 
134 
 
may negatively impact physical function, preventing household members from 
participating in their ADLs.  Additionally, domestic work may be redistributed among 
other household members, resulting in declines in their physical function.  
Alternatively, declines in physical function may stop people from participating in the 
PSNP, preventing them from benefitting from the program.  Understanding the 
relationship between physical function and the PSNP is crucial because it may have far-
reaching consequences for beneficiary households. 
 Despite having inconclusive validity, the ADL questionnaire demonstrated that 
participants had difficulty completing their ADLs.  This finding should not be 
disregarded, and additional research should be conducted to determine the underlying 
causes.  Any barrier to fulfilling ADLs can negatively affect the health and well-being of 
the respondent and their household.  Furthermore, understanding the underlying 
causes will allow interventions to address them, enabling more people in rural highland 
Ethiopia to fulfill their ADLs and improve their quality of life. 
 Finally, future research should examine body composition, physical activity, 
physical function, and diet quality in the same study population.  All these 
characteristics are interrelated with overlapping drivers.  Therefore, understanding 
them is crucial to addressing malnutrition in LMIC.  While underweight is associated 
with higher morbidity and mortality rates, BMI is an insufficient measure in Ethiopia, 
meaning the significance of having a BMI less than 18.5 kg/ 2m  in this context is not 
known.  Therefore, it is more important to determine whether a person has inadequate 
or excess adiposity.  Additionally, excess adiposity is caused by a sedentary lifestyle 
and excess energy intake.  Therefore, quantifying MVPA is a crucial component in 
understanding BMI-related health risks.   
135 
 
FFM and physical activity affect physical function, which can affect physical 
activity and then FFM.  Understanding the functional significance of malnutrition is a 
critical component in determining who is malnourished and how to address it.  Without 
understanding it, researchers and policymakers cannot fully address malnutrition in 
rural highland Ethiopia.  The tools validated in this study, BIA and SFT prediction 
equations, the calibrated 24-hour recall of time use and perceived energy exertion, and 
the STS, will be invaluable. 
 The work should be repeated in more populations, with more participants, and 
over extended periods to assess changes in body composition and physical function.  
With an unlimited budget and time, I would like to run an extensive study that has 
repeated measures of body composition, physical activity, physical function, and 
detailed dietary intake [6,7].  In addition to calories, macronutrients, and 
micronutrients, quantifying amino acid intake is necessary due to their different roles in 
muscle protein synthesis [8–10].  Additional measures and biomarkers of skeletal 
muscle well-being would be interesting to include [11–13].  Research has demonstrated 
there are genetic differences in body composition and physical function.  However, 
existing research has mainly focused on populations of European descent and not those 
of African descent [14]. 
 Addressing malnutrition in Ethiopia is challenging.  Efforts are complicated 
because researchers have limited tools to identify malnutrition, factors that influence 
malnutrition, and the functional significance of being malnourished.  This results of this 
dissertation provides researchers with innovative valid tools to characterize the causes 
and consequences of malnutrition.  The additional information collected using these 
tools will be invaluable in creating and evaluating interventions to improve nutrition in 
Ethiopia. 
136 
 
 
 
REFERENCES 
1.  Sinaga, M.; Yemane, T.; Tegene, E.; Lidstrom, D.; Belachew, T. Performance of 
Newly Developed Body Mass Index Cut-off for Diagnosing Obesity among 
Ethiopian Adults.  J Physiol Anthropol 2019, 38, 14, doi:10.1186/s40101-019-0205-2. 
2.  Angaw, D.A.; Ali, R.; Tadele, A.; Shumet, S. The Prevalence of Cardiovascular 
Disease in Ethiopia: A Systematic Review and Meta-Analysis of Institutional and 
Community-Based Studies.  BMC Cardiovasc Disord 2021, 21, 37, 
doi:10.1186/s12872-020-01828-z. 
3.  Bishu, K.G.; Jenkins, C.; Yebyo, H.G.; Atsbha, M.; Wubayehu, T.; Gebregziabher, 
M. Diabetes in Ethiopia: A Systematic Review of Prevalence, Risk Factors, 
Complications, and Cost.  Obes 2019, 15, 100132, doi:10.1016/j.obmed.2019.100132. 
4.  Long, A.E.; Prewitt, T.E.; Kaufman, J.S.; Rotimi, C.N.; Cooper, R.S.; McGee, D.L. 
Weight–Height Relationships among Eight Populations of West African Origin: 
The Case against Constant BMI Standards. Int J Obes 1998, 22, 842–846, 
doi:10.1038/sj.ijo.0800670. 
5.  Luke, A.; Bovet, P.; Forrester, T.E.; Lambert, E.V.; Plange-Rhule, J.; Dugas, L.R.; 
Durazo-Arvizu, R.A.; Kroff, J.; Richie, W.N.; Schoeller, D.A. Prediction of Fat-Free 
Mass Using Bioelectrical Impedance Analysis in Young Adults from Five 
Populations of African Origin. Eur J Clin Nutr 2013, 67, 956–960, 
doi:10.1038/ejcn.2013.123. 
6.  Gebru, Y.A.; Hyun-II, J.; Young-Soo, K.; Myung-Kon, K.; Kwang-Pyo, K. Variations 
in Amino Acid and Protein Profiles in White versus Brown Teff (Eragrostis Tef) 
Seeds, and Effect of Extraction Methods on Protein Yields.  Foods 2019, 8, 202, 
doi:10.3390/foods8060202. 
7.  Blum, J.; Epstein, R.; Watts, S.; Thalacker-Mercer, A. Importance of Nutrient 
Availability and Metabolism for Skeletal Muscle Regeneration.  Front Physiol 2021, 
12, 696018, doi:10.3389/fphys.2021.696018. 
8.  Walker, D.K.; Dickinson, J.M.; Timmerman, K.L.; Drummond, M.J.; Reidy, P.T.; 
Fry, C.S.; Gundermann, D.M.; Rasmussen, B.B. Exercise, Amino Acids and Aging 
in the Control of Human Muscle Protein Synthesis.  Med Sci Sports Exerc 2011, 43, 
2249–2258, doi:10.1249/MSS.0b013e318223b037. 
9.  Katsanos, C.S.; Kobayashi, H.; Sheffield-Moore, M.; Aarsland, A.; Wolfe, R.R. A 
High Proportion of Leucine Is Required for Optimal Stimulation of the Rate of 
Muscle Protein Synthesis by Essential Amino Acids in the Elderly.  Am J Physiol 
Endocrinol Metab 2006, 291, E381–E387, doi:10.1152/ajpendo.00488.2005. 
137 
 
10.  Millward, D.J.; Garlick, P.J.; Nnanyelugo, D.O.; Waterlow, J.C. The Relative 
Importance of Muscle Protein Synthesis and Breakdown in the Regulation of 
Muscle Mass.  Biochemical Journal 1976, 156, 185–188, doi:10.1042/bj1560185. 
11.  Nie, J.; Tong, T.K.; George, K.; Fu, F.H.; Lin, H.; Shi, Q. Resting and Post-Exercise 
Serum Biomarkers of Cardiac and Skeletal Muscle Damage in Adolescent Runners.  
Scand J Med Sci Sports 2011, 21, 625–629, doi:10.1111/j.1600-0838.2010.01096.x. 
12.  Finsterer, J. Biomarkers of Peripheral Muscle Fatigue during Exercise.  BMC 
Musculoskelet Disord 2012, 13, 218, doi:10.1186/1471-2474-13-218. 
13.  Siracusa, J.; Koulmann, N.; Banzet, S. Circulating MyomiRs: A New Class of 
Biomarkers to Monitor Skeletal Muscle in Physiology and Medicine.  J Cachexia 
Sarchopeni 2018, 9, 20–27, doi:10.1002/jcsm.12227. 
14.  Campbell, M.C.; Tishkoff, S.A. African Genetic Diversity: Implications for Human 
Demographic History, Modern Human Origins, and Complex Disease Mapping.  
Annu Rev Genomics Hum Genet 2008, 9, 403–433, 
doi:10.1146/annurev.genom.9.081307.164258. 
 
 
 
  
138 
 
APPENDIX 1: Supplementary body composition tables 
 
 
  
139 
 
 
  
140 
 
 
 
 
 
 
 
  
141 
 
APPENDIX 2: Aim 2 sensitivity analyses 
 
Table A2.1: Study population characteristics by round in which participants appear 
Characteristic Both round PSNP Round Agricultural 
only Round only 
N 64 25 27 
Age 30 29 31 
Any education 47.7 % 48.0% 37.0% 
BMI (kg/ 2m ) 19.5 19.2 19.3 
Underweight (BMI < 18.5 kg/ 2m ) 26.6 % 44.0 % 29.6% 
Partner lives in household 87.5 % 88.0 % 81.5 % 
Food insecure during past 6 31.2 % 28.0 % 26.9 % 
months 
Benefits from public works 46.9 % 44.0 % 48.1% 
safety net 
 
 
  
142 
 
Table A2.2: GPAQ validity disaggregated by round in which participants participated 
Exertion level Pearson Spearman’s  GPAQ Accelerometry  correlation rank  coefficient correlation coefficient 
PSNP Round     
Both rounds     
N  64 64   
Proportion of time at 
sedentary/light exertion 92.9 90.0 0.07 0.12 
Proportion of time at 
moderate exertion  5.7 % 9.6 % 0.02 0.10 
Proportion of time at 
vigorous exertion 1.4 % 0.4 % 0.08 0.12 
Proportion of time at 
MVPA 7.1 % 10.0 % 0.07 0.12 
PSNP Round only     
N 25 25   
Proportion of time at 
sedentary/light exertion 89.3 90.0 0.08 0.05 
Proportion of time at 
moderate exertion 8.9 % 9.6 % -0.09 -0.06 
Proportion of time at 
vigorous exertion 1.7 % 0.4 % 0.49 0.37 
Proportion of time at 
MVPA 10.7 % 10.0 % 0.08 0.05 
Agricultural Round      
Both rounds     
N 64 64   
Proportion of time at 
sedentary/light exertion 87.1 91.0 -0.15 -0.14 
Proportion of time at 
moderate exertion 11.7 % 8.5 % -0.14 -0.16 
Proportion of time at 
vigorous exertion 1.2 % 0.5 % -0.14 0.01 
Proportion of time at 
MVPA 12.9 % 10.0 % -0.15 -0.14 
Agricultural Round only     
N 27 27   
Proportion of time at 
sedentary/light exertion 83.7 89.0 -0.14  -0.10 
Proportion of time at 
moderate exertion 13.1 % 10.4 % -0.29 -0.20 
Proportion of time at 
vigorous exertion 3.2 % 0.6 % 0.19 0.25 
Proportion of time at 
MVPA 16.3 % 11.0 % -0.14 -0.10 
 
143 
 
 
 
Table A2.3: 24-hour recall validity disaggregated by round in which participants 
participated 
Exertion level Spearman’s 
 24-hour Pearson 
 recall Accelerometry  correlation 
rank 
coefficient correlation coefficient 
PSNP Round     
Both rounds     
N  64 64   
Proportion of time at 
sedentary/light exertion 86.2 % 89.1 % 0.34 0.32 
Proportion of time at 
moderate exertion  13.2 % 10.5 % 0.27 0.27 
Proportion of time at 
vigorous exertion 0.6 % 0.4 % 0.02 0.20 
Proportion of time at MVPA 13.8 % 10.9 % 0.34 0.32 
PSNP Round only     
N 25 25   
Proportion of time at 
sedentary/light exertion 76.3 % 88.8 % 0.06 0.07 
Proportion of time at 
moderate exertion 23.4 % 10.9 % 0.06 0.07 
Proportion of time at 
vigorous exertion 0.3 % 0.4 % -0.07 0.12 
Proportion of time at MVPA 23.7 % 11.2 % 0.06 0.07 
Agricultural Round     
Both rounds     
N 64 64   
Proportion of time at 
sedentary/light exertion 75.4 % 85.9 % 0.13 0.15 
Proportion of time at 
moderate exertion 20.6 % 13.3 % 0.09 0.09 
Proportion of time at 
vigorous exertion 4.0 % 0.8 % 0.06 0.21 
Proportion of time at MVPA 24.6 % 14.1 % 0.13 0.15 
Agricultural Round only     
N 27 27   
Proportion of time at 
sedentary/light exertion 62.8 % 83.1 % 0.25 0.18 
Proportion of time at 
moderate exertion 29.3 % 16.1 % 0.12 0.03 
Proportion of time at 
vigorous exertion 7.9 % 0.9 % 0.15 0.20 
Proportion of time at MVPA 37.2 % 16.9 % 0.25 0.17 
 
  
144 
 
APPENDIX 3: Physical function  
Table A3.1: Summation of results 
 Feasibility Reliability Validity 
Psts    
UGS   N/A 
ADL    Inconclusive 
A grey cell indicates the test met the criteria, N/A indicates that was 
not evaluated 
 
  
145 
 
 
 
Table A3.2: Study population characteristics sensitivity analysis 
 PSNP round Agricultural round  
One Both One Both 
Characteristic 
round rounds round rounds 
N 51 127 11 127 
Age 29 30 30 30 
Any education 41.2% 53.5% 45.5% 51.2% 
BMI (kg/ 2m ) 18.7 19.6 20.2 19.5 
Underweight (BMI < 18.5 kg/ 2m ) 51.0% 29.1% 27.3% 32.0% 
Partner lives in household 88.2% 84.3% 72.7% 84.3% 
Food insecure during past 6 27.5% 35.4% 18.2% 27.6% 
months 
Benefits from public works safety 33.3% 52.8% 18.2% 47.2% 
net 
Psts 9.9 9.9 9.8 9.9 
UGS (seconds) 8.7 8.7 8.6 8.5 
ADL travel 13.7% 14.2% 18.2% 21.2% 
ADL clean 9.8% 13.4% 27.3% 19.7% 
ADL food 9.8% 13.4% 27.3% 20.5% 
Any ADL 17.6% 17.3% 27.3% 26.0% 
Long term illness 19.6% 14.2% 9.1% 15.0% 
Short term illness 13.7% 29.1% 18.2% 27.6% 
 
 
146 
 
Table A3.3:  Validity of ADLs in relationship to the power sit to stand variable for 
those in both rounds sensitivity analysis 
 All PSNP round Agricultural round 
ADL 
Coefficient p-value Coefficient p-value Coefficient p-value 
Question 
N  254 127 127 
Travel -1007 0.31 705 0.66 -2182 0.08 
Cleaning 62 0.95 2691 0.10 -1786 0.16 
Preparing 219 0.83 2770 0.09* -1507 0.23 
food 
Any ADL -99 0.91 2105 0.16 -1652 0.15 
 
 
 
147