Interrelationships Among Gastrointestinal Infection, Stunting

Interrelationships among gastrointestinal infection, stunting, and their socio-ecological determinants in impoverished Panamanian preschool children: A spatio-temporal and ecohealth approach

Carli Halpenny

Institute of Parasitology & School of Environment

McGill University,

Montreal, Canada

2012

A thesis submitted to McGill University in partial fulfilment of the requirements of the degree of Doctor of Philosophy

Abstract

Background: Although growth stunting, height for age Z score (HAZ) <-2SD, results from sustained poor diet and frequent infection both of which are influenced by social and biophysical factors, few studies have used a transdisciplinary ecohealth framework for a comprehensive analysis of this relationship. Objective: To examine the interrelationships between preschool child stunting and gastrointestinal infections within the biophysical, social and spatial context of extreme poverty among the Ngäbe in Western Panama where conditional food voucher (FV) and cash transfer (CT) programs occurred. Methods: A 16-mo longitudinal study of 356 preschool children involved two reinfection cycles following albendazole treatment. Data collection included repeated fecal samples, household socio-behavioural questionnaires, multiple dietary records and anthropometric measures, water samples, GPS and participatory workshops. An asset-based household wealth index (HW), an index of household dispersion (HD), index of chronicity of diarrhea (CDI) and protozoan infection (CPI), and dietary pattern scores were generated and incorporated into spatial cluster analysis and multiple regression models of anthropometric and infection outcomes. Influence diagrams created during small group workshops identified participant perceptions of health. Results: Households with higher HWI had a latrine, aqueduct access, cell phone, and/or stove and HD ranged from 5–113 households/km2. High prevalence clusters of hookworm and Trichuris (but not Ascaris) occurred in regions with lowest HWI and HD. Ascaris and hookworm reinfection was driven by individual susceptibility traits (stunting) whereas Trichuris reinfection was dependent on household (maternal education) and regional (high prevalence cluster) characteristics. The high frequency of diarrhea was consistent with community perceptions of health priorities and of the link between diarrhea and poor water quality. Both E. coli counts and CDI

i were higher in households without aqueduct access. Sixty percent of children were stunted and 22% were underweight. The detrimental effect of CPI on HAZ was driven by household density but moderated by household wealth. Diet also influenced growth. The basic diet (rice, beans, coffee, sugar) was supplemented with fruits and vegetables (FV region) or market snacks (CT region). A “Market” diet pattern was concentrated in the CT region close to the road. Meat was beneficial for linear growth but only in the FV region whereas carbohydrates (sweets in FV region and root vegetables in CT region) were detrimental, even after controlling for infection and socio-economic status. With regard to weight gain, fish and eggs were beneficial in the FV region whereas, in the CT region, milk products were beneficial, but chips and sweets were detrimental. Implications: This transdisciplinary research highlighted key public health messages necessary to improve growth and reduce infection in this vulnerable population.

ii Abrégé

Contexte: Le retard de croissance, correspondant à une valeur centrée réduite (z-score) de la taille par rapport à l'âge (ZTA) inférieur à deux écarts type, découle d'une diète pauvre et d'infections fréquentes qui sont influencées par des facteurs sociaux et biophysiques. Seulement peu d'études ont utilisé une approche multidisciplinaire santé-écologie pour analyser la relation entre ces facteurs et la fréquence des arrêt de croissance. Objectif: Examiner les relations entre le retard de croissance chez les enfants en âge préscolaire et les infections gastro-intestinales avec le contexte social, biophysique et spatial dans des conditions de pauvreté extrêmes chez les peoples Ngabe de l'ouest du Panamá, où prennent place des programmes d'aide alimentaire (AA) et de transfert monétaire (TM).Méthodes: Dans le carde d'une étude de 16 mois, nous avons suivi 356 enfants d'âge préscolaire impliqués dans deux cycles de réinfections à la suite d'un traitement à l'albendazole. Les données recueillies inclues des échantillons répétés de matière fécale, questionnaires sur les comportements sociaux des ménages, journaux alimentaires et mesures anthropométriques, échantillons d'eau, données de système de localisation GPS et ateliers participatifs. Des indices de richesse des ménages (IRM) basé sur le patrimoine, de dispersion des ménages (IDM), de chronicité des la diarrhée (ICD) et de l'infection aux protozoaires (CIP), et une note sur les habitude alimentaires ont été calculés. Ces indices ont été inclus dans une analyse spatiale de regroupement et des analyses de régression linéaire multiple pour prédire les mesures anthropométriques et le résultat des infections. Les diagrammes d'influence créés durant les ateliers ont permis d'évaluer les perceptions des participants par rapport à la santé. RésultatsLes ménages avec les IRM élevés avaient une toilette, l'accès à l'aqueduc, un téléphone cellulaire et/ou un poêle. Leur IDM variait de 5-113 ménages/km2. Des regroupements de prévalence d'ankylostome et de Trichuris (mais pas d'Ascaris) étaient présents dans les régions avec les IRM et les IDM les plus faibles. Les réinfections par Ascaris et par

iii des ankylostomes étaient influencées par des traits de sensibilité personnelle (retard de croissance): la réinfection par Trichuris dépendait des caractéristiques du ménage (éducation maternelle) et de la région (regroupement de prévalence élevée). La fréquence élevé de diarrhée correspondait avec la perception de la communauté sur les priorité en matière de santé et avec la mauvaise qualité des eaux. Les dénombrements d'E. coli et le ICD étaient plus élevés dans les ménages qui n'avaient pas accès à l'aqueduc. Soixante-six pour cent des enfants étaient atteints d'arrêt de croissance et 22% souffraient d'insuffisance pondérale. L'effet nuisible de la CIP sur la ZTA était principalement influencé par la densité de ménage mais aussi par le patrimoine du ménage. Le régime alimentaire avait aussi un effet sur la croissance. Le régime de base (riz, haricots, café, sucre) était complété avec des fruits et des légumes (région AA) ou des friandises du marché (région TM). Un régime « du marché » était concentré dans la région TM proche de la route. La présence de viande avait un effet bénéfique sur la croissance linéaire seulement dans la région AA alors que les glucides (sucreries dans la région AA et légumes racines dans le région TM) avait un effet néfaste, même en tenant compte des infections et du statut socio-économique. Le poisson et les œufs avaient un effet bénéfique sur le gain de masse corporelle dans la région AA. Dans la région TM, les produits laitiers avait un effet bénéfique sur la prise de masse alors que les croustilles et les sucreries avaient un effet néfaste. Implications: Cette étude multidisciplinaire souligne des messages important en santé publique pour améliorer la croissance et diminuer les infections dans cette population vulnérable.

iv Acknowledgements

First and foremost I would like to thank the communities that opened their doors to me during the 2 years in Panamá. Their generous participation in the research project and their willingness to share their daily activities with me made the experience valuable in ways beyond the research alone. I would especially like to thank the men and women who acted as my translators, guides and educators. Their commitment to the project as well as the stories and observations they shared on those long hikes opened my eyes to new dimensions of the comarca. I am grateful to the nutritionists who stepped outside their comfort zones to experience a new part of Panamá and to the lab technicians who worked well into the night on the big collection days. In Panamá City the informative support and guidance of Emerita Pons, Flavia Fontes and Odalis Sinisterra made the collaboration between MINSA and McGill University not only possible but so productive. I will forever be grateful to Vicky Valdés, whose mentorship during my stay in Panamá not only provided invaluable logistical advice but also sound cultural counsel.

In Montreal, I have been lucky to have the opportunity to work with my supervisors, Dr. Marilyn Scott and Dr. Kristine Koski who have challenged me to develop as a researcher and given me the freedom to explore such a diverse project. I respect their unflagging energy and enthusiasm for research as well as their willingness to delve into such a transdisciplinary project. I have enjoyed sharing these last years with my lab mates, past and present (Felipe, Doris, Rachel, Marie-Pierre, Lisa, Javier, Candice and Maurice), who have kept all aspects of this project fun. Shannon, Christina, Shirley and all the members of the Institute of Parasitology and School of Environment thank you for the smiles, hallway conversations and helping hands.

I gratefully acknowledge the National Sciences and Engineering Research Council (NSERC), the International Development Research Centre (IDRC) and the

v Secretaria Nacional de Ciencia, Tecnología y Innovación (SENACYT) for the financial support that made this research possible. I am also appreciative of the financial assistance provided by the Lynden Laird Lyster Memorial Fellowship from the Institute of Parasitology and the travel grants provided by the Centre for Host-Parasite Interactions (CHPI).

Countless discussions around dinner tables in Panamá and Montreal have been indispensable to the navigation of scientific, cultural and educational themes over these past 6 years. Thank you to Verena, Anne-Marie and Daniel. I wish to thank my family for their unconditional love and support during yet another one of my adventures. I enjoyed sharing Panamá with you and valued your listening ears throughout my PhD work. Finally, to Rob who’s infinite curiosity has kept me inspired through these final years. Thank you for enduring the difficult times with me, keeping me grounded and reminding me to keep playing in life.

vi Contribution of Authors

This thesis was written by Carli Halpenny. The thesis consists of four manuscripts, all of which were co-authored with my supervisors, Dr. Marilyn E. Scott and Dr. Kristine G. Koski as the research was developed in collaboration with them. In addition, Victoria E. Valdés co-authored the first (Chapter 3) and second (Chapter 4) manuscripts as she contributed to the development of the research during my time in Panamá and was the PI on funding received through SENACYT. The second manuscript (Chapter 4) was also co-authored with Claire Paller, an undergraduate research assistant who assisted in developing the spatial analysis of the STH data. The first manuscript is published in the American Journal of Tropical Medicine and Hygiene, the second manuscript has been submitted for publication to PLoS Neglected Tropical Diseases and the third and fourth manuscripts will be submitted in the near future.

I was responsible for all the field work, statistical analysis of the data and data interpretation. In the field, I identified and trained local field workers, MINSA nutritionists and MINSA lab personnel, the latter received training in two new methods for fecal analysis of STH infections (FLOTAC and Kato Katz). Local field workers collected fecal and water samples in addition to administering questionnaires and translating for MINSA nutritionists. MINSA lab personnel conducted fecal analyses and MINSA nutritionists conducted 24 hour recalls, demographic questionnaires and took anthropometry measures. I accompanied field staff on a rotating basis to verify data collection methods and to meet the participants. I was responsible for coordinating the logistics of data collection and analysis as well as reviewing questionnaires and lab work for data quality. I conducted all water sample analysis and collected GPS data. I was present in all participatory workshops and trained local women to translate and assist in the delivery of these workshops. During field work, guidance was provided by Victoria E. Valdés for the nutritional components. Nilofar Hariri, Dianna Mohid

vii and Sara Wing assisted in developing a novel nutrient database that was used to analyse data for Chapter 5. Financial resources for the field work were obtained through two research grants that I wrote under the guidance Dr. M. E. Scott, Dr. K. G. Koski and collaborators in Panamá. Dr. M.E. Scott and Dr. K.G. Koski also provided financial resources during data analysis, they were involved in the design of experiments, critiqued data presentation and analyses, and pre-edited all written manuscripts and the thesis.

viii Statement of Originality

This doctoral research study has provided a comprehensive transdisciplinary analysis of the inter-relationship between infection and malnutrition. Previous longitudinal studies have incorporated dietary, anthropometric and intestinal infection measures, and review articles have described the negative cycle of infection and malnutrition by integrating observations from a variety of studies, but to my knowledge, this is the first study on intestinal infections and stunting that has integrated such a broad range of measures from parasitology, nutrition and geography and that has incorporated a mixed methods Ecohealth approach involving both qualitative, quantitative approaches.

Several methodological advances and new applications were used that led to novel findings. 1. I developed a new tool to characterize chronicity of diarrhea (Chronicity of Diarrhea Index, CDI) and of protozoan infections (Chronicity of Protozoa, CPI) based on frequency of the condition based on 4-7 fecal samples collected over a 16 mo period. The development of these chronicity indices allowed us to show the importance of recurring protozoan infection on child stunting. 2. Household density estimates were used to explain chronicity of protozoan infection. Despite better access to latrines and aqueducts, densely populated regions had more chronic protozoan infection than dispersed regions. Interestingly, gastrointestinal nematode transmission was higher in more dispersed households. 3. Principle components derived household asset-based wealth indices (HWI) were used to differentiate among households all experiencing extreme poverty. The discriminatory power of this index was strong

ix enough to describe patterns in child height-for-age (HAZ), soil- transmitted helminth (STH) transmission and CPI. 4. This is the first time that Dietary Pattern Analysis (DPA) has been used to describe and differentiate diets of preschool children in a developing country. We identified two distinct food consumption patterns that we labelled as a “Beans and Rice” and a “Market” pattern. The latter was particularly prevalent near the access road in the Cash Transfer region.

Additional original findings are listed below.

1. Intriguing species-specific differences in STH reinfection dynamics emerged that revealed stronger similarities of Ascaris with hookworm rather than Trichuris. Ascaris and hookworm reinfection was driven primarily by stunting, an individual level susceptibility trait whereas Trichuris reinfection rates were determined by household (maternal education) and regional (spatial clusters) characteristics. 2. This research considered both gastrointestinal parasite infection and food consumption in comparisons of preschool child anthropometry between two types of conditional transfer programs. After controlling for parasite infection as well as regional socio-economic traits, we demonstrated that food-based predictors of linear growth and weight gain differed between the program regions. Meat consumption improved and sweets decreased linear growth in the FV region whereas in the CT region root vegetables were detrimental to growth but bread and pasta were beneficial. Fish and eggs improved weight gain in the FV region but milk products were important in the CT region where sweets and chips decreased weight gain. 3. Through participatory methodologies, we found that communities had a good sense of priority health concerns. Communities understood the causal pathways associated with the etiology of diarrhea but they had x misconceptions about the potability of water from aqueducts. They assumed that water from aqueducts did not need to be treated at the household level. Although E. coli counts were lower in the aqueduct water, they were well above recommended levels for drinking water. Discussions also revealed pivotal aspects requiring attention including the need to address taste of treated water.

Finally, through this research, novel approaches to collaboration between the Ministry of Health and local indigenous peoples developed with implications for future collaborations. This was the first time participatory research was conducted in this region of Panamá. Furthermore, the involvement of national level institutions (Panamanian Ministry of Health (MINSA)), regional level health authorities (MINSA comarcal), local leaders and local participants was novel for health research in Panamá. Certain novel elements of the study were particularly of note: 1) the inclusion of local health perspectives to inform the design of a health investigation and furthermore the use of quantitative data to verify these findings; 2) the involvement of local translators with MINSA nutritionists so that interviews could be conducted in Ngäbere (the local language); 3) training of local leaders to conduct questionnaires and collect samples; 4) small group workshops that used participatory methods and were conducted in Ngäbere. As a consequence, considerable local capacity building and empowerment occurred. Furthermore, the respect of MINSA for local capabilities improved markedly and is changing the nature of collaborations.

xi Table of Contents

Abstract ...... i

Abrégé ...... iii

Acknowledgements ...... v

Contribution of Authors ...... vii

Statement of Originality ...... ix

Table of Contents ...... xii

List of Tables...... xviii

List of Figures ...... xxi

List of Abbreviations ...... xxiii

CHAPTER 1

1.1 Introduction ...... 1

1.2 Rationale and Research Objectives ...... 3

1.3 References ...... 5

CHAPTER 2

Literature Review - Foundations of child health in a rural impoverished context . 8

2.1 Chronic growth failure: Nutrition and infection ...... 8

2.1.1 Stunting as a composite measure for child health in a region...... 8

2.1.2 Implications for child cognitive and motor development ...... 11

2.1.3 Long term consequences of stunting ...... 13

2.2 Gastrointestinal infections: Biology and risk factors for transmission ...... 14

2.3 Rural poverty landscape ...... 17

2.3.1 Rural development and the changing shape of the rural landscape ... 17

2.3.2 Rural development and health ...... 19

xii 2.4 Towards an integrated view of health ...... 22

2.5 Tools for deciphering the complexity of health ...... 25

2.5.1 Mixed methods ...... 25

2.5.2 Descriptive spatial mapping ...... 29

2.5.3 Variable reduction methods ...... 30

2.6 Panamá: Comarca Ngäbe Buglé ...... 37

2.6.1 Socio-political background ...... 37

2.6.2 Indigenous peoples of Panamá – Focus on the Ngäbe Buglé ...... 38

2.7 References ...... 42

CHAPTER 3

Household density and asset-based indices predict child health in impoverished indigenous villages in rural Panamá ...... 60

3.3.1 Study population ...... 64

3.3.2 Ethical considerations ...... 65

3.3.3 Study Procedure ...... 65

3.3.4 Spatial survey ...... 66

3.3.5 Household wealth characterization...... 66

3.3.6 Anthropometry ...... 67

3.3.7 Fecal samples ...... 67

3.3.8 Water quality and behavioural risk factors ...... 68

3.3.9 Data analysis ...... 69

3.4 Results ...... 70

3.4.1 Household and child variables: overview and comparisons between political regions...... 70

xiii 3.4.2 Spatial comparison of household characteristics...... 72

3.4.3 Child health comparisons ...... 72

3.5 Discussion ...... 74

3.6 Acknowledgements ...... 83

3.7 References ...... 83

CHAPTER 4

Regional, household and individual factors that influence soil transmitted helminth reinfection dynamics in preschool children from rural indigenous Panamá ...... 108

4.1 Abstract ...... 109

4.2 Introduction ...... 110

4.3 Materials and Methods ...... 112

4.3.1 Ethics Statement ...... 112

4.3.2 Study area...... 112

4.3.3 Study Design and Protocol ...... 113

4.3.4 Fecal samples ...... 115

4.3.5 Characterization of infection risk factors ...... 115

4.3.6 Statistical analysis ...... 117

4.4 Results ...... 118

4.4.1 Methodological comparison ...... 118

4.4.2 Household and demographic characteristics of participants ...... 119

4.4.3 Reinfection dynamics...... 119

4.4.4 Factors associated with spatial clusters of infection and with individual reinfection intensity...... 120

4.5 Discussion ...... 122

xiv 4.6 Acknowledgements ...... 129

4.7 References ...... 131

Connecting Statement II ...... 146

CHAPTER 5

Comparison of preschool child diet and anthropometry between a cash transfer and food voucher conditional transfer programs in rural Panamá ...... 147

5.1 Abstract ...... 148

5.2 Introduction ...... 150

5.3 Materials and Methods ...... 151

5.3.1 Study area and population...... 151

5.3.2 Ethical considerations ...... 153

5.3.3 Study participants ...... 153

5.3.4 Study design and protocol ...... 154

5.3.5 Household and child characterization...... 154

5.3.6 Anthropometry ...... 154

5.3.7 Diet analysis ...... 155

5.3.8 Spatial analysis ...... 157

5.3.9 Statistical analysis ...... 158

5.4 Results ...... 159

5.4.1 Study population description ...... 159

5.4.2 Diet description ...... 160

5.4.3 Dietary pattern description ...... 161

5.4.4 Predictors of preschool linear growth...... 161

5.4.5 Predictors of preschool child weight gain ...... 162

xv 5.5 Discussion ...... 163

5.6 Acknowledgements ...... 171

5.7 References ...... 172

Connecting Statement III ...... 199

CHAPTER 6

A mixed methods approach reveals understanding of the interrelationships between diarrhea and water quality among the Panamanian Ngäbe Buglé ..... 200

6.1 Abstract ...... 201

6.2 Introduction ...... 203

6.3 Materials and Methods ...... 204

6.3.1 Study area and population ...... 204

6.3.2 Study design ...... 205

6.3.3 Phase I Workshops ...... 206

6.3.4 Phase II Surveys ...... 207

6.3.5 Phase III Workshops ...... 208

6.3.6 Ethical considerations ...... 209

6.3.7 Statistical analysis ...... 210

6.4 Results ...... 210

6.4.1 Phase I workshop – Community health perceptions ...... 210

6.4.2 Phase II survey data ...... 212

6.4.3 Phase III workshop – Barriers to water treatment ...... 214

6.5 Discussion ...... 215

6.6 Acknowledgements ...... 223

6.7 References ...... 224

xvi CHAPTER 7

General Discussion ...... 235

7.1 Major Findings ...... 235

7.2 Methodological Advances ...... 237

7.3 Reflections on the Methodological Approach ...... 239

7.3.1 Transdisciplinarity ...... 239

7.3.2 Participation ...... 240

7.4 Limitations ...... 242

7.5 Public Health Implications ...... 245

7.6 References ...... 247

Appendix ...... 251

xvii List of Tables

Chapter 3

Table 1. Timeline of sample collection during 10 household visits from June 2008 – October 2009...... 89

Table 2. Household variables included in the Household Wealth Index (HWI), their respective scoring factors and the percent of households in each wealth category owning items...... 91

Table 3. Summary of household and child characteristics and health outcomes according to political region and household density category...... 92

Table 4. Predictors of height-for-age Z-score in Panamanian preschool children from a stepwise multiple regression model...... 95

Table 5. Predictors of chronicity of protozoan infection in Panamanian preschool children based on a Zero-inflated Poisson regression model...... 97

Table 6. Predictors of chronicity of diarrhetic stool in Panamanian preschool children from a Stepwise Poisson regression model...... 99

Chapter 4

Table 1. Sample sizes for two STH reinfection cycles among Panamanian preschool children...... 138

Table 2. Comparison of characteristics between households within and outside high prevalence clusters...... 139 Table 3. Final multiple logistic regression models predicting household presence within Cycle 2 high prevalence clusters...... 140

xviii Table 4. Comparison of regional, household and individual factors influencing STH reinfection of Panamanian preschool children...... 141

Table 5. Negative binomial regression models of Ascaris and hookworm reinfection intensity in Panamanian preschool children ...... 142

Chapter 5

Table 1. Food groups consumed by at least 10% of the population (A) and 3-10% of the population (B). Groups were generated from foods recorded in 24 hr recalls for Dietary pattern analysis using Principal Components Analysis (PCA)...... 179

Table 2. Factor weights of the 2 Principle Component Analysis derived diet patterns detected among Panamanian preschool children...... 183

Table 3. Summary statistics of household and child variables comparing Food Voucher (FV) and Cash Transfer (CT) regions...... 184

Table 4. Mean amount (g/d) of the food groups consumed by Panamanian preschool children in the Food Voucher (FV) and Cash Transfer (CT) regions during 2008 and 2009 for food groups consumed by more than 3% of the population and that differed in mean consumption between programs and/or years...... 186

Table 5. Growing seasons (primary – light gray, secondary – medium gray) and harvest periods (dark gray) of primary food crops and wild foods in Besiko, comarca Ngäbe Buglé (X – burn period, diagonal slash – overlap of growing and harvest periods)...... 188

Table 6. Percentage of households who bought each food group for those households in which the food was consumed...... 189

xix Table 7. Hierarchical multiple linear regression models of HAZ scores (A) and change in HAZ scores (B) of Panamanian preschool children by Conditional Transfer region...... 190

Table 8. Hierarchical multiple linear regression models of WAZ scores (A) and change in WAZ scores (B) of Panamanian preschool children by Conditional Transfer region...... 193

Chapter 6

Table 1. Summary of the top health priority identified by each of 68 small groups in Soloy and Chorcha through the cue card exercise...... 228

Table 3. Summary of water treatment knowledge and practices and link between water treatment and diarrhea symptoms based on questionnaires administered to primary care giver of preschool children...... 230

xx List of Figures

Chapter 1

Figure 1. Conceptual framework of the biophysical and socio-economic factors influential in preschool child health in the comarca Ngäbe-Buglé, Panamá...... 7

Chapter 3

Figure 1. Household density clusters in study area located in western Panamá ...... 100

Figure 2. Prevalence of symptomatic diarrhea (a) and protozoan infection (b) over the duration of the study...... 102 Figure 3. Frequency distribution of density of participating households...... 103

Figure 4. Frequency distribution of Household Wealth Index (HWI) by household density group...... 104

Figure 5. Mean child health outcomes by household density group and wealth percentile...... 105

Chapter 4

Figure 1. Prevalence and intensity for Ascaris, hookworm and Trichuris in Panamanian preschool children...... 144

Figure 2. Spatial clusters of households with high prevalence of hookworm and Trichuris infection ...... 145

xxi Chapter 5

Figure 1. Scree plot of eigenvalues after Principle Component Analysis (PCA) of dietary intake data...... 196

Figure 2. Percentage of Panamanian preschool children in the Food Voucher (A) and Cash Transfer (B) programs that consumed specific foods or food groups in 2008 and 2009 consumed by at least 10% of the population in 2008 or 2009...197

Chapter 6

Figure 1. Examples of cause and prevention influence diagrams for diarrhea from community workshops...... 231

Figure 2. Longitudinal prevalence with 95% CI of diarrhetic samples collected from preschool children in the CT and FV region...... 232

Figure 3. Average E.coli concentration in water samples collected from households who used aqueduct or ground water sources in the CT and FV regions...... 233

Figure 4. Examples of felt board “stories” created by participants during Phase III workshops...... 234

xxii List of Abbreviations

ABZ Albendazole

AIC Akaike’s Information Criteria

ANOVA Analysis of Variance

CDI Chronic Diarrhea Index cfu colony forming units

CI Confidence Interval

CPI Chronic Protozoan Index

CT Cash Transfer

CTP Conditional Transfer Program

DALY Disability Adjusted Life Years

DDS Diet Diversity Score

DHS Demographic Health Survey

DPA Dietary Pattern Analysis epg Eggs per gram

ERR Egg Reduction Rate

FA Factor Analysis

FV Food Voucher

FVS Food Variety Score

xxiii g grams

GDP Gross Domestic Product

GI Gastrointestinal

GPS Geographic Positioning System

HAZ Height-for-age Z score

HWI Household Wealth Index

INCAP Instituto de Nutrición de Centro America y Panama

IUGR Intrauterine Growth Restriction kcal Kilocalories km Kilometers

LSMS Living Standards in Measurement Survey

MEDUCA Ministerio de Educación

MIDA Ministerio de Desarrollo Agropecuario

MIDES Ministerio de Desarrollo Social

MINSA Ministerio de Salud mL Milliliters mo Month

NA Not Applicable

NE Not entered

PCA Principle Components Analysis

xxiv SEM Standard Error of the Mean

SES Socio-economic Status

STH Soil Transmitted Helminth

USDA United States Department of Agriculture

UTM Universal Transverse Mercator

WAZ Weight-for-age Z score

WHO World Health Organization

WHZ Weight-for-height Z score

xxv CHAPTER 1

1.1 Introduction Health inequalities are considered to be the “leading health problem” in Latin America [1]. In Panamá specifically, the socio-economic, cultural and geographic barriers to health among Panama’s Indigenous Peoples have resulted in an extremely poor health profile characterized by a self-reinforcing negative cycle of poverty, chronic malnutrition, and multiple gastro-intestinal (GI) parasitic infections. Ninety percent of rural indigenous populations in Panamá live in extreme poverty (less than $1.75/day) [2] and in the comarca Ngäbe-Buglé of western Panamá, the average household income is approximately $300/year, primarily from agricultural labour. Recent surveys [3,4] of the Ngäbe-Buglé have documented low diet diversity, inadequate intakes of energy, fat and several vitamins and minerals (as defined by the Instituto de Nutrición de Centroamérica y Panamá) [3,4], as well as very high prevalence of chronic malnutrition (50% [4] & 61% [3] stunted) and GI nematodes (over 80%) in pre-school children [3].

One of the often over-looked consequences of chronic malnutrition is impaired immunity that increases susceptibility to a wide range of infections including GI nematodes. These parasites in turn cause malnutrition by impairing absorption, thus limiting growth and physical, social, mental and cognitive development [5-7]. In addition, in chronically infected populations, the parasite specific immune response may be modulated to balance damage caused by immunopathology [8].

It is well recognized that the malnutrition – infection interaction does not occur in isolation but additional biophysical as well as socio-economic variables contribute to these afflictions [5,9]. Ehrnberg and Ault [10] suggest that biological factors such as nutritional status often synergize in a negative way with extrinsic determinants of health such as poverty, and environmental factors. For example, poor access to health services exacerbates the increased

1 susceptibility to infection that results from dietary deficiencies [6,11], which can have long lasting effects on development and productivity [5,12]. Current efforts to control GI parasite infection focus primarily on community wide anthelmintic delivery focusing primarily on school age children [13]; however, the integration of parasite control programs within social development programs may further reduce rates of re-infection.

In Panamá, the Red de Oportunidades aims to address the high levels of poverty and malnutrition in the Indigenous and rural populations by assigning individual corregimientos to either a conditional food voucher program a conditional cash transfer program. Both conditional transfer programs are a coordinated effort between the Ministerios de Salud (MINSA), de Educación (MEDUCA), and Desarollo Agropecuario (MIDA) which provide additional income in exchange for participation in health and development programs. Women from families living in extreme poverty and/or in poor conditions can receive $50 US/month in food vouchers or cash if their children stay in school and are current on their vaccinations, if women receive regular sexual and reproductive health checks, if parents have attended meetings of Padres de Familia en la Escuela (MIDES) and if one adult member of the household has participated in organized agricultural production training workshops. Mandatory school attendance reduces the number of children migrating to work in coffee plantations during the school year, and provides the additional immediate benefit of a school lunch program and school-administered deworming and nutrient supplements. While the requirements of both transfer programs are similar, vouchers can only be redeemed by the designated woman in each household for specified food items at certain local stores, whereas use of cash transfers is unrestricted.

It is recognized that coordinated, multi-sectoral programs have an increased potential to improve health in neglected populations of Latin America [10] due to the sustained, synergistic effect of investment in education, health 2 and nutrition on human capital and community economy [14]. Indeed, evidence based on cross-sectional studies of conditional cash transfer programs in other countries reveals improvement in growth, nutritional status, school attendance and health system usage [15-17]; however, longitudinal studies of impact on a broad range of indicators has not been undertaken, and no studies have compared the two types of conditional programs. In addition, despite the inclusion of anthelmintic delivery, the impact of conditional transfer programs on GI parasite infection has not been characterized. Finally, distinguishing the synergies between program components is an important fundamental question in understanding the increased benefit of conditional transfer programs [18].

In order to capture these important interactions between program components as well as the immediate and sustained impact of the Panamanian programs, a comprehensive evaluation is necessary and must consider social, economic and biophysical components across multiple temporal scales. This is possible using an ecosystem approach to health, which combines ecological theory, in particular systems thinking, with participatory research [19,20] to create systems descriptions of the poverty – malnutrition – infection cycle in the Ngäbe Buglé comarca. This will allow the identification of contextual factors that balance or reinforce this cycle and are important targets for intervention programs [19].

1.2 Rationale and Research Objectives The comarca Ngäbe-Buglé provided an opportunity to examine the GI parasite infection – nutrition interrelationship within the context of a multi- dimensional poverty reduction program. In particular, there was an opportunity to examine the importance of biophysical and socio-economic factors to this relationship. Within this conceptual framework (Figure 1), the following objectives were investigated:

3 - Determine whether GI parasite infection increases preschool child stunting and whether stunting in turn increases susceptibility to GI parasite reinfection. - Identify the individual, household and regional level factors that influence preschool child stunting and infection status. - Examine the interactions between biophysical, social and spatial factors that influence preschool child nutrition and GI parasite reinfection.

Chapter 2 consists of a review of the relevant literature from the fields of parasitology, nutrition and geography to set the context for the thesis. In Chapter 3, I investigated the relationship between preschool child stunting and chronic diarrhea and protozoan infection. Furthermore, I examined how this relationship was influenced by components of the social and biophysical environment, specifically sanitation and hygiene behaviours, aqueduct access, household poverty and household density. This manuscript has been published in The American Journal of Tropical Medicine and Hygiene. In Chapter 4, I investigated reinfection dynamics of three Soil Transmitted Helminth (STH) infections and how regional, household and individual level traits influenced susceptibility and exposure to reinfection. This manuscript has been submitted to PLoS Neglected Tropical Diseases. In Chapter 5, I examined the influence of food consumption on child linear growth and weight gain in addition to comparing food consumption and anthropometry between 2 types of conditional transfer programs. Chapter 6 examines the concordance between local perceptions of health and quantitative study results in relation to diarrhea and water quality. These final two manuscripts (Chapters 5 and 6) are in preparation for submission. In Chapter 7, I expand upon overarching themes that have emerged from this research in a General Discussion.

4 1.3 References

1. PAHO (1998) Leading Pan-American Health. Washington: Pan American Health Organization. 2. Ministerio de Economía y Finanzas (2010) La Distribución del Ingreso en Los Hogares de Panamá:Encuesta de Niveles de Vida 2008. Panamá: Ministerio de Economía y Finanzas. 34 p. 3. Payne L, Koski KG, Ortega-Barria E, Scott ME (2007) Benefit of vitamin A supplementation on Ascaris re-infection is less evident in stunted children. J Nutr 137: 1455-1459. 4. Ortega A, Fontes F, Sinisterra O, Valdés V (2006) Evaluación nutricional en pre- escolares y escolares de los distritos de Mironó, comarca Ngobe-Buglé y Santa Fe, Veraguas. Panama City, Panama: Secretaría del Plan Alimentario Nacional (SENAPAN). 46 p. 5. Stephenson LS, Latham MC, Ottesen EA (2000) Malnutrition and parasitic helminth infections. Parasitology 121: S23-S38. 6. Koski KG, Scott ME (2001) Gastrointestinal nematodes, nutrition and immunity: Breaking the negative spiral. Annu Rev Nutr 21: 297-321. 7. Dewey KG, Begum K (2011) Long-term consequences of stunting in early life. Matern Child Nutr 7: 5-18. 8. Turner JD, Faulkner H, Kamgno J, Kennedy MW, Behnke J, et al. (2005) Allergen-specific IgE and IgG4 are markers of resistance and susceptibility in a human intestinal nematode infection. Microbes and Infection 7: 990- 996. 9. Hughes RG, Sharp DS, Hughes MC, Akau'ola S, Heinsbroek P, et al. (2004) Environmental influences on helminthiasis and nutritional status among Pacific schoolchildren. International Journal of Environmental Health Research 14: 163-177.

5 10. Ehrenberg JP, Ault SK (2005) Neglected diseases of neglected populations: thinking to reshape the determinants of health in Latin America and the Caribbean. BMC public health [electronic resource] 5: 119. 11. Scrimshaw NS, SanGiovanni JP (1997) Synergism of nutrition, infection, and immunity: An overview. Am J Clin Nutr 66. 12. Crompton DWT, Nesheim MC (2002) Nutritional impact of intestinal helminthiasis during the human life cycle. Annu Rev Nutr 22: 35-59. 13. WHO (2001) Report of the 54th World Health Assembly Control of Schistosomiasis and Soil-transmitted Helminth Infections. WHO. Resolutions No. 54.19 Resolutions No. 54.19. 14. Villatoro P (2005) Conditional cash transfer programmes: experiences from Latin America. CEPAL Review 86: 83-96. 15. Skoufias E (2005) PROGRESA and its impacts on the welfare of rural households in Mexico. Research Report of the International Food Policy Research Institute: 1-98. 16. Hoddinott J, Skoufias E (2004) The impact of PROGRESA on food consumption. Econ Dev Cult Change 53: 37-61. 17. Behrman JR, Skoufias E (2006) Mitigating myths about policy effectiveness: Evaluation of Mexico's antipoverty and human resource investment program. Ann Am Acad Pol Soc Sci 606: 244-275. 18. Rawlings L, Rubio G (2005) Evaluating the Impact of Conditional Cash Transfer Programs. The World Bank Research Observer 20. 19. Waltner-Toews D (2004) Ecosystem sustainability and health: A practical approach. Cambridge: Cambridge University Press. 138 p. 20. Lebel J (2003) Health: An ecosystem approach. http://www.idrc.ca/openebooks/012-8/.

Figure 1. Conceptual framework of the biophysical and socio-economic factors influential in preschool child health in the comarca Ngäbe-Buglé, Panamá.

7 CHAPTER 2

Literature Review - Foundations of child health in a rural impoverished context

2.1 Chronic growth failure: Nutrition and infection

2.1.1 Stunting as a composite measure for child health in a region.

In developing countries, 32% of children under 5 years old are unable to reach their growth potential [1]. Africa has the highest prevalence of stunting (40%) however the large population in Asia results in the greatest number of children who are small for their age (112 million) [1]. This chronic restriction to a child’s potential growth is the cumulative effect of poor nutrition and frequent infection, making stunting an indicator of the long term impact of a child’s environment on their health [2,3].

Early investigations into the nutritional basis of stunting focused primarily on protein and energy malnutrition through supplementation studies in the 1970’s and 1980’s which found differing results on height/length gain (reviewed in Allen[4]). There was stronger evidence for the benefits of protein than energy supplementation on length gain however growth faltering was still detected in individuals with protein and energy sufficient diets [4]. This led to the more recent focus on the contribution of micronutrient deficiencies in the dietary mechanisms of stunting. Meta analyzes of combination or nutrient specific supplementation found that multiple micronutrient supplementation but not

8 vitamin A or iron supplementation significantly improved child linear growth [5,6]. Studies of zinc supplementation have had differing results. In populations at risk of zinc deficiency, supplementation greatly improved child height gain [7,8] however no effect was detected in populations with less severe deficiencies [6]. In addition to specific macro and micro nutrient deficiencies, greater dietary quality as indicated by a more diverse diet [9,10] and more animal source foods [11], also improves child growth, likely due to the increased nutrient adequacy of higher quality diets [12].

Child stunting in developing countries cannot be explained through dietary mechanisms alone as infection also plays a crucial role in child growth [13,14]. The anorexia, nutrient malabsorption and diarrhea associated with gastrointestinal infections have direct consequences on child growth by limiting the nutrients consumed and absorbed [15,16]. Furthermore, the immune response to the infection may also inhibit growth due to the effect of chronic cytokine exposure on bone deposition [17]. Cross sectional field studies have noted an association between gastrointestinal (GI) infection and stunting for soil transmitted helminth (STH) infections [18-20] as well as Giardia [21,22]. Case- control longitudinal studies that examined a causal link between infection and child growth have had differing results. Treatment for STH and Giardia infections were followed by improved height gain in some studies [23] while others only detected weight gain [24,25].

As stunting is considered an indicator of the cumulative nutrition and infection status of a child it follows that the duration and frequency of infection also play an important role in determining the severity of growth failure. Indeed, a recent pooled analysis found that the odds of stunting at 2 yrs of age increased with each prior diarrheal episode as well as the cumulative duration of diarrheal symptoms [26]. Even subclinical infections caused by the chronic ingestion of pathogenic microorganisms can cause structural changes to the intestine which

9 leads to growth faltering [27,28], a phenomenon known as environmental enteropathy [29].

In addition to the above studies, large scale national studies and community level investigations that created composite models of child stunting further support the relationship between stunting, diet and infection. Furthermore, these studies have highlighted the importance of community and household factors for child nutritional status. Large scale studies that analyzed Demographic Health Survey (DHS) and national census data have shown the importance of poverty and health care access to child chronic malnutrition. In Africa and parts of the Andes, community level indicators of poverty such as illiteracy rates [30] and economic inequality [31] were correlated with stunting in multivariate analysis. Furthermore, regional geographic traits such as a dry climate [32] and being in the highlands as well as indigenous ethnicity [31] were related to higher levels of stunting. At the household level, poor household wealth [20,30,33], low maternal education [30,32,33] and poor health seeking behaviours [31,32] or lack of access to health care [33] also contributed to child stunting. Community level studies also documented the negative influence of poverty on stunting [34,35]. Furthermore, they highlighted the importance of nutrient density in weaning foods [36], diet quality [37], protein consumption [20] and diet diversity [38], however the importance of dietary diversity was limited to urban areas [38]. Multiple studies demonstrated the negative effect of infection on stunting, either polyparasitism [20], protozoan infection [39,40] or STH infection [41,42]. A study in Mexico demonstrated the importance of the family environment, especially type or stability of profession for child stunting and furthermore showed that the risk factors for stunting differed between rural and urban areas [43]. Thus, at the national, regional and household level, poverty and poverty related traits such as employment, education attainment and health care access are particularly influential for child stunting.

10 Thus, child stunting is of greatest concern in children frequently exposed to infection who consume a diet of insufficient nutritional quality. This relationship has been called a negative spiral for the way in which infection leads to a decline in nutritional status which further increases susceptibility to infection [15]. However, this synergistic interaction can also be used to improve child growth when nutritional interventions reduce the adverse effects of infection on growth [14].

2.1.2 Implications for child cognitive and motor development.

As growth faltering reflects the cumulative effect of nutritional deprivation and frequent infection, it is associated with poor child development [44,45]. Child motor activity is considered a developmental milestone and plays an important role in cognitive development by enabling children to explore their environment. Studies of child gross motor development in Nepal [46], Guatemala [47] and Pakistan [48] have demonstrated that children who are shorter for their age [46] or who have slower linear growth in the first year of life [47,48] began walking at a later age. The fine motor skills needed to build block towers however, were not affected by post natal growth [48]. Another cross- sectional study among Zanzibari infants found that among children who were already walking, stunted children had less total motor activity and spent less time in locomotion than their taller counter parts [49]. Another longitudinal study in Guatemala that standardized infant development scores to assess mental and psychomotor development 4 times through the first 3 years of life found that psychomotor development was strongly associated with HAZ as well as growth throughout the first 2 years of life but that this relationship was less strong for mental development [50]. The authors concluded that the weaker association between growth and mental development may be due to the fact

11 that the effect of chronic growth faltering on mental development may occur later than the effect on motor development.

Neuropsychological tests suggest that the relationship between stunting and cognitive performance varies by cognitive function [51]. Specifically, while stunted children performed poorly on most neuropsychological tests, the rate of cognitive development differed between functions. Executive and visuo-spatial functions, considered higher level processing had a particularly slow rate of development in stunted children. In contrast, verbal comprehension as well as learning and memory for verbal and visual material were similar in stunted and adequately nourished children.

Longitudinal studies that specifically examined the lasting effect of early childhood stunting on cognitive scores later in life have uncovered the importance of severity and persistence of stunting on cognitive function. A study that examined the relationship between stunting and cognitive skills at multiple time points in a child’s life highlighted that while stunting early in life was related to lower verbal and quantitative abilities, being stunted at school entry had a stronger negative impact on cognitive test scores [52]. Mendez et al [53] further demonstrated that child stunting in the first 2 years of life impacted school enrolment, school performance and cognitive test scores at age 8 and again at age 11. Importantly this effect was pronounced for children that were more severely stunted and for whom stunting started earlier in life. Importantly, schooling may act to buffer the negative effects of early malnourishment in stunted children as the effect of schooling on cognitive test scores was the same for stunted and non-stunted children. Thus, in order to minimize the negative effect of childhood stunting on cognitive development, it is important to address the social perceptions that often lead to late enrolment of stunted children [54].

12 2.1.3 Long term consequences of stunting.

Stunted children are more likely to become shorter adults and for women especially this has a direct effect on the size of their children [55]. Maternal stunting leads to Intrauterine Growth Restriction (IUGR) [55] which increases the risk of complications during pregnancy [1] as well as fetal and neonatal morbidity and mortality [56,57]. Furthermore, mothers who are short for their age are more likely to have stunted children [57] creating an intergenerational cycle of poor growth.

In addition to the physical consequences of stunting on adult size, the detrimental effects on cognition early in life have long lasting implications for intellectual functioning and school achievement later in life [45]. A recent review and meta analysis by Victora et al [58] reported that slow growth in early childhood had long term consequences for adult educational achievement, graduation rates and high school performance. Adult economic capacity and human capital have also been linked to childhood stunting. In Brazil and Guatemala, 1 SD increase in HAZ was associated with an 8% increase in adult income and in India adults who were taller as children had more household assets later in life [58]. A follow up study to the INCAP Oriente nutrition intervention in Guatemala also provided the opportunity to examine the consequences of childhood stunting on adult human capital and economic productivity. First, it was shown that children who received a high energy and protein supplement compared to a lower energy, protein deficient drink in the first 3 years of their lives had a higher growth rate and were taller for their age [59]. The human capital follow up study re-examined these children between the ages of 26 and 42 and found that adults who had received the nutrition intervention as children had improved reading comprehension and intelligence tests, more years of schooling(women), bigger body size and capacity for heavy work (VO2 max) as well as 46% higher wages (men) [60]. Therefore, the negative

13 physical and cognitive effects of stunting in early childhood have intergenerational implications.

In summary, nearly 1/3 of children under 5 in developing countries are unable to reach their growth potential due to a nutritionally inadequate diet and frequent clinical or subclinical infections. Importantly, poor health during this crucial growth period has long lasting effects on cognitive and motor development which in turn influence educational and economic capacity. Furthermore, maternal short stature not only increases the risk of adverse pregnancy outcomes but also leads to IUGR and smaller children thus continuing an intergenerational cycle of poor growth. Accordingly, interventions that address the interaction of malnutrition and infection not only improve child health but can contribute to poverty reduction.

2.2 Gastrointestinal infections: Biology and risk factors for transmission

Gastrointestinal infections (GI) not only contribute to stunting but stunted children are also more susceptible to infection [1,15,61].

Soil Transmitted Helminths (STH). The three most common STH infections Ascaris lumbricoides, Trichuris trichiura and the hookworms, Ancylostoma duodenale and Necator americanus, infect 2-3 billion people globally [62] and are the target of preventive chemotherapy due to their impact on child and community development [62]. For all three infections, people become infected when infective stages from the external environment enter the body either through the ingestion of eggs or when larvae enter the bloodstream through the skin. After entry into the host, larval stages make their way to the intestine to mature into adult parasites. Adult worms lay eggs that are passed with the feces and develop in the external environment into infective eggs or larvae (hookworm) [63]. Several factors uniquely distinguish the STH infections.

14 Ascaris is particularly prolific, producing over 10 times the eggs of the other STH infections [63] whereas hookworm exists as a larval stage in the external environment and is thus more sensitive to desiccation in dry conditions [64].

The STH infections share epidemiological traits. Of particular note is their aggregated distribution within the host population in endemic areas that results in a few individuals carrying the majority of the infection burden [65]. Age in particular has been associated with aggregation of STH infection in a population. For Ascaris and Trichuris, 5-10 year old children harbour the greatest burden of infection whereas for hookworm, infection burdens peak in adolescence and early adulthood [65]. Predisposition to infection, the correlation between an individual’s pre and post treatment infection burden, is a trait of STH infections that contributes to this overdispersed distribution of the parasites in the population [66]. Mechanisms of predisposition relate to exposure to infectious stages and susceptibility to infection. Susceptibility to infection can be of genetic [67-69] or nutritional origin [70-72]. Individuals predisposed to heavy infection have differences in the genetic regulation of B cell activation and immunoglobulin secretion have been related to predisposition to heavy infection [69] and impaired immune function due to micro and macronutrient deficiencies [15]. Behavioural factors are strongly linked with the exposure mechanism of susceptibility. Factors such as geophagy [73,74], poor hygiene [75], household crowding [76,77] and not wearing shoes [78] directly increase the likelihood of an individual being exposed to an egg or larva. Furthermore, factors associated with poverty [79,80] such as poor latrine access [79,81] and poor household construction [82] increase the likelihood that infectious stages reach the environment thus increasing risk of exposure to infection.

Spatial clustering of infection has also been noted for STH infections at the household [83], regional and national levels [84,85]. Household level socio- economic factors are considered to be the most influential of small scale spatial clustering [86-88] however at larger scales biophysical and climatic factors 15 influence parasite dispersal. Specifically, soil with lower clay content is associated with a higher prevalence of hookworm infection, likely due to the increased porosity of the soil that allows larvae to migrate vertically in the soil and avoid migration [89]. Vegetation cover also provides the moisture and shade that promote egg and larval survival [88,89] and climatic factors such as rainfall and temperature are also associated with STH infection [84,90,91]. Thus, it is a combination of individual, household and regional factors that determine the burden of STH infection and therefore need to be addressed for effective control.

Diarrhea and protozoan infection. Diarrhea is responsible for 15% of all deaths in children under 5 years of age [92]. Viral, bacterial and protozoan pathogens cause diarrhea through the fecal-oral transmission route, spread most often through water [75]. Pathogens spread by the fecal oral route are transmitted from a focus of infection (excreta) to the host primarily through soil, water, arthropods and hands [93]. Importantly, if the point of infection is isolated from the susceptible population, transmission cannot occur [93]. Indeed, the primary risk factors for diarrheal and protozoan infection are related to sanitation and hygiene, specifically poor hygiene practices [94,95] and lack of sanitation infrastructure [96,97]. Interventions that address these risk factors by improving drinking water, access and use of sanitation facilities as well as water quality and hygiene behaviours have all been shown to reduce diarrheal disease, especially point of use water quality interventions [98,99].

In summary, GI infections are transmitted through a fecal-oral pathway that is influenced by socio-economic and biophysical. Interventions that improve access to clean water and sanitation infrastructure will be important in helping endemic regions escape poverty.

16 2.3 Rural poverty landscape 2.3.1 Rural development and the changing shape of the rural landscape

Malnutrition and ill health are recognized as two of the manifestations of poverty however increasingly, a multidimensional characterization of poverty is being used. The Programme of Action of the World Summit for Social Development (United Nations, 2006, resolution 1, annex II) recognized that poverty results in low access to education and basic services as well as resources and safe housing in addition to the lack of participation in civil, social and cultural life. Consequently, poverty reduction strategies have made steps towards the integration of economic and social policies that promote sustainable economic growth at the same time as increasing employment opportunities [100]. Programs such as microfinance and conditional cash transfers have been widely implemented in an effort to increase people’s economic and human capital. Specifically, by providing a short term economic incentive in combination with required participation in education and health programs it is believed that recipients will have the capabilities and resources to effectively escape poverty [100,101]. Central to the success of these programs is the ability to access schools and health facilities as well as markets with sufficient employment and commercial potential.

Transportation infrastructure is instrumental in providing access to the markets and social programs that generate economic and human capital. Indeed, improved rural roads help to lower the cost and time needed for transportation, making it easier to access financial markets and goods as well as increasing employment opportunities in the agricultural and non-agricultural sectors [102-104]. Furthermore, rural road networks aid in the establishment and attendance of schools and health centres as well as the creation of sanitation and water systems [102]. Studies that examine rural road development projects suggest that roads are valuable tools in rural poverty

17 reduction. In China, Jalan et al [104] found that road density was one of six geographic variables that could predict the existence of a “poverty trap”, ensuring that the household’s consumption could not rise over time. A more recent longitudinal case control study in Bangladesh quantified the poverty reduction benefits associated with 2 rural road development initiatives [103]. Participant communities had an average poverty reduction 3-6% greater than control communities over 5 years. Furthermore, Khandker and colleagues demonstrated improvements in employment, education and household consumption after controlling for village level characteristics. Specifically, household transport costs were reduced, crop output and price indices increased and employment opportunities and wages increased in communities with improved access, as did male and female secondary school enrolment and household per capita consumption. Importantly, the poorest households benefited the most from improved road access. Thus, the authors demonstrated a significant positive, short term impact of 2 rural road improvement programs however they also pointed out that understanding long term effects such as the expansion of rural markets and rural-urban migration will provide further insights on the impact of road development on rural landscapes.

The combination of increased transportation networks and the ensuing economic opportunities are instrumental in shaping regional rural development through changes to population mobility and migration. Historically, demographic studies examined migration from a dichotomous rural –urban perspective and considered population movement to be motivated through economic self-interest or influenced by market economics. More recently however, analysis of migration that includes cultural considerations has painted a more nuanced picture of migration and how it shapes the socio-spatial rural landscape [105,106]. An in depth analysis of regional development in the Amazon highlighted that population mobility was one of the most important catalysts of regional change and that physical infrastructure was influential in

18 migration and settlement patterns [106]. Specifically, out migration increased with proximity to a road, however rural communities that were larger and had more livelihood enhancing services and infrastructure had less migration. Furthermore, these analyses highlighted that migration to seek off –farm employment was a way of diversifying income sources and as such localized rural-rural migration that led to the development of small rural towns was common. Within this context, the traditional rural/urban dichotomy is considered insufficient to characterize the more complex socio-spatial rearrangements occurring in the changing rural landscape and as such it is suggested that policies should be regionally based rather than focused on “urban” or “rural” development [106].

Thus, as rural poverty reduction programs are put into place, the rural landscape is also changing. Improved transportation networks improve economic and educational opportunities but also increase mobility and migration, transforming the demographics of rural areas.

2.3.2 Rural development and health

These changes to the rural landscape also have implications for human health. In addition to economic and educational benefits, rural roads improve access to health facilities as well as sanitation and hygiene infrastructure [102]. Roads reduce travel time and travel costs, both of which are influential in decisions to use health care among rural women [107]. Indeed, access to public transport roads was more important than distance to a health facility for decreasing infant mortality in rural Kenya, presumably due to the effect it had on ease of transport [108]. Importantly, despite better physical access to health clinics, social and economic barriers at the clinic itself may reduce the positive effects associated with better roads [109].

Improved access to sanitation and hygiene infrastructure, considered “forgotten foundations in health” [75], also help to reduce child morbidity and

19 mortality [110,111]. Better water supply, latrines and sewage systems directly reduce the transmission of viral, bacterial and parasitic infections [75,98,112], in addition to indirectly impacting disease by encouraging behaviours such as hand washing [113,114]. Roads often improve access to sanitation and hygiene infrastructure because of the ease of transporting building materials and furthermore, they increase population mobility resulting in greater exposure to new ideas and the “urban lifestyle” which may in turn increase latrine adoption and diffusion [115,116].

In contrast, roads have also been linked with negative health outcomes. The construction of roads themselves as well as changes to population mobility and migration can increase the prevalence of multiple types of infections. Deforestation associated with road construction acts to increase breeding grounds for mosquitoes and other vectors, thus leading to an increase in vector borne diseases [117,118]. Sexually transmitted diseases on the other hand are related to increased human mobility and rapid population growth that increases contact between infected and uninfected individuals [118-120]. Furthermore, the increased population density in some settlement areas that results from higher migration rates is also an important determinant of pathogen transmission [121,122]. Two studies from rural Ecuador specifically examined the link between enteric pathogens and road proximity [123] or population density [124] in a rural developing country context. Eisenberg and colleagues [123] detected higher rates of bacterial, viral and parasitic enteric pathogens in villages closer to the main road compared to more remote villages. Possible causal pathways were the reintroduction of pathogens from increased population movement outside the village or poor community management of sanitation and hygiene infrastructure that resulted from lower social cohesion in these communities. Bates et al [124] demonstrated that diarrheal prevalence was higher in more densely populated communities but also highlighted the importance of social connections for reducing infection risk.

The shift to a more commercial economy which often accompanies road access also has positive and negative consequences for child nutritional status, primarily through the influence on diet. Diets in urban areas as well as rural areas with increased commercial activity tend to have fewer seasonal caloric fluctuations as well as more processed foods and animal products [125,126] but less grain and wheat products [125]. In comparison, more traditional diets are often higher in fruits and vegetables as well as complex starches such as tubers [126]. The shifts in diet as well as lifestyle that accompany the changing community economy are on one side associated with a reduction in child malnutrition [125] but on the other an increase in the prevalence of obesity and overweight [127,128]. This phenomenon, known as the nutrition transition, is common in countries undergoing the transition to more commercial economies [128] and is characterized by a dietary shift from a high consumption of complex carbohydrates and fiber to higher intakes of fats and sugars [127]. Traditionally, high fat diets were unique to wealthier societies however with an increase in the availability of cheap vegetable oils, fat consumption in low income societies has also increased and is considered a growing public health concern in developing nations [127].

Taken together, it is well recognized that poverty is a multidimensional phenomenon and thus poverty reduction strategies focus on improving both economic and human capital. Rural development programs that improve economic opportunities and access to physical infrastructure also change the demographics of these rural landscapes, with positive and negative consequences for child health. Thus, studying child health in these dynamic environments necessitates a more comprehensive view of health.

21 2.4 Towards an integrated view of health In the 1970’s growing awareness of the influence of social, cultural and economic factors on health outcomes led to the goal of Health for All By the Year 2000 which emphasized the importance of intersectoral actions for health. This resulted in health initiatives that brought together researchers from multiple disciplines especially from the health and social sciences and that could largely be characterized as multi-disciplinary or interdisciplinary [129]. Multi- disciplinary research is defined as researchers working in parallel or sequentially from discipline-specific backgrounds [129,130] such as entomologists consulting social-psychologists for insight into behaviour change theory in a malaria control program. Interdisciplinary research differs in that researchers work together on a common problem however this is still done from discipline-specific backgrounds. While useful in short term problem solving, the seminal work of Rosenfield [129] highlighted the limitations of both multidisciplinary and interdisciplinary approaches for generating long term programmatic changes and generating theory that could cope with a dynamic view of human health, and introduced the concept of transdisciplinarity. In contrast to multi- and interdisciplinary work, using a transdisciplinary approach, researchers work jointly on a shared conceptual framework that incorporates disciplinary-specific theories to address a common problem [129]. Since this initial introduction, transdisciplinarity has been further developed to become an integrative approach [131] that requires a collaborative process [131,132] that incorporates social, economic political and environmental factors to adequately consider health from an individual to a societal level [129,130,133].

The increasing acknowledgement of the complexity of health problems saw the adoption of a more “ecological” perspective in multiple disciplines as well as the development of truly transdisciplinary frameworks. In public health, eco-epidemiology sought to establish local understandings of the multiple interactive systems that determined human health rather than finding universal

22 explanations. Ecological public health also focused on interactions and interdependence between humans, health and the physical and social environments but saw health as a pattern of relations rather than a quantitative outcome. In health promotion, an ecological perspective recognized the need for interventions that focused on multiple levels in order to improve change the multiple interacting factors that influenced health behaviours [134]. The fields of epidemiology, medical geography and medical sociology were particularly interested in the relationship between poverty and health and therefore focused on understanding the relationship between health and place [135-137].

Traditional approaches saw the relationship between people’s health and their location as either a collection of the traits of the individuals who lived there (compositional) or as a result of the characteristics of the place (contextual) [138-140]. More recently however, the composition/context dichotomy has been labelled as a false dualism that neglects the interactions that occur between an individual and where they live [140,141]. For example, children growing up in a neighbourhood are shaped by the social norms, values and educational opportunities of that area which in turn are likely to affect how that child will interact with that place as an adult [141]. This has led to a shift in the theoretical and conceptual models of ‘health and place’ research that emphasizes the dynamic, multidimensional nature of this relationship [140,141]. By focusing on the processes and interactions occurring between people and places over time this school of thought moves away from using scale specific, static indicators of place (i.e. politically designated regions) and seeks to recognize how an individual’s relationship with their environment may change over time (i.e. movement patterns) and furthermore how the structural context itself is influenced by political, economic and social forces [141]. It is believed that this multifaceted view of health will be especially beneficial in determining the appropriate scope for public health interventions as well as identifying the

23 individual traits that can moderate contextual factors for improved health in impoverished environments.

One research approach stands out in its emphasis on transdisciplinarity. Ecohealth or Ecosystem Approaches to Health evolved from an ecological perspective and broadens the definition of health to view “human health as embedded in the wider pursuit of ecosystem health” [142]. This body of work draws on ecological and complex systems theory to create a conceptual model of interconnected, nested systems increasing in scale from genes and cells to organisms, species, communities and ecosystems [142-144]. ‘Health’ is considered to emerge from the interactions between these systems and is greater than the sum of the individual parts. Thus, similar to the ‘health and place’ literature, Ecohealth aims to assess the relationships and interactions between social and ecological factors across temporal and spatial scales to create an integrated analysis of the variables influencing human and ecosystem health [145,146]. It is because of this broad definition of health, that Ecohealth has made transdisciplinarity central to the research approach. Specifically, it is recognized that investigations require multiple, diverse perspectives to create a richer understanding of the interactions of interest [147]. Thus, research projects often draw from multiple disciplines (i.e. biology, public health, anthropology) and include participatory methodology to include participant perspectives of health in addition to those of the researchers [146,148].

This multi-dimensional, dynamic view of health is considered particularly worthwhile for pragmatic fields due to its utility for identifying management and policy priorities as well as designing public health interventions. In practice, the dynamic nature of transdisciplinary research approaches can lead to difficulties in implementation. First, the fact that transdisciplinarity is considered to create an evolving, context-specific view of health results in a product (and process) that is continuously scrutinized, evaluated and reviewed [130]. In maintaining an adaptive framework, transdisciplinarity also lacks specific “how to” models of the 24 research process [132,133]. The transdisciplinary teamwork inherently requires more time, cooperation, physical proximity and bureaucratic flexibility than disciplinary research and all of these factors have been highlighted as barriers to this form of research approach [132,133]. Finally, the need to frame research more narrowly for publication and a lack of evaluative tools that capture systemic processes are difficulties that need to be overcome [132,133].

2.5 Tools for deciphering the complexity of health Although transdisciplinary research projects related to health may differ in their conceptual approaches, they share several methodological and analytic tools.

2.5.1 Mixed methods

Traditionally, health research has used 2 basic types of inquiry: quantitative and qualitative. Quantitative research seeks an objective perspective and relies on the numerical measurement of phenomena through controlled experiments and representative sampling procedures. In contrast, qualitative research aims to gain “in-depth” perspectives of certain group’s experiences, whereby recognizing the inherent biases present in observation [149]. More recently, a third methodology that combines quantitative and qualitative methods has been recognized in its own right as multi-methods, multi-strategy or mixed methods [150,151].

In an effort to promote the development and validation of mixed methods research there has been an increased effort to standardize research designs to clarify the intent to combine both quantitative and qualitative data in a study [152]. The six common study designs are differentiated by three design components where the implementation of quantitative and qualitative data

25 collection can occur sequentially or concurrently, with a priority on either or on both depending on the theoretical framework.

Sequential studies have two distinct phases of data collection and can either be explanatory, exploratory or transformative in nature. Explanatory studies put the priority on the quantitative data and the qualitative data is used primarily to assist in explaining and interpreting the quantitative results. In contrast, exploratory studies place priority on the qualitative data collection and the quantitative findings assist to test elements of a new theory or to generalize qualitative findings to different samples. In transformative studies, the theoretical grounding drives the choice of design and methodology thus an explanatory or exploratory process could be adopted depending upon the theoretical framework.

Concurrent studies have one phase of data collection in which quantitative and qualitative data are collected simultaneously. Concurrent triangulation studies seek to offset the weaknesses inherent in either quantitative or qualitative study designs by using both methods to confirm and corroborate findings within a single study. Concurrent nested studies give priority to either study methodology. The less dominant form of inquiry either answers different questions or seeks information on different levels than the dominant methodology. Concurrent transformative study design is similar to the sequential transformative design in that it can adopt either a nested or triangulation format depending on the theoretical construct driving the investigation.

More recently an effort has been made to move to an integrative design that will generate results that go beyond description to conceptualization [153]. This research design gives an equal emphasis to qualitative and quantitative inquiry and seeks to increase the parallelism between these methods throughout the study design. This process begins with a unified conceptualization of the

26 information sought which is followed by an integrated data collection and puts an emphasis on recontextualizing statistically derived results to the qualitative context for a more fully integrated, richer understanding of the results that goes beyond the sole use of qualitative or quantitative inquiry.

From its beginnings as the use of qualitative methods to interpret quantitative results, mixed methods research has grown in popularity [150,154]. In practice, mixed methods studies fall on a gradient from qualitative dominant to quantitative dominant [151] where the degree of qualitative versus quantitative methods depends on the research objectives. In a review of 75 mixed methods studies in Health Services Research from the UK, O’Cathain et al [154] found that methodological complementarity was the primary reason for incorporating mixed methods in a research project. Specifically, researchers sought to increase the breadth of questions that could be addressed and to assist in the development of a more comprehensive research design. Quantitative methods were used primarily to test effectiveness of an intervention, describe patterns, explain variability and determine sample sizes for qualitative components. Whereas, qualitative methods were used for exploratory purposes, to develop an intervention or explain relationships detected in quantitative studies. Importantly, mixed methods are considered particularly useful in applied fields due to the pragmatic advantages for addressing ‘complex’ health environments [154,155].

The pragmatic, comprehensive nature of mixed methods research has made it an integral component of “health and place” and Ecohealth research [139,148]. To capture the dynamic relationship between people and place, Cummins et al [141] call for a “relational approach” that assesses an individual’s experience or perceptions of a place rather than the use of static geographically bound descriptions of a location. This approach recognizes that how an individual relates to their environment may be just as important as the physical space they occupy and furthermore that different groups of individuals may 27 relate differently to the same physical environment. Indeed, by incorporating qualitative methods to inform quantitative analysis, researchers have demonstrated that mental health may influence sensitivity to the effects of poverty [156] and that the racial or ethnic composition of neighbourhoods can influence food shopping patterns [157]. The more personal understanding of an individual’s experience in an environment and what this means for health that comes from mixed methods research is especially relevant in the development of health interventions [141].

Ecohealth research views health promotion as the management of socio- ecological systems [142], thus it is grounded in ecological and complex systems theory. The participation of stakeholders in an Ecohealth research project is considered essential in identifying the diverse processes influencing health and furthermore, is believed to generate the skills and interest in ongoing management practice [142,148]. Studies in the Peruvian Amazon and Kathmandu have demonstrated the importance of including multiple perspectives for understanding the structure of the socio-ecological context of health. In Peru, conventional land use analyses described a system dominated by small-scale cattle ranchers, however after incorporating multiple stakeholders in the investigation, timber extraction was identified as the main economic activity of the region [158]. In Kathmandu, focus groups generated diagrammatic representations of the social and ecological processes influencing Echinoccocus transmission which served to identify the social and cultural factors that led to the failure of an existing public health intervention [159]. Thus, mixed methods research is of great benefit in health studies that seek to include a multi-dimensional view of the dynamic processes influencing health. Specifically, the use of qualitative inquiry to inform quantitative analysis is especially beneficial to develop policy that best reaches the target population.

28 2.5.2 Descriptive spatial mapping

Central to multi-dimensional investigations of health is the need for multiple levels into the analysis. By including individual and contextual variables in multi-level or contextual analysis researchers are able to examine the relative influence of each type of variable on health outcomes as well as how individual traits may moderate how their environment impacts their health [139]. To most accurately answer these questions, the choice of context or “neighbourhood” should be based on the proposed mechanism of influence that this context has on the health outcome [139], however this is not always an easy task. Especially in large scale investigations, researchers often rely on politically defined administrative boundaries that may not capture the specific mechanism of interest [138]. Including mapping in health research has helped to define contextual variables that relate specifically to the health mechanisms of interest. For example, mapping pesticide exposure risks [160,161] or food accessibility variables (ie grocery stores, healthy food outlets) [162-164] has allowed health researchers to link cancer and obesity to the most salient context variables.

Disease surveillance is another mapping approach commonly used in health research that identifies spatial and temporal changes in disease patterns to classify areas of research interest or areas in need of public health action [165,166]. Scan statistics are one of the most commonly used methods to identify clusters of disease and do so by comparing disease distribution within a scanning window against an expected rate of disease occurrence over the entire study area [167,168]. In addition to detecting epidemic outbreaks [169], small scale analyses can be useful in isolating small scale patterns that differ from larger generalized trends. For example, household level data collection found high risk zones of malaria transmission that persisted despite a general decrease in infection at the village level, thereby facilitating a targeted intervention [170]. A study of Giardia infection in Canada found that clusters of infection did not correlate strongly with the common risk factors, thus discovering an area where 29 additional research was needed to elucidate the transmission pathway [171]. Thus, by identifying context variables that are of particular relevance to disease transmission and by locating specific areas with unique disease patterns, descriptive mapping is a valuable tool in public health research.

2.5.3 Variable reduction methods

Studies of poverty and health often rely on the comparison of measured data with standard values to determine an absolute status of poverty [172] or dietary adequacy [173]. In areas where poverty levels are high or diets are well below the developed standards, these tools do not allow for the distinction between people that are poor and less poor in “wealth” or dietary adequacy. To address this issue and allow the study of context relevant patterns in wealth and diet, nutrition and poverty researchers have developed variable reduction methods to characterize households or individuals with respect to their assets [174,175] or foods consumed [176,177]. The primary variable reduction procedures used in these fields are Principal Component Analysis (PCA) and Factor Analysis (FA) that while similar in procedure differ in their assumption of a causal mechanism. FA assumes that there are underlying latent variables that cause the correlation between the measured variables whereas PCA makes no assumption about how the variables group. PCA is the most commonly used method in the nutrition and poverty fields thus I will focus on this method in a brief overview before describing discipline specific details below. PCA is a multivariate variable reduction strategy that is used to create artificial variables or components that account for most of the variance in measured variables that have a degree of redundancy. Perpendicular components are derived to account for the maximum amount of variance among the observed variables where each subsequent component captures an additional, unique dimension of the data. For each component, weighted coefficients are determined for all measured variable indicating their relative contribution to the variance in the measured data. The final number of components is equal to the number of 30 measured variables and plots of the variance explained by each component are used to determine the number of components kept for subsequent analyses. Asset-based indices of household wealth use only the first component [175] whereas multiple components reflect different dietary patterns present consumed by a population [177].

There are two main approaches to measuring poverty, absolute poverty indices indicate where a person or household falls relative to a minimum standard, whereas measurements of relative poverty indicate a person’s socio- economic position relative to their community at a given time [172]. Measurement of absolute poverty is relevant in countries or regions where poverty is not wide spread, however in extremely impoverished areas relative poverty indices enable a distinction between poor and less poor, something especially important in the assessment of poverty reduction strategies [175]. In theory, the best indicator of household welfare is actual consumption (sum of household production minus any items attained from other sources) [175] however this indicator is especially difficult and time consuming to measure. Measures of income or expenditure are commonly used as proxies for consumption however they do not necessarily reflect the living standard as they only assess money coming into the household but do not capture bartered or self produced goods [175]. Furthermore, in areas where household income is often inconsistent these measurements may not give an accurate estimation of household wealth [172].

In response to the above mentioned problems, indices that indicate ownership of household assets as well as dwelling characteristics have been developed to reflect longer term household capital [172]. A variety of scores of varying complexity have been created ranging from a simple summation of items possessed, to qualitatively weighted sums of possessions, to statistically weighted indices [172,174,175,178]. Statistically weighted asset-based indices of

31 wealth have become the most commonly used method for measuring relative poverty in developing countries, with PCA being the primary statistical method of generating asset weights [175,179]. It is important to recognize that due to regional and cultural variability in the range of possessions, the resultant index is location specific and should not be used to compare between countries [172,175] or even between vastly different regions within countries such as urban and rural areas [179]. Despite these caveats, when assets selected are correlated and have sufficient variability in their distribution among households [179] they may have a stronger explanatory power than expenditure data for rural health outcomes [178]. Indeed, in recent public health studies in rural areas asset-based indices of wealth predicted health outcomes such as nutritional status [35,180] and parasitic infections [181,182].

Nutritional epidemiology seeks to understand the relationship between food and nutrient intake with disease and anthropometric outcomes [173]. Research in dietetics and nutrition relies on two types of dietary assessment tools. The first provides a quantitative measure of daily consumption either through dietary food records or 24 hr recalls. Food records are completed by the individual as they consume the foods and as such may influence dietary choices whereas 24 hr recalls are conducted by a trained interviewer and ask the individual to recall food consumption from the previous day [173]. Thus, recalls do not influence food choices however they rely on respondent memory for foods and portion sizes consumed [183].

The second group of dietary assessment tools provides a retrospective analysis of patterns of food use over a longer period of time, using Food Frequency Questionnaires (FFQ) or diet histories. Food frequency questionnaires provide qualitative data on usual intakes of foods or food groups over a longer period of time. When portion sizes are included in the FFQ, they become semi- quantitative and can be used to assess nutrient and energy intake however 32 greater measurement error in FFQ results in less accurate measures of intake than those calculated using recalls or records. Dietary histories combine a 24 hour recall of actual intake with information on the usual eating pattern to describe usual food intakes over a longer time period [173,183].

Studies on the validity of dietary assessment methods have compared reported energy and nutrient intake with standards such as biological markers or recorded or weighed amounts. Generally, 24 hour recalls and dietary records were found to be more accurate than FFQ in assessing dietary intake [183-185]. Both 24 hr recalls and food records were found to under report energy consumption by 3-34% whereas the range of underreporting of protein consumption is between 11-28% [183]. Over reporting of energy intake has also been reported, occurring by 7-11% in studies reviewed by Burrows et al [184]. Despite the error inherent in recalls and food records, when done over multiple days they are still considered the more accurate method for assessing energy intake in young children [184].

Estimating usual nutrient intake in developing countries poses additional challenges to nutritional epidemiologists due to irregular food consumption patterns. In particular, subsistence farming populations that rely on seasonally available, locally produced foods have highly variable diets, especially due to irregular inclusion of animal products [186]. As a result, estimates for macronutrient and micronutrient consumption in pregnant women in Malawi required 8-23 days and 95-213 days of recall data respectively in order to estimate macronutrient consumption within an error range of ± 20% [186]. Importantly, dietary variability varies among developing countries. Studies in Indonesia estimated that 6 days of recall data were sufficient for estimates of macronutrient consumption, likely due to the limited number of foods consumed in this population [187]. Thus, accurate assessment of usual nutrient intake in developing countries requires an understanding of the primary food sources in the study region. 33 Measurement of usual food and nutrient intake at the individual level is especially of interest to nutritional epidemiologists who examine relationships between diet and health. Usual intake data requires knowledge of dietary consumption over a period of time thus requiring multiple days of dietary recalls or FFQs. Importantly, a priori knowledge of the intra-individual variability in food consumption is important in choosing the most accurate method of dietary assessment. If the foods or nutrients of interest are consumed infrequently a longer period of analysis is necessary to capture intake of these items [173]. For this reason, the longer time frame included in a single FFQ makes this method an easier tool to administer for determining patterns in diet however if resources allow, multiple recalls provide a more accurate assessment of usual intake.

The traditional approaches described above have focused on quantifying specific nutrients and foods however this neglects the potential interactions or synergies that may occur between nutrients in whole food diets and is less helpful for public health dietary recommendations [176,177]. In response to this concern, the assessment of eating patterns has emerged in nutritional epidemiological studies with 2 overarching approaches to identify dietary patterns: score-based or data driven. Score-based diet patterns are tallies that assess the diversity or variety in a diet or assess the degree of similarity between the diet and current nutrition guidelines. In contrast, data driven diet metrics use multivariate statistical analyses to group foods consumed together (factor analysis or PCA) or individuals that eat similarly (cluster analysis). Thus, score- based diet patterns often measure adherence to dietary guidelines whereas data driven patterns reflect foods being consumed and do not necessarily reflect optimal diet patterns [176,188]. Although dietary pattern analysis has positive implications for nutritional epidemiology, there are several limitations for data- driven approaches, primarily relating to the subjectivity that comes from the arbitrary decisions made by the researcher. In particular, during data analysis the researcher must make decisions about how to group foods into categories,

34 which quantities to use (grams, frequency, % energy), the number of diet scores (factors) to keep and how to label components. In addition, the results of the analysis are context specific, likely to vary with age, sex, SES and ethnicity and studies on the validity and reproducibility of diet patterns are still rare. Of the studies that examined validity and reproducibility, the results were promising. Two studies found reproducible patterns using 2 FFQ and both found that the major patterns identified using FFQ data were validated by patterns generated using diet records [189,190]. Furthermore, Hu et al [189] found that the identified food patterns (Prudent and Western) were correlated with plasma concentrations of biomarkers.

Despite the shortcomings mentioned above, recent reviews of Dietary Pattern Analysis (DPA) in the adult population suggest that whole food diet patterns are correlated with nutrient intake and beneficial in predicting risk of chronic diseases [176,177,188,191]. Specifically, despite differences in the actual foods that comprised a “healthy” diet pattern, across cultural settings these patterns were associated with lower levels of disease, cancers, mortality and obesity [177]. Thus, while further operational research is needed, it is agreed that DPA is a valuable tool in nutritional epidemiology when it addresses appropriate questions and detailed reports of the subjective steps in the analysis are provided [176,188].

Although DPA has proven beneficial for studying adult health, the use of DPA in investigations of preschool child diet is less common, especially in developing countries. In a review of child diet pattern research in developed countries, Smithers et al [192] found that similar to adult studies, “healthy”, “unhealthy” and traditional patterns emerge from DPA of children under 5 years but that few studies have linked diet patterns to health outcomes. For studies that did, the findings were mixed. Higher “healthy” index scores were associated with infant weight gain [193], better health status [194] and increased lean mass but not BMI [195]. “Junk food” patterns were associated with hyperactivity 35 [196] and meat patterns were associated with obesity [197] and overweight [194]. To date, the use of DPA for preschool children has focused on investigations of obesity in the developed world however the cultural specificity of DPA may be beneficial in capturing consumption patterns that may differ greatly from dietary guidelines. Furthermore, DPA avoids the use of nutrient composition data that may be inaccurate [198] or ineffective in capturing geographic or species variability in food nutrient composition [199,200]. Diet diversity (DDS) and food variety scores (FVS), considered whole food indices of dietary quality [201], have been used in developing countries to examine the association between diet and health outcomes. A large scale analysis of Demographic Health Survey (DHS) data found that in multiple countries diet diversity influenced child stunting independent of SES [9]. More recently, community studies have demonstrated some additional subtleties to this relationship. Studies in Bolivia and Burkina Faso, suggested that dietary diversity was most important in children under 2 years of age [37,202] whereas in Mali the beneficial effect of DDS on child growth was only detected in the urban areas [38]. It was suggested that the poor diet diversity in rural areas limited the ability to detect the effect on child growth however despite an overall poor diet in Bangladesh, high DDS was still associated with reduced odds of being stunted for children under 5 years of age [10]. Thus, whole food indices of dietary quality have shown promising results for linking diet and child health in developing countries however the inclusion of methods that capture quantities of food groups consumed rather than just counts will be beneficial in furthering our understanding of diet and health outcomes.

In summary, studies that recognize the complex, dynamic nature of the relationship between human health and the socio-ecological environment rely on a diverse tool box to investigate health. The integration of quantitative and qualitative forms of inquiry provides greater resolution in identifying the factors most influential to health. Descriptive mapping can be particularly useful in

36 identifying priority areas or variables for health research, primarily by examining spatial and temporal clusters in health outcomes or risk factors. Furthermore, indices of asset-based wealth and whole food diet patterns provide context specific metrics that identify heterogeneities among populations that are commonly grouped as “rural” and “impoverished”.

2.6 Panamá: Comarca Ngäbe Buglé 2.6.1 Socio-political background

Located where the Central American isthmus narrows, Panamá’s strategic geographical position has contributed to its position as one of the fastest growing economies in Latin America. Fuelled primarily by income generated from the Panamá canal (3.4% of GDP in 2009), Panamá experienced an 8% growth in Gross Domestic Product (GDP) from 2006-2010. However despite this growth, Panamá falls below the Latin American region in poverty reduction with one third of the population still living in poverty. These inconsistencies are largely considered a result of inequalities in income distribution reflected in Panamá’s position among the top 20 countries in income inequality [100]. Poverty is concentrated in the rural areas, particularly the rural indigenous areas where 96% of the population live in poverty compared to 51% of the rural non- indigenous population and 18% of the urban population [203]. A World Bank Poverty Assessment [204] that examined data from the Living Standards Measurement Survey (LSMS) found that the poverty and inequality recorded in Panamá reflected disparities in labour, education and health care as well as physical and financial assets. Rural indigenous regions, where 8% of the population accounted for 35% of those in extreme poverty, had the poorest living standards. Less than half of the indigenous population had access to an improved water supply and sanitation infrastructure, 16% of individuals entered secondary education [204], infant mortality rates were 3 times the national

37 average and 57% of children under 5 yrs of age are chronically malnourished [205].

In 2005, the Torrijos administration adopted the “Strategic Vision of Economic and Employment Toward 2009”, making poverty reduction and equal income distribution key priorities [205]. The main tool in this effort, the Red de Oportunidades, was launched in 2005 and includes conditional transfer (CT) programs combined with basic infrastructure expansion. In a 2007 assessment by the World Bank, it was noted that the CT programs and social services were being provided but infrastructure improvements were still lacking [205].

2.6.2 Indigenous peoples of Panamá – Focus on the Ngäbe Buglé

Panamá’s indigenous peoples make up 7 indigenous groups; Kuna, Bri-Bri, Wounaan, Emberra, Ngäbe, Bugle and Naso. Starting in 1972 the Panamanian government began recognizing indigenous territories as comarca, which gave the groups the exclusive land rights and considerable autonomy [205]. Political structure in the comarca consists of elected local and regional leaders that form a general assembly (Congreso General) however national government still controls public expenditure and tax revenues in the comarca.

Of Panamá’s 7 indigenous groups, the Ngäbe Buglé make up nearly 2/3 of the indigenous population (2004 population estimate-128,978) but are also the most impoverished with 91% of the population living in extreme poverty (< US$ 1.75/day) [203]. The comarca Ngäbe Bugle, established in 1997 is located in western Panamá and covers 6,994 km2 on the Pacific and Caribbean slopes of the Central Cordillera. The comarca is further subdivided into 7 districts or corregimientos: Besiko, Kankintú, Kusapín, Mirono, Müna, Nole Duima and Ñürüm. The corregimientos of Besiko, Mirono, Müna, Nole Duima and Ñürüm are located on the Pacific slope where land cover is a combination of grasses and tropical forest cover and there are a dry (December to April) and wet season.

38 The Caribbean slope is predominantly tropical forest owing largely to the year round wet season.

Traditionally the Ngäbe Buglé lived in dispersed family groups and practiced subsistence agricultural that relied on long fallow cycles to restore the nutrients in the poor soil of the steep cordilleran slopes [206]. As populations increased and arable land became scarce, fallow periods shortened and productivity declined. At same time “national development” projects initiated by the Torrijos government in the 1960’s and resource extraction projects (oil pipeline, mine and hydroelectric) led to regional development projects that brought roads, schools and health facilities into the previously remote Ngäbe Buglé territory. The combination of these two factors led to a shift in livelihoods for the Ngäbe Buglé as they became more involved with the national political and economic realms. Specifically, wage labour became the most common means of securing cash resources, but required extended absences from home, a phenomenon that continues until today [206]. Contemporary livelihoods range from dispersed subsistence farming households with minor cash inputs from migratory labour to small store owners in more densely populated communities that are found close to the access roads.

Health outcomes in the comarca Ngäbe Buglé are still below national averages. Life expectancy in the comarca has increased slightly in the past decade from 67 in 2004 to 69 in 2009 [207], compared to a national average of 76 [208]. Despite a considerable increase since 2004, professional health care providers only present at 53% of births in the comarca and doctors are still in short supply (27/10,00 inhabitants) [207]. In children under 5 years of age, the primary causes of death are respiratory and skin infections, diarrhea, gastrointestinal parasitism and malnutrition [207]. The national average of chronic malnutrition was 18% in 2008 [209], however in the comarca Ngäbe Buglé the prevalence of stunting recorded in recent studies was 60% [210] and 61% [211]. Recent dietary assessment found the average child diet is based 39 mainly on complex carbohydrates (rice, beans, plantains and vegetables) and thus is low in animal protein and dietary diversity [210,211]. Not surprisingly, dietary intake of energy and fat as well as, vitamin A, B12, thiamine and folate were well below recommendations set by the Nutritional Institute of Central America and Panamá (INCAP) [210,211]. Furthermore, GI nematodes were detected in greater than 80% of preschool children [210]. Thus preschool children in this region consume a diet low in diversity, fat and animal sources and chronic malnutrition as well as GI nematode infection are of concern.

In 2005, the Red de Oportunidades began in the comarca to address the high levels of poverty and malnutrition and is currently delivered to the over 90% of families in extreme poverty. There are two variants of the Red de Oportunidades programs, the cash transfer (CT) program provides an additional US$ 50/month and is present in 4 of the 7 corregimientos in the comarca. The food voucher (FV) program, in the other 3 corregimientos, provides $50/month in food vouchers to be reimbursed at local stores. Transfers are provided to a designated woman from each household and in exchange, a member of the household must participate in health and skills training, women must be up-to- date on their health checks (pregnancy and reproductive health), and children must attend school and be current on their vaccinations. Training workshops in the CT program focus primarily on preventive health and financial management whereas training in the FV region focuses on agricultural themes. Despite recent changes in government, the Red de Oportunidades is still in operation, reaching 14,766 households in the comarca Ngäbe Buglé in 2010 [212].

The interrelationship between gastrointestinal parasite infection and malnutrition lends itself well to a transdisciplinary approach due to the complex set of variables that act at the individual to societal levels to modulate this relationship. The Ecohealth approach was identified as a particularly appropriate research framework due to its emphasis on mixed methodologies and in particular the inclusion of participatory methodologies as a way of capturing 40 multiple perspectives on health issues. Specifically, this longitudinal research project incorporated quantitative and qualitative methods in exploratory and explanatory sequential research designs to gain a rich understanding of the quantitative data as well as to generalize qualitative findings. Thus, this doctoral thesis research will implement an Ecohealth approach that relies on the novel methodological integration of discipline specific tools to conduct a comprehensive transdisciplinary analysis of preschool child health in the comarca Ngäbe Buglé of western Panamá.

This conceptual framework will first be used to determine whether GI parasite infection increases preschool child stunting and whether stunting in turn increases susceptibility to GI parasite reinfection. Second, individual, household and regional level factors that modulate this relationship will be identified; and finally, the interactions between the biophysical, social and spatial factors that influence preschool child nutrition and GI parasite reinfection will be examined.

2.7 References 1. Black R, E., Allen L, H., Bhutta Z, A., Caulfield L, E. , de Onis M, et al. (2008) Maternal and child undernutrition: global and regional exposures and health consequences. Lancet 371: 243-260.

2. Golden MHN (1996) Specific deficiencies versus growth failure: Type I and Type II. J Nutr Environ Med 6: 301.

3. Dewey KG, Begum K (2011) Long-term consequences of stunting in early life. Matern Child Nutr 7: 5-18.

4. Allen LH (1994) Nutritional Influences on Linear Growth - a General-Review. Eur J Clin Nutr 48: S75-S89.

5. Ramakrishnan U, Aburto N, McCabe G, Martorell R (2004) Multimicronutrient interventions but not vitamin A or iron interventions alone improve child growth: Results of 3 meta-analyses. J Nutr 134: 2592-2602.

6. Ramakrishnan U, Nguyen P, Martorell R (2009) Effects of micronutrients on growth of children under 5 y of age: Meta-analyses of single and multiple nutrient interventions. Am J Clin Nutr 89: 191-203.

7. Brown KH, Peerson JM, Rivera J, Allen LH (2002) Effect of supplemental zinc on the growth and serum zinc concentrations of prepubertal children: A meta-analysis of randomized controlled trials1-3. Am J Clin Nutr 75: 1062- 1071.

8. Imdad A, Bhutta ZA (2011) Effect of preventive zinc supplementation on linear growth in children under 5 years of age in developing countries: A meta- analysis of studies for input to the lives saved tool. BMC Public Health 11.

9. Arimond M, Ruel MT (2004) Dietary diversity is associated with child nutritional status: Evidence from 11 Demographic and Health Surveys. J Nutr 134: 2579-2585.

10. Rah JH, Akhter N, Semba RD, Pee SD, Bloem MW, et al. (2010) Low dietary diversity is a predictor of child stunting in rural Bangladesh. Eur J Clin Nutr 64: 1393-1398.

11. Grillenberger M, Neumann CG, Murphy SP, Bwibo NO, Weiss RE, et al. (2006) Intake of micronutrients high in animal-source foods is associated with better growth in rural Kenyan school children. Br J Nutr 95: 379-390.

42 12. Torheim LE, Ouattara F, Diarra MM, Thiam FD, Barikmo I, et al. (2004) Nutrient adequacy and dietary diversity in rural Mali: association and determinants. Eur J Clin Nutr 58: 594-604.

13. Lunn PG (2002) Growth retardation and stunting of children in developing countries - Invited commentary. Br J Nutr 88: 109-110.

14. Dewey KG, Mayers DR (2011) Early child growth: How do nutrition and infection interact? Matern Child Nutr 7: 129-142.

15. Koski KG, Scott ME (2001) Gastrointestinal nematodes, nutrition and immunity: Breaking the negative spiral. Annu Rev Nutr 21: 297-321.

16. Lunn PG, Northropclewes CA, Downes RM (1991) Intestinal Permeability, Mucosal Injury, and Growth Faltering in Gambian Infants. Lancet 338: 907-910.

17. Stephensen CB (1999) Burden of infection on growth failure. J Nutr 129: 534S-538S.

18. Oberhelman RA, Guerrero ES, Fernandez ML, Silio M, Mercado D, et al. (1998) Correlations between intestinal parasitosis, physical growth, and psychomotor development among infants and children from rural Nicaragua. Am J Trop Med Hyg 58: 470-475.

19. Quihui-Cota L, Valencia ME, Crompton DWT, Phillips S, Hagan P, et al. (2004) Prevalence and intensity of intestinal parasitic infections in relation to nutritional status in Mexican schoolchildren. Trans R Soc Trop Med Hyg 98: 653-659.

20. Saldiva SR, Silveira AS, Philippi ST, Torres DM, Mangini AC, et al. (1999) Ascaris-Trichuris association and malnutrition in Brazilian children. Paediatr Perinat Epidemiol 13: 89-98.

21. Sackey ME, Weigel MM, Armijos RX (2003) Predictors and nutritional consequences of intestinal parasitic infections in rural Ecuadorian children. J Trop Pediatr 49: 17-23.

22. Simsek Z, Zeyrek FY, Kurcer MA (2004) Effect of Giardia infection on growth and psychomotor development of children aged 0-5 years. J Trop Pediatr 50: 90-93.

23. Stephenson LS, Latham MC, Adams EJ, Kinoti SN, Pertet A (1993) Physical- Fitness, Growth and Appetite of Kenyan School Boys with Hookworm, Trichuris-Trichiura and Ascaris-Lumbricoides Infections Are Improved 4 Months after a Single-Dose of Albendazole. J Nutr 123: 1036-1046.

43 24. Stephenson LS, Crompton DWT, Latham MC, Schulpen TWJ, Nesheim MC, et al. (1980) Relationships between Ascaris infection and growth of malnourished preschool children in Kenya. Am J Clin Nutr 33: 1165-1172.

25. Adams EJ, Stephenson LS, Latham MC, Kinoti SN (1994) Physical-Activity and Growth of Kenyan School-Children with Hookworm, Trichuris-Trichiura and Ascaris-Lumbricoides Infections Are Improved after Treatment with Albendazole. J Nutr 124: 1199-1206.

26. Checkley W, Buckley G, Gilman RH, Assis AM, Guerrant RL, et al. (2008) Multi- country analysis of the effects of diarrhoea on childhood stunting. Int J Epidemiol 37: 816-830.

27. Humphrey JH (2009) Child undernutrition, tropical enteropathy, toilets, and handwashing. The Lancet 374: 1032-1035.

28. Goto R, Mascie-Taylor CGN, Lunn PG (2009) Impact of intestinal permeability, inflammation status and parasitic infections on infant growth faltering in rural Bangladesh. Br J Nutr 101: 1509-1516.

29. McKay S, Gaudier E, Campbell DI, Prentice AM, Albers R (2010) Environmental enteropathy: New targets for nutritional interventions. International Health 2: 172-180.

30. Adekanmbi VT, Kayode GA, Uthman OA (2011) Individual and contextual factors associated with childhood stunting in Nigeria: A multilevel analysis. Matern Child Nutr.

31. Larrea C, Freire W (2002) Social inequality and child malnutrition in four Andean countries. Rev Panam Salud Publica 11: 356-364.

32. Pongou R, Ezzati M, Salomon JA (2006) Household and community socioeconomic and environmental determinants of child nutritional status in Cameroon. BMC Public Health 6.

33. Agee MD (2010) Reducing child malnutrition in Nigeria: combined effects of income growth and provision of information about mothers' access to health care services. Soc Sci Med 71: 1973-1980.

34. Engebretsen IMS, Tylleskär T, Wamani H, Karamagi C, Tumwine JK (2008) Determinants of infant growth in Eastern Uganda: A community-based cross-sectional study. BMC Public Health 8.

35. Hong R, Mishra V (2006) Effect of wealth inequality on chronic undernutrition in Cambodian children. J Health Popul Nutr 24: 89-99.

44 36. Leonard WR, Dewalt KM, Stansbury JP, McCaston MK (2000) Influence of dietary quality on the growth of highland and coastal Ecuadorian children. Am J Hum Biol 12: 825-837.

37. Benefice E, Monroy SL, Jiménez S, López R (2006) Nutritional status of Amerindian children from the Beni River (lowland Bolivia) as related to environmental, maternal and dietary factors. Public Health Nutr 9: 327- 335.

38. Hatløy A, Hallund J, Diarra MM, Oshaug A (2000) Food variety, socioeconomic status and nutritional status in urban and rural areas in Koutiala (Mali). Public Health Nutr 3: 57-65.

39. Boeke CE, Mora-Plazas M, Forero Y, Villamor E (2010) Intestinal protozoan infections in relation to nutritional status and gastrointestinal morbidity in Colombian school children. J Trop Pediatr 56: 299-306.

40. Moffat T (2003) Diarrhea, respiratory infections, protozoan gastrointestinal parasites, and child growth in Kathmandu, Nepal. Am J Phys Anthropol 122: 85-97.

41. Casapia M, Joseph SA, Nunez C, Rahme E, Gyorkos TW (2006) Parasite risk factors for stunting in grade 5 students in a community of extreme poverty in Peru. Int J Parasitol 36: 741-747.

42. Ahmed A, Al-Mekhlafi H, Al-Adhroey A, Ithoi I, Abdulsalam A, et al. (2012) The nutritional impacts of soil-transmitted helminths infections among Orang Asli schoolchildren in rural Malaysia. Parasites & Vectors 5: 119.

43. Reyes H, Pérez-Cuevas R, Sandoval A, Castillo R, Santos JI, et al. (2004) The family as a determinant of stunting in children living in conditions of extreme poverty: A case-control study. BMC Public Health 4.

44. Grantham-McGregor S (2002) Linear growth retardation and cognition. Lancet 359: 542.

45. Grantham-McGregor S, Cheung YB, Cueto S, Glewwe P, Richter L, et al. (2007) Developmental potential in the first 5 years for children in developing countries. Lancet 369: 60-70.

46. Siegel EH, Stoltzfus RJ, Kariger PK, Katz J, Khatry SK, et al. (2005) Growth indices, anemia, and diet independently predict motor milestone acquisition of infants in South Central Nepal. J Nutr 135: 2840-2844.

47. Kuklina EV, Ramakrishnan U, Stein AD, Barnhart HH, Martorell R (2004) Growth and diet quality are associated with the attainment of walking in rural Guatemalan infants. J Nutr 134: 3296-3300.

45 48. Cheung YB, Yip PSF, Karlberg JPE (2001) Fetal growth, early postnatal growth and motor development in Pakistani infants. Int J Epidemiol 30: 66-72.

49. Olney DK, Pollitt E, Kariger PK, Khalfan SS, Ali NS, et al. (2007) Young Zanzibari children with iron deficiency, iron deficiency anemia, stunting, or malaria have lower motor activity scores and spend less time in locomotion. J Nutr 137: 2756-2762.

50. Kuklina EV, Ramakrishnan U, Stein AD, Barnhart HH, Martorell R (2006) Early childhood growth and development in rural Guatemala. Early Hum Dev 82: 425-433.

51. Kar BR, Rao SL, Chandramouli BA (2008) Cognitive development in children with chronic protein energy malnutrition. Behav Brain Funct 4: art. no. 31. doi: 10.1186/1744-9081-1184-1131.

52. Crookston BT, Dearden KA, Alder SC, Porucznik CA, Stanford JB, et al. (2011) Impact of early and concurrent stunting on cognition. Matern Child Nutr 7: 397-409.

53. Mendez MA, Adair LS (1999) Severity and timing of stunting in the first two years of life affect performance on cognitive tests in late childhood. J Nutr 129: 1555-1562.

54. Fentiman A, Hall A, Bundy D (2001) Health and cultural factors associated with enrolment in basic education: A study in rural Ghana. Soc Sci Med 52: 429-439.

55. Kramer MS (1987) Determinants of low birth weight: Methodological assessment and meta-analysis. Bull World Health Organ 65: 663-737.

56. Kramer MS, Olivier M, McLean FH, Willis DM, Usher RH (1990) Impact of intrauterine growth retardation and body proportionality on fetal and neonatal outcome. Pediatrics 86: 707-713.

57. Özaltin E, Hill K, Subramanian SV (2010) Association of maternal stature with offspring mortality, underweight, and stunting in low- to middle-income countries. JAMA - Journal of the American Medical Association 303: 1507- 1516.

58. Victora CG, Adair L, Fall C, Hallal PC, Martorell R, et al. (2008) Maternal and child undernutrition: consequences for adult health and human capital. The Lancet 371: 340-357.

59. Habicht JP, Martorell R (2010) Probability, plausibility, and adequacy evaluations of the oriente study demonstrate that supplementation improved child growth. J Nutr 140: 407-410.

46 60. Mara D, Lane J, Scott B, Trouba D (2010) Sanitation and health. PLoS Med 7: e1000363.

61. Scrimshaw NS, SanGiovanni JP (1997) Synergism of nutrition, infection, and immunity: An overview. Am J Clin Nutr 66.

62. WHO (2005) Deworming for Health and Development - Report of the third global meeting of the partners for parasite control. Geneva.

63. Bethony J, Brooker S, Albonico M, Geiger SM, Loukas A, et al. (2006) Soil- transmitted helminth infections: ascariasis, trichuriasis, and hookworm. Lancet 367: 1521-1532.

64. Udonsi JK, Atata G (1987) Necator americanus: Temperature, pH, light, and larval development, longevity, and desiccation tolerance. Exp Parasitol 63: 136-142.

65. Anderson RM (1986) The population dynamics and epidemiology of intestinal nematode infections. Trans R Soc Trop Med Hyg 80: 686-696.

66. Holland CV (2009) Predisposition to Ascariasis: Patterns, mechanisms and implications. Parasitology 136: 1537-1547.

67. Quinnell RJ, Pullan RL, Breitling LP, Geiger SM, Cundill B, et al. (2010) Genetic and household determinants of predisposition to human hookworm infectionin a Brazilian community. J Infect Dis 202: 954-961.

68. Williams-Blangero S, Subedi J, Upadhayay RP, Manral DB, Rai DR, et al. (1999) Genetic analysis of susceptibility to infection with Ascaris lumbricoides. Am J Trop Med Hyg 60: 921-926.

69. Williams-Blangero S, VandeBerg JL, Subedi J, Jha B, Corrêa-Oliveira R, et al. (2008) Localization of multiple quantitative trait loci influencing susceptibility to infection with Ascaris lumbricoides. J Infect Dis 197: 66- 71.

70. Hagel I, Lynch NR, Di Prisco MC, Ṕerez M, Sánchez JE, et al. (1999) Helminthic infection and anthropometric indicators in children from a tropical slum: Ascaris reinfection after anthelmintic treatment. J Trop Pediatr 45: 215- 220.

71. Saldiva SRM, Carvalho HB, Castilho VP, Struchiner CJ, Massad E (2002) Malnutrition and susceptibility to enteroparasites: Reinfection rates after mass chemotherapy. Paediatr Perinat Epidemiol 16: 166-171.

72. Hesham Al-Mekhlafi M, Surin J, Atiya AS, Ariffin WA, Mohammed Mahdy AK, et al. (2008) Pattern and predictors of soil-transmitted helminth

47 reinfection among aboriginal schoolchildren in rural Peninsular Malaysia. Acta Trop 107: 200-204.

73. Luoba AI, Geissler PW, Estambale B, Ouma JH, Alusala D, et al. (2005) Earth- eating and reinfection with intestinal helminths among pregnant and lactating women in western Kenya. Trop Med Int Health 10: 220-227.

74. Saathoff E, Olsen A, Kvalsvig JD, Geissler WP (2002) Geophagy and its association with geohelminth infection in rural schoolchildren from northern KwaZulu-Natal, South Africa. Trans R Soc Trop Med Hyg 96: 485- 490.

75. Bartram J, Cairncross S (2010) Hygiene, sanitation, and water: forgotten foundations of health. PLoS Med 7: e1000367.

76. Pullan RL, Bethony JM, Geiger SM, Cundill B, Correa-Oliveira R, et al. (2008) Human helminth co-infection: Analysis of spatial patterns and risk factors in a Brazilian community. PLoS Neglect Trop D 2: e352.

77. De Souza EA, Da Silva-Nunes M, Dos Santos Malafronte R, Muniz PT, Cardoso MA, et al. (2007) Prevalence and spatial distribution of intestinal parasitic infections in a rural Amazonian settlement, Acre State, Brazil. Cad Saude Publica 23: 427-434.

78. Chongsuvivatwong V, Pas-Ong S, McNeil D, Geater A, Duerawee M (1996) Predictors for the risk of hookworm infection: Experience from endemic villages in southern Thailand. Trans R Soc Trop Med Hyg 90: 630-633.

79. Traub RJ, Robertson ID, Irwin P, Mencke N, Thompson RCA (2004) The prevalence, intensities and risk factors associated with geohelminth infection in tea-growing communities of Assam, India. Trop Med Int Health 9: 688-701.

80. Holland CV, Taren DL, Crompton DWT, Nesheim MC, Sanjur D, et al. (1988) Intestinal helminthiases in relation to the socioeconomic environment of Panamanian children. Soc Sci Med 26: 209-213.

81. Stothard JR, Imison E, French MD, Sousa-Figueiredo JC, Khamis IS, et al. (2008) Soil-transmitted helminthiasis among mothers and their preschool children on Unguja Island, Zanzibar with emphasis upon ascariasis. Parasitology 135: 1447-1455.

82. Hotez P (2008) Hookworm and Poverty. Ann N Y Acad Sci 1136: 38-44.

83. Forrester JE, Scott ME, Bundy DAP, Golden MHN (1988) Clustering of Ascaris lumbricoides and Trichuris trichiura infections within households. Trans R Soc Trop Med Hyg 82: 282-288.

48 84. Brooker S, Kabatereine NB, Tukahebwa EM, Kazibwe F (2004) Spatial analysis of the distribution of intestinal nematode infections in Uganda. Epidemiol Infect 132: 1065-1071.

85. Koroma JB, Peterson J, Gbakima AA, Nylander FE, Sahr F, et al. (2010) Geographical distribution of intestinal schistosomiasis and soil- transmitted helminthiasis and preventive chemotherapy strategies in Sierra Leone. PLoS Neglect Trop D 4: e891.

86. Walker M, Hall A, Basáñez MG (2011) Individual predisposition, household clustering and risk factors for human infection with Ascaris lumbricoides: New epidemiological insights. PLoS Neglect Trop D 5: e1047

87. Pullan RL, Kabatereine NB, Quinnell RJ, Brooker S (2010) Spatial and genetic epidemiology of hookworm in a rural community in Uganda. PLoS Neglect Trop D 4: e713.

88. Raso G, Vounatsou P, Gosoniu L, Tanner M, Goran E, et al. (2006) Risk factors and spatial patterns of hookworm infection among schoolchildren in a rural area of western Cote d'Ivoire. Int J Parasitol 36: 201-210.

89. Saathoff E, Olsen A, Kvalsvig JD, Appleton CC, Sharp B, et al. (2005) Ecological covariates of Ascaris lumbricoides infection in schoolchildren from rural KwaZulu-Natal, South Africa. Trop Med Int Health 10: 412-422.

90. Gunawardena GSA, Karunaweera ND, Ismail MM (2004) Wet-days: Are they better indicators of Ascaris infection levels? J Helminthol 78: 305-310.

91. Nwosu ABC, Anya AO (1980) Seasonality in human hookworm infection in an endemic area of Nigeria, and its relationship to rainfall. Tropenmed Parasitol 31: 201-208.

92. Black RE, Cousens S, Johnson HL, Lawn JE, Rudan I, et al. (2010) Global, regional, and national causes of child mortality in 2008: a systematic analysis. The Lancet 375: 1969-1987.

93. Wagner EG, Lanoix JN (1958) Excreta disposal for rural areas and small communities. Geneva: World Health Organization.

94. Rinne S, Rodas EJ, Galer-Unti R, Glickman N, Glickman LT (2005) Prevalence and risk factors for protozoan and nematode infections among children in an Ecuadorian highland community. Trans R Soc Trop Med Hyg 99: 585- 592.

95. Herbst S, Fayzieva D, Kistemann T (2008) Risk factor analysis of diarrhoeal diseases in the Aral Sea area (Khorezm, Uzbekistan). Int J Environ Heal R 18: 305-321.

49 96. Genser B, Strina A, Teles CA, Prado MS, Barreto ML (2006) Risk factors for childhood diarrhea incidence: dynamic analysis of a longitudinal study. Epidemiology 17: 658-667.

97. Cifuentes E, Suarez L, Solano M, Santos R (2002) Diarrheal diseases in children from a water reclamation site in Mexico city. Environ Health Perspect 110: A619-A624.

98. Fewtrell L, Kaufmann RB, Kay D, Enanoria W, Haller L, et al. (2005) Water, sanitation, and hygiene interventions to reduce diarrhoea in less developed countries: a systematic review and meta-analysis. Lancet Infect Dis 5: 42-52.

99. Curtis V, Cairncross S (2003) Effect of washing hands with soap on diarrhoea risk in the community: a systematic review. Lancet Infect Dis 3: 275-281.

100. United Nations (2009) Rethinking Poverty: Report on the World Social Situation 2010. Geneva: United Nations.

101. Fiszbein A, Schady N (2009) Conditional cash transfers: reducing present and future poverty. Washington, DC: World Bank. 383 p.

102. Brenneman A, Kerf M (2002) Infrastructure and Poverty Linkages: A Literature Review. Washington, DC: The World Bank.

103. Khandker SR, Bakht Z, Koolwal GB (2009) The poverty impact of rural roads: evidence from Bangladesh. Econ Dev Cult Change 57: 685-722.

104. Jalan J, Ravallion M (2002) Geographic poverty traps? A micro model of consumption growth in rural China. Journal of Applied Econometrics 17: 329-346.

105. De Haan A (1999) Livelihoods and poverty: The role of migration - a critical review of the migration literature. J Dev Stud 36: 1-47.

106. Barbieri AF, Monte-Mor RL, Bilsborrow RE (2007) Town in the jungle: exploring linkages between rural-urban mobility, urbanization and development in the Amazon. Urban Population, Devleopment and Environment Dynamics in Developing Countries Nairobi, Kenya: PRIPODE Workshop. pp. 26.

107. Vissandjée B, Barlow R, Fraser DW (1997) Utilization of health services among rural women in Gujarat, India. Public Health 111: 135-148.

108. Ombok M, Adazu K, Odhiambo F, Bayoh N, Kiriinya R, et al. (2010) Geospatial distribution and determinants of child mortality in rural western Kenya 2002-2005. Trop Med Int Health 15: 423-433.

50 109. Airey T (1992) The impact of road construction on the spatial characteristics of hospital utilization in the Meru district of Kenya. Soc Sci Med 34: 1135- 1146.

110. Denno D (2011) Global child health. Pediatr Rev 32: e25-e38.

111. Hunter PR, MacDonald AM, Carter RC (2010) Water supply and health. PLoS Med 7: e1000361.

112. Milroy CA, Borja PC, Barros FR, Barreto ML (2001) Evaluating sanitary quality and classifying urban sectors according to environmental conditions. Environ Urban 13: 235-255.

113. Curtis V, Kanki B, Mertens T, Traore E, Diallo I, et al. (1995) Potties, pits and pipes: Explaining hygiene behaviour in Burkina Faso. Soc Sci Med 41: 383- 393.

114. Huda TMN, Unicomb L, Johnston RB, Halder AK, Yushuf Sharker MA, et al. (2012) Interim evaluation of a large scale sanitation, hygiene and water improvement programme on childhood diarrhea and respiratory disease in rural Bangladesh. Soc Sci Med.

115. Jenkins MW, Curtis V (2005) Achieving the 'good life': why some people want latrines in rural Benin. Soc Sci Med 61: 2446-2459.

116. Jenkins MW, Cairncross S (2010) Modelling latrine diffusion in Benin: towards a community typology of demand for improved sanitation in developing countries. J Water Health 8: 166-183.

117. Vittor AY, Gilman RH, Tielsch J, Glass G, Shields T, et al. (2006) The effect of deforestation on the human-biting rate of Anopheles darlingi, the primary vector of falciparum malaria in the Peruvian Amazon. Am J Trop Med Hyg 74: 3-11.

118. Barcellos C, Feitosa P, Damacena GN, Andreazzi MA (2010) Highways and outposts: Economic development and health threats in the central Brazilian Amazon region. International Journal of Health Geographics 9.

119. Hacker MA, Leite I, Friedman SR, Carrijo RG, Bastos FI (2009) Poverty, bridging between injecting drug users and the general population, and "interiorization" may explain the spread of HIV in southern Brazil. Health Place 15: 514-519.

120. Carswell JW, Lloyd G, Howells J (1989) Prevalence of HIV-1 in east African lorry drivers. AIDS 3: 759-761.

121. Luque Fernández MA, Mason PR, Gray H, Bauernfeind A, Fesselet JF, et al. (2010) Descriptive spatial analysis of the cholera epidemic 2008-2009 in

51 Harare, Zimbabwe: a secondary data analysis. Trans R Soc Trop Med Hyg 105: 38-45.

122. Ferrari MJ, Djibo A, Grais RF, Bharti N, Grenfell BT, et al. (2010) Rural-urban gradient in seasonal forcing of measles transmission in Niger. Proc Biol Sci 277: 2775-2782.

123. Eisenberg JNS, Cevallos W, Ponce K, Levy K, Bates SJ, et al. (2006) Environmental change and infectious disease: How new roads affect the transmission of diarrheal pathogens in rural Ecuador. Proc Natl Acad Sci U S A 103: 19460-19465.

124. Bates SJ, Trostle J, Cevallos WT, Hubbard A, Eisenberg JNS (2007) Relating diarrheal disease to social networks and the geographic configuration of communities in rural Ecuador. Am J Epidemiol 166: 1088-1095.

125. Aguirre-Arenas J, Escobar-Pérez M, Chávez-Villasana A (1998) Evaluation of the food consumption and nutrition in 4 rural communities. Salud Publica Mex 40: 398-407.

126. Romaguera D, Samman N, Rossi A, Miranda C, Pons A, et al. (2008) Dietary patterns of the Andean population of Puna and Quebrada of Humahuaca, Jujuy, Argentina. Br J Nutr 99: 390-397.

127. Drewnowski A, Popkin BM (1997) The Nutrition Transition: New Trends in the Global Diet. Nutr Rev 55: 31-43.

128. Bermudez OI, Tucker KL (2003) Trends in dietary patterns of Latin American populations. Cad Saude Publica 19: S87-S99.

129. Rosenfield PL (1992) The potential of transdisciplinary research for sustaining and extending linkages between the health and social sciences. Soc Sci Med 35: 1343-1357.

130. Annerstedt M (2010) Transdisciplinarity as an inference technique to achieve a better understanding in the health and environmental sciences. International Journal of Environmental Research and Public Health 7: 2692-2707.

131. Dankwa-Mullan I, Rhee KB, Stoff DM, Pohlhaus JR, Sy FS, et al. (2010) Moving toward paradigm-shifting research in health disparities through translational, transformational, and transdisciplinary approaches. Am J Public Health 100: S19-S24.

132. Leung Z, Middleton D, Morrison K (2012) One Health and EcoHealth in Ontario: A qualitative study exploring how holistic and integrative

52 approaches are shaping public health practice in Ontario. BMC public health 12.

133. Kessel F, Rosenfield PL (2008) Toward Transdisciplinary Research. Historical and Contemporary Perspectives. Am J Prev Med 35: S225-S234.

134. McLaren L, Hawe P (2005) Ecological perspectives in health research. J Epidemiol Community Health 59: 6-14.

135. Kaplan GA, Lynch JW (1997) Whither studies on the socioeconomic foundations of population health? Am J Public Health 87: 1409-1411.

136. Kaplan GA (1996) People and places: Contrasting perspectives on the association between social class and health. Int J Health Serv 26: 507-519.

137. Carstairs V, Morris R (1989) Deprivation and mortality: An alternative to social class? Community Med 11: 210-219.

138. Duncan C, Jones K, Moon G (1998) Context, composition and heterogeneity: Using multilevel models in health research. Soc Sci Med 46: 97-117.

139. Diez Roux AV (2001) Investigating neighborhood and area effects on health. Am J Public Health 91: 1783-1789.

140. Macintyre S, Ellaway A, Cummins S (2002) Place effects on health: How can we conceptualise, operationalise and measure them? Soc Sci Med 55: 125-139.

141. Cummins S, Curtis S, Diez-Roux AV, Macintyre S (2007) Understanding and representing 'place' in health research: A relational approach. Soc Sci Med 65: 1825-1838.

142. Nielsen NO (2001) Ecosystem approaches to human health. Cad Saúde Pública,: 69-75.

143. Kay JJ, Regier HA, Boyle M, Francis G (1999) An ecosystem approach for sustainability: Addressing the challenge of complexity. Futures 31: 721- 742.

144. Holling CS (2001) Understanding the complexity of economic, ecological, and social systems. Ecosystems 4: 390-405.

145. Kay J, Schneider E (1994) Embracing complexity, the challenge of the ecosystem approach. Alternatives 20: 32-39.

146. Waltner-Toews D (2004) Ecosystem sustainability and health: A practical approach. Cambridge: Cambridge University Press. 138 p.

53 147. Funtowicz S, Ravtez J (2008) Beyond Complex Systems: Emergent Complexity and Social Solidarity. In: Waltner-Toews D, Kay J, Lister NM, editors. The Ecosystem Approach: Complexity, Uncertainty, and Managing for Sustainability. New York: Columbia University Press. pp. 309-322.

148. Forget G, Lebel J (2001) An ecosystem approach to human health. Int J Occup Environ Health 7: S3-S36.

149. Creswell JW (2007) Qualitative inquiry and research design: Choosing among five approaches, second edition. Thousand Oaks, CA: Sage Publications Ltd.

150. Bryman A (2006) Integrating quantitative and qualitative research: How is it done? Qualitative Research 6: 97-113.

151. Johnson RB, Onwuegbuzie AJ, Turner LA (2007) Toward a Definition of Mixed Methods Research. Journal of Mixed Methods Research 1: 112- 133.

152. Creswell JW (2008) Research Design: Qualitative, quantitative and mixed methods approaches. Nerbraska: Sage Publications.

153. Castro FG, Kellison JG, Boyd SJ, Kopak A (2010) A methodology for conducting integrative mixed methods research and data analyses. Journal of Mixed Methods Research 4: 342-360.

154. O'Cathain A, Murphy E, Nicholl J (2007) Why, and how, mixed methods research is undertaken in health services research in England: A mixed methods study. BMC Health Services Research 7.

155. Shaw JA, Connelly DM, Zecevic AA (2010) Pragmatism in practice: Mixed methods research for physiotherapy. Physiother Theory Pract 26: 510- 518.

156. Fagg J, Curtis S, Stansfeld S, Congdon P (2006) Psychological distress among adolescents, and its relationship to individual, family and area characteristics in East London. Soc Sci Med 63: 636-648.

157. Cannuscio CC, Weiss EE, Fruchtman H, Schroeder J, Weiner J, et al. (2009) Visual epidemiology: Photographs as tools for probing street-level etiologies. Soc Sci Med 69: 553-564.

158. Waltner-Toews D (2001) An ecosystem approach to health and its applications to tropical and emerging diseases. Cad Saúde Pública,: 7-36.

54 159. Neudoerffer RC, Waltner-Toews D, Kay JJ, Joshi DD, Tamang MS (2005) A diagrammatic approach to understanding complex eco-social interactions in Kathmandu, Nepal. Ecol Soc 10.

160. Carozza SE, Li B, Elgethun K, Whitworth R (2008) Risk of childhood cancers associated with residence in agriculturally intense areas in the United States. Environ Health Perspect 116: 559-565.

161. Gunier RB, Ward MH, Airola M, Bell EM, Colt J, et al. (2011) Determinants of agricultural pesticide concentrations in carpet dust. Environ Health Perspect 119: 970-976.

162. Moore LV, Diez Roux AV, Nettleton JA, Jacobs DR (2008) Associations of the local food environment with diet quality - A comparison of assessments based on surveys and geographic information systems. Am J Epidemiol 167: 917-924.

163. Powell LM, Auld MC, Chaloupka FJ, O'Malley PM, Johnston LD (2007) Associations Between Access to Food Stores and Adolescent Body Mass Index. Am J Prev Med 33: S301-S307.

164. Bodor JN, Rose D, Farley TA, Swalm C, Scott SK (2008) Neighbourhood fruit and vegetable availability and consumption: The role of small food stores in an urban environment. Public Health Nutr 11: 413-420.

165. Kim AA, Martinez AN, Klausner JD, Goldenson J, Kent C, et al. (2009) Use of sentinel surveillance and geographic information systems to monitor trends in HIV prevalence, incidence, and related risk behavior among women undergoing syphilis screening in a jail setting. J Urban Health 86: 79-92.

166. Rainey JJ, Omenah D, Sumba PO, Moormann AM, Rochford R, et al. (2007) Spatial clustering of endemic Burkitt's lymphoma in high-risk regions of Kenya. Int J Cancer 120: 121-127.

167. Kuldorff M (2011) SaTScan v9.0: Software for the spatial and space-time scan statistics. Information Management Services, Inc. http://www.satscan.org p.

168. Robertson C, Nelson TA (2010) Review of software for space-time disease surveillance. International Journal of Health Geographics 9.

169. Sarkar R, Prabhakar AT, Manickam S, Selvapandian D, Raghava MV, et al. (2007) Epidemiological investigation of an outbreak of acute diarrhoeal disease using geographic information systems. Trans R Soc Trop Med Hyg 101: 587-593.

55 170. Gaudart J, Poudiougou B, Dicko A, Ranque S, Toure O, et al. (2006) Space- time clustering of childhood malaria at the household level: A dynamic cohort in a Mali village. BMC Public Health 6.

171. Odoi A, Martin SW, Michel P, Middleton D, Holt J, et al. (2004) Investigation of clusters of giardiasis using GIS and a spatial scan statistic. International Journal of Health Geographics 3.

172. Falkingham J, Namazie C (2002) Measuring Health and Poverty: A Review of Approaches to Identifying the Poor; Centre HSR, editor. London: United Kingdom Department for International Development. 71 p.

173. Gibson RS (2005) Principles of Nutritional Assessment. New York, New York: Oxford University Press. 908 p.

174. Filmer D, Pritchett LH (2001) Estimating wealth effects without expenditure data - or tears: an application to educational enrollments in states of India. Demography 38: 115-132.

175. Gwatkin D, Wagstaff A, Yazbeck A (2005) Reaching the poor with health nutrition and population services: What works, what doesn't and why. Washington, DC: World Bank. 354 p.

176. Hu FB (2002) Dietary pattern analysis: A new direction in nutritional epidemiology. Curr Opin Lipidol 13: 3-9.

177. Newby PK, Tucker KL (2004) Empirically derived eating patterns using factor or cluster analysis: A review. Nutr Rev 62: 177-203.

178. Rutstein S, Johnson K (2004) The DHS Wealth Index. Calverton, Maryland: ORC Macro, MEASURE DHS+.

179. Vyas S, Kumaranayake L (2006) Constructing socio-economic status indices: how to use principal components analysis. Health Policy Plan 21: 459-468.

180. Van De Poel E, Hosseinpoor AR, Speybroeck N, Van Ourti T, Vega J (2008) Socioeconomic inequality in malnutrition in developing countries. Bull World Health Organ 86: 282-291.

181. Raso G, Utzinger J, Silué KD, Ouattara M, Yapi A, et al. (2005) Disparities in parasitic infections, perceived ill health and access to health care among poorer and less poor schoolchildren of rural Côte d'Ivoire. Trop Med Int Health 10: 42-57.

182. Nundy S, Gilman RH, Xiao L, Cabrera L, Cama R, et al. (2011) Wealth and its associations with enteric parasitic infections in a low-income community in Peru: use of principal component analysis. Am J Trop Med Hyg 84: 38- 42.

56 183. Thompson FE, Subar AF (2001) Dietary assessment methodology. Nutrition in the Prevention and Treatment of Disease 2.

184. Burrows TL, Martin RJ, Collins CE (2010) A Systematic Review of the Validity of Dietary Assessment Methods in Children when Compared with the Method of Doubly Labeled Water. J Am Diet Assoc 110: 1501-1510.

185. McPherson RS, Hoelscher DM, Alexander M, Scanlon KS, Serdula MK (2000) Dietary assessment methods among school-aged children: Validity and reliability. Prev Med 31: S11-S33.

186. Nyambose J, Koski KG, Tucker KL (2002) High intra/interindividual variance ratios for energy and nutrient intakes of pregnant women in rural Malawi show that many days are required to estimate usual intake. J Nutr 132: 1313-1318.

187. Persson V, Winkvist A, Ninuk T, Hartini S, Greiner T, et al. (2001) Variability in nutrient intakes among pregnant women in Indonesia: Implications for the design of epidemiological studies using the 24-h recall method. J Nutr 131: 325-330.

188. Moeller SM, Reedy J, Millen A, Dixon LB, Newby PK, et al. (2007) Dietary Patterns: Challenges and Opportunities in Dietary Patterns Research. J Am Diet Assoc 107: 6.

189. Hu FB, Rimm E, Smith-Warner SA, Feskanich D, Stampfer MJ, et al. (1999) Reproducibility and validity of dietary patterns assessed with a food- frequency questionnaire. Am J Clin Nutr 69: 243-249.

190. Khani BR, Ye W, Terry P, Wolk A (2004) Reproducibility and validity of major dietary patterns among Swedish women assessed with a food-frequency questionnaire. J Nutr 134: 1541-1545.

191. Kant AK, Graubard BI (2005) A comparison of three dietary pattern indexes for predicting biomarkers of diet and disease. J Am Coll Nutr 24: 294-303.

192. Smithers LG, Golley RK, Brazionis L, Lynch JW (2011) Characterizing whole diets of young children from developed countries and the association between diet and health: A systematic review. Nutr Rev 69: 449-467.

193. Baird J, Poole J, Robinson S, Marriott L, Godfrey K, et al. (2008) Milk feeding and dietary patterns predict weight and fat gains in infancy. Paediatr Perinat Epidemiol 22: 575-586.

194. Shin KO, Oh SY, Park HS (2007) Empirically derived major dietary patterns and their associations with overweight in Korean preschool children. Br J Nutr 98: 416-421.

57 195. Robinson S, Marriott L, Poole J, Crozier S, Borland S, et al. (2007) Dietary patterns in infancy: The importance of maternal and family influences on feeding practice. Br J Nutr 98: 1029-1037.

196. Wiles NJ, Northstone K, Emmett P, Lewis G (2009) 'Junk food' diet and childhood behavioural problems: Results from the ALSPAC cohort. Eur J Clin Nutr 63: 491-498.

197. Friedman LS, Lukyanova EM, Serdiuk A, et al (2009) Social-environmental factors associated with elevated body mass indexin Ukranian cohort of children. Int J Pediatr Obes 4: 81-90.

198. McBurney RPH, Griffin C, Paul AA, Greenberg DC (2004) The nutritional composition of African wild food plants: From compilation to utilization. J Food Comp Anal 17: 277-289.

199. Barikmo I, Ouattara F, Oshaug A (2007) Differences in micronutrients content found in cereals from various parts of Mali. J Food Comp Anal 20: 681-687.

200. Englberger L, Schierle J, Aalbersberg W, Hofmann P, Humphries J, et al. (2006) Carotenoid and vitamin content of Karat and other Micronesian banana cultivars. Int J Food Sci Nutr 57: 399-418.

201. Ruel MT (2003) Operationalizing dietary diversity: A review of measurement issues and research priorities. J Nutr 133: 3911S-3926S.

202. Sawadogo PS, Martin-Prével Y, Savy M, Kameli Y, Traissac P, et al. (2006) An infant and child feeding index is associated with the nutritional status of 6- to 23-month-old children in rural Burkina Faso. J Nutr 136: 656-663.

203. Ministerio de Economía y Finanzas (2010) La Distribución del Ingreso en Los Hogares de Panamá:Encuesta de Niveles de Vida 2008. Panamá: Ministerio de Economía y Finanzas. 34 p.

204. World Bank (1999) Panama Poverty Assessment: Priorities and Strategies for Poverty Reduction. Washington, DC: Human Development Department, World Bank.

205. World Bank (2007) Water Supply and Sanitation in Low Income Communities, Panamá. Washington, DC: World Bank. 111 p.

206. Bort JR, Young PD (1985) Economic and political adaptations to national development among the Guaymi. Anthropol Quart 58: 12.

207. MINSA (2010) Principales indicadores demograficos y de salud en la republica de Panamá. Años: 2000-2010. Panamá: MINSA. 42 p.

58 208. World Bank (2010) Country partnership strategy for the republic of Panamá. Washington, DC: World Bank. 97 p.

209. UNICEF (2008) The State of the World's Children 2009: Maternal and Newborn Health. New York: UNICEF. 168 p.

210. Payne L, Koski KG, Ortega-Barria E, Scott ME (2007) Benefit of vitamin A supplementation on Ascaris re-infection is less evident in stunted children. J Nutr 137: 1455-1459.

211. Ortega A, Fontes F, Sinisterra O, Valdés V (2006) Evaluación nutricional en pre-escolares y escolares de los distritos de Mironó, comarca Ngobe- Buglé y Santa Fe, Veraguas. Panama City, Panama: Secretaría del Plan Alimentario Nacional (SENAPAN). 46 p.

212. MIDES (2010) Informe de cobertura y avances de Red de Oportunidades: 2010. Panamá: MIDES. 16 p.

59 CHAPTER 3

Household density and asset-based indices predict child health in impoverished indigenous villages in rural Panamá

Carli M. Halpenny1, Kristine G. Koski2, Victoria E. Valdés3 and Marilyn E. Scott1

1Institute of Parasitology and McGill School of Environment and 2School of Dietetics and Human Nutrition, Macdonald Campus of McGill University, Ste- Anne de Bellevue, Quebec, Canada; 3Escuela de Nutrición y Dietética, Facultad de Medicina, Universidad de Panamá, Panamá.

Published in American Journal of Tropical Medicine and Hygiene:

Halpenny CM, Koski KG, Valdés VE, Scott ME (2012) Prediction of child health by household density and asset-based indices in impoverished indigenous villages in Rural Panamá. Am J Trop Med Hyg 86: 280-291.

3.1 Abstract Chronic infection over a 16 month period and stunting of preschool children were compared between more spatially “dense” versus “dispersed” households in rural Panamá. Chronic protozoan infection was associated with higher household density, lower household wealth index, poor household water quality, yard defecation and the practice of not washing hands with soap before eating. Models for chronic diarrhea confirmed the importance of household wealth and water quality as well as sanitation and hygiene practices. Furthermore, chronic protozoan infection was an important predictor for low height-for-age (HAZ), along with low household wealth index scores (HWI) but not household density. Thus, despite better access to health related infrastructure in the more densely populated households, chronic protozoan infection was more common, and was associated with higher rates of child stunting, compared with more dispersed households. However, within regions of higher population density, those households with more wealth are at an advantage as evidenced by better child growth.

3.2 Introduction Improved child health in combination with greater financial capital is considered integral in breaking the intergenerational cycle of poverty [1,2]. As such, development of physical infrastructure is central to poverty alleviation. Not only does it increase access to economic opportunities through road construction and access to financial markets and employment opportunities [3-5] but physical infrastructure also improves health through the addition of sanitation and water systems as well as providing better access to schools and health facilities [6]. Access to health facilities reduces child morbidity and mortality [7,8] particularly when education initiatives enhance the utilization of the available services [9]. Moreover, access to sanitation and water infrastructure, considered “forgotten foundations in health” [10], reduces child morbidity and mortality by encouraging the formation of well-being behaviours [11-13]. Conversely, infrastructure development is usually coupled with increased population density and spatial proximity of households which can increase transmission of infectious diseases [14,15]. For enteric pathogens in particular, factors such as crowding within the home [16,17] and higher household density have been associated with increased risk of infection [18]. This has been shown to be of particular importance in urban areas where sanitation infrastructure is insufficient [19,20] but household spatial dispersion is of likely relevance for studying child health in rural areas as well. Child growth in particular, is indicative of the long term impact of a child’s environment on their health where growth failure is associated with factors such as prolonged dietary insufficiency or frequent infection [21,22]. Thus, examining frequency of infection and child height for age can contribute to an understanding of the cumulative impact of physical assets (including infrastructure) and spatial proximity of households on child health.

Furthermore, as economic opportunities increase in extremely impoverished rural areas, accurately measuring household wealth is important to fully characterize the risk factors associated with child health outcomes. Measuring wealth in areas of extreme poverty is considered a challenge, primarily due to the fact that household income is often inconsistent, making reported income or consumption estimations of wealth inaccurate [23]. Statistically weighted asset- based indices of wealth [24,25] are considered a proxy for long term income to determine relative poverty [23] and may have stronger explanatory power than expenditure data for rural health outcomes [25]. Our investigation in western Panamá provided an opportunity to compare access to physical infrastructure and child health outcomes between more spatially concentrated and spatially dispersed rural areas and furthermore to examine the influence of household wealth on child stunting and infection outcomes in a region of extreme poverty. In the comarca Ngäbe Buglé, contemporary livelihoods range from subsistence farming with minor cash inputs in spatially dispersed households to short term wage labour from small stores or temporary employment in more spatially concentrated households. Importantly, recent studies in the comarca reported that over half of Ngäbe children are stunted [26,27]. Thus, our study aimed to: 1) compare intestinal protozoan infection, diarrhea and stunting in preschool children between spatially dispersed and more spatially concentrated households; and 2) determine whether household density and degree of poverty were associated with severity of stunting and chronicity of protozoan infections and diarrhea. Specifically, we hypothesized that households in the more densely populated regions would have higher average wealth, better access to sanitation, hygiene and health care, and therefore that children living in these households would have less severe stunting, and less chronic gastrointestinal protozoan infections and diarrhea.

3.3 Materials & Methods 3.3.1 Study population. The comarca Ngäbe-Buglé is a semi-autonomous political region inhabited primarily by the Ngäbe and Buglé indigenous groups (2004 population estimate of 128,978) of whom 91% live in extreme poverty (< US$ 1.75/day) [28]. In 2005, the Conditional Transfer (CT) program Red de Oportunidades began in the comarca, providing an additional US$ 50/month in either cash or food vouchers to the over 90% of families in extreme poverty in exchange for participation in health and education programs. The data presented are a subset of a larger study that was conducted in 2008 and 2009 in 2 corregimientos or political regions (Soloy - cash transfer, Emplanada de Chorcha - food voucher) within the district of Besiko within the CT project area (Figure 1). Using a random number generator, six villages were randomly selected within each corregimiento from those that were within 2 hrs walking distance of the health centre and had between 25-100 households registered in a recent government survey. All households in selected villages were approached and those that met the inclusion criteria of having at least one child 4 years or younger and living in extreme poverty, defined as having participated in a CT program since 2005, were given an orientation to the study in Spanish and Ngäbere (the local language). All but 3 eligible households agreed to participate resulting in a total of 262 households with 373 preschool children (161 households with 1 eligible child, 91 with 2, and 10 households with 3 eligible children). Approximately 50 – 60% of the households from 11 villages and 37% of households from 1 village (Cerro Viejo) participated in the study for a range of 7 - 41 households/village. The lower proportion of participating households in Cerro Viejo was due to a lower proportion of households with children <5 years of age.

3.3.2 Ethical considerations. Ethical approval was obtained from the Instituto Conmemorativo de Gorgas in Panamá and McGill University in Canada. Community interactions were established in accordance with the Guía para Realizar Estudios e Investigaciones en los Pueblos Indígenas de Panamá, which included participation in introductory and results workshops in each village. Written informed consent was obtained from primary caregivers during a household visit that included an explanation of study significance, of participant requirements and rights as well as an opportunity to ask questions in Spanish and Ngäbere. According to Panamanian Ministry of Health (MINSA) protocol, primary caregivers received verbal explanation of results after each sample as well as a MINSA laboratory diagnostic form allowing them to seek treatment at a health centre. In addition, at the end of each year, names of infected children were given to MINSA personnel at the health centres.

3.3.3 Study Procedure. The results presented here are part of a larger collaborative investigation of child health conducted by MINSA, the University of Panamá and McGill University. For the purpose of this paper, long-term estimates of chronicity of protozoan infection and diarrhea were obtained using data from 7 fecal samples collected over 16 months (Table 1). In addition, child anthropometry, household demographic data, spatial survey data, and child and household sanitation and hygiene practices were recorded during visits throughout the study period (Table 1). Data collection was conducted by MINSA personnel with previous experience in the area together with local translators who had been oriented to the specific project goals and methods prior to the initiation of the study, during which previously untested questionnaires were piloted with local communities. Quality of data collection was verified by rotating field supervisor visits as well as

questionnaire revision after each day of field work. Laboratory procedures were conducted during each week of field work.

3.3.4 Spatial survey. Geographical longitude and latitude points were recorded during visits 7 – 10 at households (n = 250), health facilities (n = 2) and schools (n = 10); access roads were recorded as tracks with a handheld Geographic Positioning System (GPS) (Garmin eTrex Vista HCX, Olathe KS). Geographic coordinates were converted to metric coordinates using the Universal Transverse Mercator (UTM) 17N projection and analyzed in ArcGIS Map 9.3.1. Household geographic coordinates were used to characterize the spatial dispersion of participant homes and distances from roads, health facilities and schools. Distances were calculated using Proximity Analysis and household density was calculated using a point density estimate for a circular neighbourhood with a 250 m radius. Resulting density estimates for each household describe the number of other study participant homes found within this circular area around their home, expressed in homes per square kilometer. As household density is one of the primary comparisons in the current study, analysis is restricted to households for which spatial data was available (250 households and 356 children; 153 households with 1 eligible child, 88 with 2, and 9 households with 3 eligible children).

3.3.5 Household wealth characterization. The primary caregivers of participating children were interviewed in Spanish or Ngäbere at visit 5, using field tested wealth characterization questionnaires based on government questionnaires previously used in the area. Demographic information on household construction (dirt floor, absence of walls, solid walls), possessions (radio, cell phone, bicycle, sewing machine, stove, hoe) and access to running water and latrine was used to calculate a Household 66

Wealth Index (HWI) according to Demographic Health Survey (DHS) methodology [25], based on Principal Components Analysis (PCA) as developed by Filmer and Pritchett [24]. The HWI was calculated for each house as the sum of the possessions and housing quality scores after being multiplied by their relative weights, as determined by the first component of the PCA analysis (n=229). Households were then subdivided into quintiles and categorized as lowest (lowest 40%), middle (middle 40%) and highest (top 20%) wealth groupings, as suggested by Filmer and Pritchett [24].

3.3.6 Anthropometry. Weight and height/length of participating children were measured at visit 4 by trained nutritionists using Seca anthropometry scales (Seca 750, Birmingham, UK), Portable Stadiometers (Seca 214, Birmingham, UK) and Measuring Mats (Seca 210, Birmingham, UK). Using sex and birth date information taken from health cards, child height-for-age (n = 285), weight-for- age (n = 288) and weight-for-height Z scores (n = 280)1 were calculated from WHO growth reference standards [22] using WHO Anthro 3.1. Children were classified as stunted if their height-for-age Z score (HAZ) ≤was -2SD, underweight if their weight-for-age Z score (WAZ) was≤ -2SD, and wasted or overweight/obese if their weight-for-height Z score (WHZ) was≤ -2 SD or ≥ 2S D respectively. Years of school attended by the mother was also collected at visit 4 (n=240).

3.3.7 Fecal samples. During household visits, labelled collection containers were given to each caregiver on the day prior to fecal sample collection with instructions on how to

1 Variability in anthropometry sample sizes is due to error in data recording (height n=7, weight n=4).

collect fecal samples. The following morning samples were collected from the home and transported on ice to the Parasitology Laboratory at the Hospital General del Oriente, Chiriqui, Panamá where MINSA technicians followed MINSA’s standard protocol for clinical diagnosis of diarrhea (single, liquid or semi-liquid stool sample). One direct smear per sample was analyzed to record the presence or absence of the protozoan infections, Giardia sp and/or Entamoeba histolytica/dispar. Up to 7 fecal samples per child were collected from July 2008-October 2009 (Table 1). The number and timing of stool samples was defined by the design of the larger study. Chronic protozoan and diarrhea indices. To assess risk factors for frequent protozoan infection or symptomatic diarrhea, we calculated a Chronic Protozoan Index (CPI) and a Chronic Diarrhea Index (CDI) as the number of positive stool samples divided by the total number of samples provided by each child for whom we had at least 4 fecal samples (n=292). The provision of at least 4 stool samples was deemed sufficient to reflect frequency of infection over the 16 months of the study. It is important to note that the CPI and CDI did not distinguish between new or persistent infections.

3.3.8 Water quality and behavioural risk factors. To assess risk factors for stunting and infection, child defecation habits (n = 279) and hand washing practices (n = 268) were recorded as part of questionnaires administered to the mother during visits 2 and 3 respectively. Water samples were collected during visit 6 from the point of consumption in each household using sealed and sterile IDEXX 100 mL collection containers. Samples were analyzed within 6 hours using Colilert/Quanti-tray® according to manufacturer’s instructions (IDEXX, Westbrook, ME). Quanti-Tray uses a modified most probable number assay to estimate the concentration of Escherichia coli (cfu/100 mL) in samples from each home (n = 206). The method

has a range of < 1 to 2,049 organisms/mL of water, and has been shown to produce similar results to traditional assays [29].

3.3.9 Data analysis.

All statistical comparisons were conducted using STATA 11.1 (College Station, TX). In all cases, the level of significance was set at p < 0.05. Data were reported as mean ± SEM, unless otherwise stated. Continuous data were compared using student t-tests and 2-way ANOVA for normally distributed variables (eg.HAZ, WAZ) or non-parametric Mann-Whitney and Kruskal-Wallis tests for non-normally distributed variables (eg. Water quality, CPI, CDI, HWI, maternal education, distances, age). In the case of multiple comparisons, if initial tests were significant, Tukey or Mann-Whitney tests were conducted as appropriate. Binomial confidence limits (95%) for prevalence data were determined using the Agresti-Coull calculation and comparisons were considered significant where confidence limits did not overlap. To examine the association of child health outcomes with individual and household traits, regression models were conducted on the total child sample and also on a randomly selected index child from each household. To determine risk factors for chronic malnutrition, stepwise multiple regression analyses were conducted with HAZ as the dependent variable. The final model included independent variables with p < 0.10. Of the 244 children for whom HAZ data was available (202 index children), a complete set of data for regression analysis was available for 213 individuals (156 index children). Poisson models determined the risk factors for chronicity of diarrhetic stool and protozoan infection where the dependent variable was the CPI or CDI score multiplied by 100 to satisfy the count criteria. Due to the large number of observed zeros in the CPI, Zero inflated Poisson models were used to distinguish between factors that influenced presence or absence of protozoan infections (Logistic portion) from those that determined the degree of chronicity

of infection (Poisson portion). The final model was chosen based on the lowest Akaike’s Information Criteria (AIC) value. Of the 292 children for whom CPI/CDI indices could be calculated (216 index children), a complete set of data for regression analysis was available for 204 individuals (149 index children). Our power calculations indicated that we required a sample size of 113 to detect an effect size of 0.15 in regression models with 9 independent variables (β=0.80, α=0.05).

3.4 Results 3.4.1 Household and child variables: overview and comparisons between political regions.

Overall, mothers had completed 3.8 ± 0.2 years of education. The majority of houses had dirt floors (91%) and 82% had walls. Latrines were found in 31% of households, but only 11% of children reportedly used them. Instead, most defecation occurred in nearby woods and streams (45%) and in the yard (28%). Piped water was available through aqueducts to only 35% of homes whereas 65% obtained their water from natural sources that included small streams or springs. E.coli was found in 98% of the household water samples, with an average of 230 ± 23 cfu / 100 ml (range = 0 – 1012 cfu / 100 mL). Hand washing with soap before eating was reported for 27% of children. The HWI ranged from - 0.82 to 2.26 with a mean value of 0.21 (± 0.04). Households from the lowest wealth category were characterized by the absence of an aqueduct, latrine, and stove, whereas those with the highest scores typically had a radio, a latrine and an aqueduct, and many had a cell phone, a sewing machine and a stove (Table 2). Households in Soloy were generally better off than those in Emplanada de Chorcha with an average HWI of 0.41 ± 0.06 compared to -0.12 ± 0.04 (Table 3). They were closer to the road, health facility and the school, they had greater

access to an aqueduct and a latrine, and mothers had almost 2 years more education than those in Emplanada de Chorcha (Table 3).

The population of preschool children in this study was 25.5 ± 0.8 months old and 49% were female. A high percentage (60%) of the preschool children was stunted, 5% were wasted, 22% were underweight and, interestingly, 11% were overweight. The prevalence or percentage of children with diarrhetic stool samples was high (45%) in samples collected between August 2008 and April 2009, but low in July 2008 as well as August and October 2009 (Figure 2a). In contrast, the prevalence of protozoan infections at each of the seven sample periods was relatively consistent (22-37%), with Giardia in 18-34% of the children and E. histolytica/ dispar in 2-5% of samples (Figure 2b). Comparisons between the political regions of Soloy and Emplanada de Chorcha revealed no difference in age, sex ratios, stunting, underweight, wasting or chronicity of diarrhetic stools. However, children from Soloy had higher CPI and lower frequency of overweight, compared to those in Emplanada de Chorcha (Table 3).

The study area was accessible through two dirt roads (see Figure 1), the western road having limited access during the rainy season (July—November). On average, the direct line distance between homes and the nearest access road was 1.5 ± 0.1 km, the distance to the nearest health centre was 2.3 ± 0.1 km, and the distance to the nearest elementary school was 0.8 ± 0.04 km. Household density varied considerably through the study area and households were characterized as being “dispersed” or “dense”, based on a natural division in the frequency distribution (Figure 3) at 50 households / km2 with 72% of the homes considered to be “dispersed” and 28% considered to be “dense”. On average, homes in Emplanada de Chorcha were more spatially dispersed than in Soloy. Of the “dispersed households 48% were in Soloy, but Soloy also had 100% of the “dense” households (Figure 1, Table 3).

3.4.2 Spatial comparison of household characteristics.

To address the first of our major objectives, we compared the biophysical / demographic conditions between the “dispersed” and “dense” households. A binary comparison confirmed that “dispersed” homes were approximately three times farther from health facilities and schools than “dense” homes, and 10 times farther from access roads (Table 3). Mothers in the “dispersed” households also had fewer years of education (p=0.05). Household access to sanitation and hygiene infrastructure was greater in the “dense” households where twice as many households had a latrine and aqueduct, and where water quality was also significantly better compared to the “dispersed” households (Table 3). The HWI had a similar range in the “dense” and “dispersed” households (Figure 4) however the average value was 5 times higher in “dense” households (Table 3).

3.4.3 Child health comparisons.

Anthropometry. The prevalence of stunting and underweight was the same in children from “dense” and “dispersed” households, however the frequency of overweight was higher in the “dispersed” households where 14% of children had a WHZ score > 2SD, compared with 3% in the “dense” households (Table 3). Despite better household access to physical infrastructure and greater household wealth, child mean WAZ was lower in the “dense” households. Furthermore, two-way ANOVA confirmed that HAZ was significantly lower in children from “dense” households compared with those from “dispersed” households (F1,162 = 7.71, p = 0.006), and in children from the lowest wealth groupings compared with the highest wealth groupings (F2,162 = 18.49, p < 0.001), with no interaction between these two factors (F2,162 = 2.69, p = 0.10) (Figure 5a). Risk factors associated with low HAZ were being older, living in a household with a lower HWI and having chronic protozoan infections (Table 4). Household

density entered the model, but only with p = 0.08. A virtually identical regression model was obtained based on a single index child per household (Table 4).

Protozoan Infection. The sample specific prevalence of protozoan infections was higher in children from “dense” than “dispersed” households in July 2008 but not in subsequent periods (Figure 2). Chronicity of protozoan infections, however, was higher in children living in “dense” households where the mean CPI was 0.37 ± 0.03 compared to 0.25 ± 0.02 in the “dispersed” households (Table 3). There was no effect of wealth category on chronicity of protozoan infections (X2 adjusted for ties = 2.51, df = 2, p = 0.28) (Figure 5b). The logistic portion of the Zero-inflated Poisson model identified that risk factors for being infected with protozoans were being younger, having worse household water quality and defecating in the yard. The associations of age and household density with being infected with protozoans were further supported by the index child model (Table 5). Predictors of chronicity of protozoan infections (Poisson portion of model) were exactly the same for the total sample model and the index child model (Table 5). Chronicity of protozoan infection was associated with being older and not washing with soap before eating, as well as being from a household with a lower HWI, poorer water quality (higher E.coli counts) or from a household with a greater surrounding population density. Surprisingly, having a mother with more years of education also emerged as a predictor of chronicity of protozoan infections. Analysis based on an index child per household was virtually identical to the total sample model.

Diarrhea: Neither the prevalence of diarrhea at individual sampling points (Figure 2) nor the chronicity of diarrhea (Table 3) differed between “dense” and “dispersed” households with an average CDI of 0.30 in children from both “dense” and “dispersed” households. Average CDI was also similar between children from households in the lowest, medium and highest wealth categories (X2 = 0.57, p = 0.75). Several risk factors for chronicity of diarrhea were similar to 73

those for protozoan infections. Specifically, greater chronicity of child diarrhea was associated with defecating in the yard and not washing with soap before eating (p=0.05) (Table 6). Younger children had higher CDI. In addition, the household factors related to chronic diarrhea were having poor water quality, having a lower HWI and having a mother with more years of schooling. Household density did not enter the model (Table 6). Index child models supported the association of sanitation and hygiene behaviours with chronic diarrhea, but neither age, HWI nor maternal education entered as significant (Table 6).

3.5 Discussion Our study provided an in-depth analysis of stunting, intestinal protozoan infection and diarrhea in a rural indigenous area of Panamá. We have confirmed that in this area of extreme poverty, households from more densely populated areas had better access to health facilities, and sanitation and water infrastructure, less fecal contamination of drinking water as well as greater, albeit still very low, household wealth as measured by possession of durable goods such as a stove, a cell phone, a latrine, and aqueduct access. Counter to our hypothesis, the better socio-economic status of homes from more densely populated areas did not always correspond with better health outcomes. Rather children from these households had a greater severity of stunting and more frequent protozoan infections over a 16 month period than children from more dispersed households. Second, we showed that within these spatially defined groups, child stunting was greater in the lowest wealth households compared to the highest wealth households but that chronicity of protozoan infections was equal across wealth groups. Regression models supported the link between infections and anthropometry outcomes by showing that chronic protozoan infections were a risk factor for low HAZ. The model for protozoan infections 74

highlighted the role of greater household density, few assets and poor water quality, as well as poor hygiene practices in increasing the risk of protozoan infections. Chronicity of child diarrhea was also linked with poor sanitation and hygiene practices as well as poor water quality, but not household density.

Panama’s Indigenous population has disproportionately high levels of poverty characterized by lower access to physical infrastructure and worse health outcomes than their non-indigenous rural counterparts. According to the 2008 Niveles de Vida report [28] 96.3% of the rural Indigenous population lived in poverty, compared to 50.7% of rural non-indigenous population and 18% of the urban population. These impoverished living conditions are accompanied by lower access to physical infrastructure with less than half of the indigenous population reported to have access to an improved water supply and sanitation infrastructure [30]. Child anthropometry outcomes in our study region were well below national averages and similar to rural indigenous statistics for Panamá [30]. At the country level, 18% of preschool children were reported as chronically malnourished (HAZ < -2 SD) [31] a figure that masks the fact that 56.6% of rural indigenous children are short for their age [26]. In our study 60% of children were stunted which is similar to the prevalence of stunting (61%) reported in a nearby region within the Comarca [27]. The previously reported prevalence of low WAZ nationally was 6% but as with HAZ, it was significantly higher in indigenous populations where nearly 25% of preschool children were considered underweight [26], similar to the prevalence recorded in our study. The low level of wasting compared to the high prevalence of stunting observed in this population is a known pattern in Latin America [32]. Stunting is often seen as an epidemiological indicator of the accumulated, long term effects of not only poor diet but also repeated infections [21]. Thus, in this population, the primary nutritional problem is not acute weight loss but rather chronic restriction of a child’s potential growth.

Recently, it has been suggested that use of political regions as the basis for understanding health risks may not be as appropriate as contextual characteristics that more directly impact child health outcomes [33]. By considering the spatial dispersion of households we were able to identify 2 groups with distinct socio-economic and physical infrastructure characteristics. Specifically, homes in the more densely populated areas were closer to a road, a health center and a school. They also had a more complex set of durable goods including radios, cell phones, sewing machines and stoves, and were more likely to have a latrine and aqueduct. However, despite the presence of aqueducts in over 50% of homes in the dense areas, the majority of the household water samples were not potable. E. coli counts in the “dense” households, though lower than in the “dispersed” households, were 100 times higher than the WHO recommendation that E.coli should be absent from drinking water [34]. This paradox is likely due to the inadequate coverage and quality of the infrastructure in the more densely populated areas where less than half of the homes had access to latrines and the majority of children did not use them. These numbers are well below the recommendation of 75% of community coverage considered to maximize community health [35]. Indeed, low coverage as well as poor quality of sanitation infrastructure has been linked to higher prevalence of parasitic infections as well as diarrhea [19]. Thus, it is believed that although homes in the more densely populated areas have better access to physical infrastructure, it was insufficient to meet the health needs of the population.

Despite the extreme poverty across our study area, characterizing household wealth was important for understanding the risk factors related to child health in the comarca Ngäbe-Buglé due to the changes in economic flows associated with physical infrastructure in rural areas [3,4]. Using our Household Wealth Index (HWI), driven primarily by presence or absence of a latrine, an aqueduct, a stove and a dirt floor, we were able to show that “dense” households had a greater

number of assets than “dispersed” households. We also showed that many of the “dense” households were similar in HWI to those in the “dispersed” households. Recently asset-based indices as a measure of “wealth” have been shown to not only be a more reliable index of wealth under conditions of extreme poverty, but also to be related to health outcomes including improved nutritional status [36,37] and reduced parasitic infections [38,39], an observation consistent with our findings. In rural areas where income is inconsistent, asset- based indices are considered a valuable proxy for relative standard of living, which was of particular interest in our study area where wage work is extremely rare and variables such as number of wage earners or income would not capture household financial status. Importantly, due to regional and cultural variability in the range of possessions, the resultant index is location specific [23,40]. To ensure the construction of a regionally appropriate index, we used questionnaires previously designed to record possessions for the comarca Ngäbe Buglé and we also field tested the questionnaire prior to beginning the study. Our resultant HWI using data on 11 possession and household construction variables explained 26% of the variability among households in ownership of these items, similar to a study in rural Africa where 12 variables accounted for 24% of the population variance [38]. The regional specificity in indices is highlighted by a recent study in Peru that used 36 variables to explain a similar degree of variance in the possession ownership (23%) [39]. This is likely due to the urban location of the study where items such as televisions, blenders and furniture were required to distinguish between households.

We found that the severity of stunting was higher in children from “dense” households, as well as those with lower HWI. The finding that older children were shorter for their age is consistent with other studies [21,41] and supports the hypothesis that low HAZ is believed to be due to the cumulative effect of undernutrition [42,43] for children living in deprived conditions. Furthermore,

the identification of chronic protozoan infections (primarily Giardia) as a risk factor for low HAZ highlighted the connection between the stunting and infection. Indeed, Giardia has been reported to have adverse affects on linear growth [44,45] with possible consequences for psychomotor development [46]. We therefore suggest that the greater chronicity of protozoan infection in children from “dense” homes is an important cause of the more severe stunting observed in these children.

Our concept of chronic protozoan infection and chronic diarrhea was based on the idea that repeated or sustained bouts of infection or diarrhea over an extended period of time are of consequence for child development and could result from repeated pathogen exposure or differential ability to access treatment. Thus, the Chronic Protozoan (CPI) and Chronic Diarrhea Indices (CDI) were created as indicators of the frequency of positive samples for protozoan infection or diarrhea over the 16 month study period and are considered to represent potential chronic pathogenicity. Our indices differ from the operational research definitions for chronic and persistent diarrhea [47] in that we did not measure the duration of diarrheal episodes or protozoan infection. Thus, although our indices cannot differentiate between new and sustained illness or classify illness as acute (abrupt onset and resolved≤14 days), chronic (>14 days due to congenital defects in absorption) or persistent (>14 days due to infection or malnutrition) [47], they do provide a long term estimate of the frequency of illness for each child. The fact that the CPI was a predictor of stunting is of particular interest given that the diagnostic methods used by MINSA have a lower sensitivity than other research protocols and may have led to an underestimation of child infection burden. If the lower sensitivity of our analysis resulted in the detection of only more severe infections, then the association of CPI with stunting may also be influenced by the severity of protozoan infection, which would be of interest to examine in future studies.

Furthermore, the similarity between model results for chronic diarrhea and protozoan infection is encouraging for the non-diagnostic definition of diarrhea used in this study.

Using these indices, we found that both protozoan infection and diarrhea were more chronic among children of more educated mothers, a somewhat unexpected finding, given that child health typically improves with increasing years of maternal education [48,49]. In our study area, the more educated women were younger, less experienced mothers (personal observation). In addition to maternal education, gender of the head of household may also influence child health outcomes [50], likely due to lower access to financial resources [51]. As our study was conducted within the context of a CT program where transfers were given to the most senior woman of each household this variable was not included in our analysis. The risk of chronic diarrhea was also higher in younger children, an observation consistent with a longitudinal study in Brazil that recorded a peak burden of diarrhea in children between 7-12 months of age after which the number of episodes/year decreased [52]. In contrast to diarrhea, we found that younger children were more likely to never have had a protozoan infection during the study period. This may have been due to the fact that exclusively breastfed children had a lower exposure to food or water borne pathogens or greater protection through breast milk. However, because the frequency of exclusively breastfed children was low in our study population (personal observation), we did not include it in our analysis. In Peru, children older than 2 years of age were more likely to be infected with protozoans [53], whereas no relationship was detected between age and persistent Giardia (lasting for at least 14 days) in Brazilian studies of urban preschool children [54] or peri-urban children between 0 and 12 years (mean age 3.4 yrs) [39].

Chronic infection and diarrhea were also higher in children with poor hygiene and from households with low water quality. Yard defecation, practiced 79

by nearly a third of children, increased the risk of chronic diarrhea and the likelihood of having had a protozoan infection over the 16 month study period. Open defecation is a recognized source of fecal contamination, especially among young children [55] and although it has been identified as a risk factor for gastrointestinal infection [56] it is often overlooked in sanitation research [57]. Our study also supported previous findings that hand washing before eating reduces soil-transmitted infections in children [58,59] especially when soap is used [60]. Furthermore, poor household water quality (measured by E. coli counts in household water samples) was identified as a risk factor for chronicity of protozoan infection and diarrhea. These results were not surprising given that fecal coliforms were present in water samples from virtually all households, and that pathogens causing both protozoan infection and diarrhea can be spread through contaminated drinking water [34] and prevented through drinking water quality interventions [61].

Finally, we uncovered interactions between household density and wealth with regard to chronic protozoan infection and chronic diarrhea. Population density is an extremely important determinant of risk of transmission of pathogens [14,15]. Indeed, enteric infections have been associated with the increased living densities in urban areas [20] or crowding within a household [16,17], however to our knowledge only one other study has investigated this relationship in rural communities. A study in rural Ecuador found that diarrheal disease prevalence was higher in more densely populated communities and also highlighted the importance of social connections in reducing the risk of diarrhea [18]. In our study, chronicity of diarrhea was independent of household density which may have been due in part to low sensitivity of our binary clinical diagnosis of diarrhea or non-infectious causes of diarrhea such as food-borne pathogens. In addition to household density we also examined the influence of household crowding on chronic infection status. In contrast to the household density

measure, the number of people/room was not significantly related to CPI or CDI in univariate or multivariate analyses.

Our data show that the negative impact of household density on chronic protozoan infection is mitigated by household wealth. Densely spaced households were more likely to have a latrine and aqueduct and also had a more complex set of durable goods including stoves. Although none of the households had potable water, the level of fecal contamination was considerably lower in households with aqueducts compared to those without, suggesting a lower pathogen exposure and helping to explain the lower chronicity of both protozoan infection and diarrhea in households with higher HWI. High values for HWI were most strongly influenced not only by presence of a latrine and an aqueduct, but also ownership of a stove. Perhaps households with a stove are more likely to boil drinking water, thus reducing exposure to water-borne pathogens. In contrast, one of the strongest negative weighting scores in the HWI was the presence of dirt floors, again a factor that would increase the likelihood of transmission of soil-borne pathogens. Our results complement a recent study [39] that examined the link between Giardia and an asset-based household wealth index in a peri-urban region of Peru. Despite the greater affluence in the Peruvian population than in our rural region of extreme poverty, and despite their measure of short-term Giardia infection persisting for more than 14 days compared with our measure of chronic infection over 16 months, they also demonstrated that Giardia was more common in households with a lower HWI. Thus, in the “dense” households where protozoan infection is greater, household wealth plays an important role in reducing chronicity of infection and consequently improving child growth.

Our use of non-validated indices of chronic protozoan infection and diarrhea differs from the duration and frequency of symptomatic episodes in child health literature. However our easily compiled indices based on MINSA field diagnostic 81

procedures proved valuable in assessing the developmental impact of the frequency of illness over a 16 month period. Further limitations of our study related to the inclusion criteria in the recruitment process. Specifically, a selection bias may have been introduced by working exclusively with households that were part of the CT program. It is believed however, that the study population is representative of the region since over 90% of households are eligible to receive the CT program in the study area. Unfortunately, we are not able to compare our sample population to the overall population as the larger census was used to identify eligible households and did not record information from those without preschool children or who had not participated in CT. Furthermore, the inclusion of multiple children per household could have inflated the influence of household level variables in our statistical models. Our total sample regression models however were supported by the virtually identical results obtained from the index child models. Finally, our estimates of household density were limited to participant households thus reflecting the density only of households with children <5 years of age. Although this led to an underestimation of the actual density of households, we believe it is of specific health relevance as the proximity of preschool children has implications for child to child transmission. Based on our intriguing results, we suggest that further research on interactions between household assets and household spacing be conducted with regard to stunting and protozoan infection in preschool children, in order to determine whether our findings can be generalized to other regions of extreme poverty.

Our investigation in the comarca Ngäbe-Buglé of western Panamá has highlighted the link between chronic infection and child growth and furthermore has demonstrated the importance of household density and asset-based household wealth for child growth in an impoverished, rural area. Despite better access to physical infrastructure, children from more spatially dense households

had a greater chronicity of protozoan infection which was associated with stunting. Thus, it is believed that the physical infrastructure available to “dense” households was insufficient to meet the health needs of the population. Finally, we showed that greater chronicity of infection associated with increased population density was mitigated by higher household wealth, suggesting that health interventions are most urgent for children from the poorest households in the more densely populated regions.

3.6 Acknowledgements We thank our collaborators in the Panamanian Ministry of Health and the Instituto Conmemorativo Gorgas de Estudios de la Salud for logistical support and guidance during field work, and community interviewers and study participants for their contributions to the study.

3.7 References 1. United Nations (2009) Rethinking Poverty: Report on the World Social Situation 2010. Geneva: United Nations.

2. Fiszbein A, Schady N (2009) Conditional cash transfers: reducing present and future poverty. Washington, DC: World Bank. 383 p.

3. Khandker SR, Bakht Z, Koolwal GB (2009) The poverty impact of rural roads: evidence from Bangladesh. Econ Dev Cult Change 57: 685-722.

4. Reardon T, Berdegué J, Escobar G (2001) Rural nonfarm employment and incomes in Latin America: overview and policy implications. World Dev 29: 395-409.

5. Escobal J, Ponce C (2004) The Benefits of Rural Roads: Enhancing Income Opportunities for the Rural Poor. Lima, Peru: Grupo de Análisis para el Desarrollo (GRADE). 25 p.

6. Brenneman A, Kerf M (2002) Infrastructure and Poverty Linkages: A Literature Review. Washington, DC: The World Bank. 83

7. Ombok M, Adazu K, Odhiambo F, Bayoh N, Kiriinya R, et al. (2010) Geospatial distribution and determinants of child mortality in rural western Kenya 2002-2005. Trop Med Int Health 15: 423-433.

8. Denno D (2011) Global child health. Pediatr Rev 32: e25-e38.

9. Agee MD (2010) Reducing child malnutrition in Nigeria: combined effects of income growth and provision of information about mothers' access to health care services. Soc Sci Med 71: 1973-1980.

10. Bartram J, Cairncross S (2010) Hygiene, sanitation, and water: forgotten foundations of health. PLoS Med 7: e1000367.

11. Jenkins MW, Curtis V (2005) Achieving the 'good life': why some people want latrines in rural Benin. Soc Sci Med 61: 2446-2459.

12. Jenkins MW, Cairncross S (2010) Modelling latrine diffusion in Benin: towards a community typology of demand for improved sanitation in developing countries. J Water Health 8: 166-183.

13. Hunter PR, MacDonald AM, Carter RC (2010) Water supply and health. PLoS Med 7: e1000361.

14. Luque Fernández MA, Mason PR, Gray H, Bauernfeind A, Fesselet JF, et al. (2010) Descriptive spatial analysis of the cholera epidemic 2008-2009 in Harare, Zimbabwe: a secondary data analysis. Trans R Soc Trop Med Hyg 105: 38-45.

15. Ferrari MJ, Djibo A, Grais RF, Bharti N, Grenfell BT, et al. (2010) Rural-urban gradient in seasonal forcing of measles transmission in Niger. Proc Biol Sci 277: 2775-2782.

16. Younas M, Shah S, Talaat A (2008) Frequency of Giardia lamblia infection in children with recurrent abdominal pain. J Pak Med Assoc 58: 171-174.

17. Chacín-Bonilla L, Barrios F, Sanchez Y (2008) Environmental risk factors for Cryptosporidium infection in an island from Western Venezuela. Mem Inst Oswaldo Cruz 103: 45-49.

18. Bates SJ, Trostle J, Cevallos WT, Hubbard A, Eisenberg JNS (2007) Relating diarrheal disease to social networks and the geographic configuration of communities in rural Ecuador. Am J Epidemiol 166: 1088-1095.

19. Milroy CA, Borja PC, Barros FR, Barreto ML (2001) Evaluating sanitary quality and classifying urban sectors according to environmental conditions. Environ Urban 13: 235-255.

20. Garenne M (2010) Urbanisation and child health in resource poor settings with special reference to under-five mortality in Africa. Arch Dis Child.

21. Lutter CK, Chaparro CM (2008) Malnutrition in infants and young children in Latin America and the Caribbean: achieving the Millenium Development Goals. Washington, DC: Pan American Health Organization. 234 p.

22. WHO Multicentre Growth Reference Study Group (2006) WHO Child Growth Standards: Methods and development: Length/height-for-age, weight- for-age, weight-for-length, weight-for-height and body mass index-for- age: Methods and development. Geneva: World Health Organization. pp. 312.

23. Falkingham J, Namazie C (2002) Measuring Health and Poverty: A Review of Approaches to Identifying the Poor; Centre HSR, editor. London: United Kingdom Department for International Development. 71 p.

24. Filmer D, Pritchett LH (2001) Estimating wealth effects without expenditure data - or tears: an application to educational enrollments in states of India. Demography 38: 115-132.

25. Rutstein S, Johnson K (2004) The DHS Wealth Index. Calverton, Maryland: ORC Macro, MEASURE DHS+.

26. Ministerio de Economía y Finanzas (2003) Encuesta de Niveles de Vida 2003. Panamá: Ministerio de Economía y Finanzas.

27. Payne L, Koski KG, Ortega-Barria E, Scott ME (2007) Benefit of vitamin A supplementation on Ascaris re-infection is less evident in stunted children. J Nutr 137: 1455-1459.

28. Ministerio de Economía y Finanzas (2010) La Distribución del Ingreso en Los Hogares de Panamá:Encuesta de Niveles de Vida 2008. Panamá: Ministerio de Economía y Finanzas. 34 p.

29. Hamilton WP, Kim M, Thackston EL (2005) Comparison of commercially available Escherichia coli enumeration tests: implications for attaining water quality standards. Water Res 39: 4869-4878.

30. World Bank (1999) Panama Poverty Assessment: Priorities and Strategies for Poverty Reduction. Washington, DC: Human Development Department, World Bank.

31. UNICEF (2008) The State of the World's Children 2009: Maternal and Newborn Health. New York: UNICEF. 168 p.

32. Victora CG (1992) The association between wasting and stunting: an international perspective. J Nutr 122: 1105-1110.

33. Fotso JC, Kuate-Defo B (2005) Socioeconomic inequalities in early childhood malnutrition and morbidity: modification of the household-level effects by the community SES. Health Place 11: 205-225.

34. World Health Organization (WHO) (2008) Guidelines for Drinking Water Quality. Geneva: World Health Organization.

35. Bateman OM, Smith S (1991) A comparison of the health effects of water supply and sanitation in urban and rural Guatemala. Washington, DC: US Agency for International Development. 37 p.

36. Van De Poel E, Hosseinpoor AR, Speybroeck N, Van Ourti T, Vega J (2008) Socioeconomic inequality in malnutrition in developing countries. Bull World Health Organ 86: 282-291.

37. Hong R, Mishra V (2006) Effect of wealth inequality on chronic undernutrition in Cambodian children. J Health Popul Nutr 24: 89-99.

38. Raso G, Utzinger J, Silué KD, Ouattara M, Yapi A, et al. (2005) Disparities in parasitic infections, perceived ill health and access to health care among poorer and less poor schoolchildren of rural Côte d'Ivoire. Trop Med Int Health 10: 42-57.

39. Nundy S, Gilman RH, Xiao L, Cabrera L, Cama R, et al. (2011) Wealth and its associations with enteric parasitic infections in a low-income community in Peru: use of principal component analysis. Am J Trop Med Hyg 84: 38- 42.

40. Gwatkin D, Wagstaff A, Yazbeck A (2005) teReaching the poor with health nutrition and population services: What works, what doesn't and why. Washington, DC: World Bank. 354 p.

41. Golden MHN (1996) Specific deficiencies versus growth failure: Type I and Type II. J Nutr Environ Med 6: 301.

42. Black R, E., Allen L, H., Bhutta Z, A., Caulfield L, E. , de Onis M, et al. (2008) Maternal and child undernutrition: global and regional exposures and health consequences. Lancet 371: 243-260.

43. Graham MA (2004) "No somos iguales": the effect of household economic standing on women's energy intake in the Andes. Soc Sci Med 58: 2291- 2300.

44. Sackey ME, Weigel MM, Armijos RX (2003) Predictors and nutritional consequences of intestinal parasitic infections in rural Ecuadorian children. J Trop Pediatr 49: 17-23.

45. Boeke CE, Mora-Plazas M, Forero Y, Villamor E (2010) Intestinal protozoan infections in relation to nutritional status and gastrointestinal morbidity in Colombian school children. J Trop Pediatr 56: 299-306.

46. Simsek Z, Zeyrek FY, Kurcer MA (2004) Effect of Giardia infection on growth and psychomotor development of children aged 0-5 years. J Trop Pediatr 50: 90-93.

47. Thapar N, Sanderson IR (2004) Diarrhoea in children: an interface between developing and developed countries. The Lancet 363: 641-641-653.

48. LeVine RA, Rowe ML (2009) Maternal literacy and child health in less- developed countries: evidence, processes, and limitations. J Dev Behav Pediatr 30: 340-349.

49. Boyle MH, Racine Y, Georgiades K, Snelling D, Hong S, et al. (2006) The influence of economic development level, household wealth and maternal education on child health in the developing world. Soc Sci Med 63: 2242-2254.

50. Haidar J, Kogi-Makau W (2009) Gender differences in the household- headship and nutritional status of pre-school children. East Afr Med J 86: 69-73.

51. McIntyre L, Rondeau K, Kirkpatrick S, Hatfield J, Islam KS, et al. (2011) Food provisioning experiences of ultra poor female heads of household living in Bangladesh. Soc Sci Med 72: 969-976.

52. Genser B, Strina A, Teles CA, Prado MS, Barreto ML (2006) Risk factors for childhood diarrhea incidence: dynamic analysis of a longitudinal study. Epidemiology 17: 658-667.

53. Silva RR, da Silva CAM, de Jesus Pereira CA, de Carvalho Nicolato RL, Negrão- Corrêa D, et al. (2009) Association between nutritional status, environmental and socio-economic factors and Giardia lamblia infections among children aged 6-71 months in Brazil. Trans R Soc Trop Med Hyg 103: 512-519.

54. Teixeira JC, Heller L, Barreto ML (2007) Giardia duodenalis infection: risk factors for children living in sub-standard settlements in Brazil. Cad Saude Publica 23: 1489-1493.

55. Lanata C, Huttly SRA, Yeager BAC (1998) Diarrhea: whose feces matter? Reflections from studies in a Peruvian shanty town. Pediatr Infect Dis J 17: 7-9.

56. Quihui L, Valencia ME, Crompton DW, Phillips S, Hagan P, et al. (2006) Role of the employment status and education of mothers in the prevalence of intestinal parasitic infections in Mexican rural schoolchildren. BMC public health 6: 225.

57. Mara D, Lane J, Scott B, Trouba D (2010) Sanitation and health. PLoS Med 7: e1000363.

58. Rinne S, Rodas EJ, Galer-Unti R, Glickman N, Glickman LT (2005) Prevalence and risk factors for protozoan and nematode infections among children in an Ecuadorian highland community. Trans R Soc Trop Med Hyg 99: 585- 592.

59. Hussein AS (2011) Prevalence of intestinal parasites among school children in northern districts of West Bank- Palestine. Trop Med Int Health 16: 240- 244.

60. Curtis V, Cairncross S (2003) Effect of washing hands with soap on diarrhoea risk in the community: a systematic review. Lancet Infect Dis 3: 275-281.

61. Fewtrell L, Kaufmann RB, Kay D, Enanoria W, Haller L, et al. (2005) Water, sanitation, and hygiene interventions to reduce diarrhoea in less developed countries: a systematic review and meta-analysis. Lancet Infect Dis 5: 42-52.

Table 1. Timeline of sample collection during 10 household visits from June 2008 – October 2009.

2008 2009

Visit 1 Visit 2 Visit 3 Visit 4 Visit 5 Visit 6 Visit 7 Visit 8 Visit 9 Visit 10 June July July July August October April May August Octobe r Household Recruit 262 Household Wealth 229 89

Water Quality 206 Mother’s Education 240 Household GPS3 250 Child Recruit 373 Fecal Sample1 229 223 237 238 244 260 244 Defecation Practices 279

Handwashing 268 Behaviours Anthropometry2 280-288

1Children that provided fecal samples in at least 4 sample periods were included in the Chronic Protozoan and 1Diarrhea Index calculations (n=292).

2 Variability in anthropometry sample sizes is due to error in data recording (height n=7, weight n=4).

3 GPS coordinates were recorded during visits 7 – 10.

Table 2. Household variables included in the Household Wealth Index (HWI), their respective scoring factors and the percent of households in each wealth category owning items.

Middle Scoring Lowest Highest 1 (Middle Factor (Lower 40%) (Upper 20%) 40%)

Variable Aqueduct 0.3911 2% 41% 90% Latrine 0.4114 2% 29% 92% Radio 0.1595 71% 82% 98% Cell Phone 0.2948 2% 7% 42% Bicycle 0.1358 1% 2% 6% Hoe -0.1156 38% 50% 17% Sewing 0.1802 38% 59% 71% Stove 0.3888 0% 7% 43% No walls -0.2779 40% 4% 0% Sol id walls 0.298 5% 59% 71% Dirt floor -0.429 100% 98% 54% HWI Mean (SE) -0.36 (0.02) 0.25 (0.02) 1.27 (0.18) Range -0.82, -0.089 -0.087, 0.67 0.69, 2.26 1 Scoring factors were determined using the variable coefficients from the first component of a Principle Component Analysis that explained 26% of the variability in the 11 variables.

Table 3. Summary of household and child characteristics and health outcomes according to political region and household density category1.

Dispersed Dense Emplanada de Soloy p Households Households p Chorcha2 ( < 50 houses/km2) ( > 50 houses/km2) Household 2.4(0.2) 0.9(0.09) 1.9(0.1) 0.2(0.02) Distance to road, km <0.001 <0.001 96 154 181 69 Distance to health facility, 2.9(0.2) 2.0(0.1) 2.8(0.1) 1.1(0.07) <0.001 <0.001 km 96 154 181 69 1.1(0.06) 0.6(0.04) 0.9(0.04) 0.4(0.02) 92 Distance to school, km <0.001 <0.001 96 154 181 69

Household Density3, 14.5(1.3) 48.1(3.0) <0.001 NA NA homes/km2 96 154 48 (40 - 55) 100(94 - 100) Households in Soloy, % NA NA <0.001 181 69 -0.12(0.04) 0.41(0.06) 0.097(0.05) 0.51(0.09) Household Wealth Index4 <0.001 <0.001 90 139 172 57 2.7(0.3) 4.6(0.3) 3.6(0.3) 4.6(0.5) Maternal Education, yrs <0.001 0.05 94 146 176 64 15(9 - 24) 41(34 - 49) 26(20 - 33) 45(33 - 58) Latrine, % <0.001 <0.001 93 142 175 60 12(7 - 20) 49(42 - 57) 27(21 - 34) 52(40 - 63) Aqueduct, % <0.001 <0.001 94 146 176 64

Water Quality, E.coli 270(39) 184(28) 243(28) 151(38) 0.002 0.004 cfu /100mL 83 123 152 54 Child

26.3(1.4) 25(1.1) 25.5(1.0) 25.7(1.6) Age, mo 0.5 0.93 139 217 268 88 47(39 - 55) 50(44 - 57) 48(42 - 54) 51(41 - 61) Female, % 0.6 0.58 139 217 268 88 0.24(0.02) 0.31(0.02) 0.25(0.02) 0.37(0.03) Protozoan Infection, CPI5 0.005 <0.001 122 170 228 64

0.3(0.02) 0.3(0.02) 0.30(0.02) 0.30(0.03) Diarrhetic Stool, CDI6 0.92 0.69 122 170 228 64 HAZ

61(53 - 70) 60(52 - 67) 58(51 - 64) 70(57 - 80) Stunting ( < -2 SD), % 0.74 0.09 122 163 225 60 93

-2.33(0.12) -2.18(0.1) -2.16(0.08) -2.56(0.18) Mean Z score 0.33 0.03 122 163 225 60 WAZ

Underweight ( < -2 SD), 22(16 - 31) 22(16 - 29) 20(16 - 26) 29(19 - 41) 0.95 0.17 % 125 163 225 63

-0.94(0.13) -0.99(0.1) -0.87(0.09) -1.3(0.2) Mean Z score 0.74 0.03 125 163 225 63 WHZ 7(3 - 13) 4(2 - 9) 6(3 - 10) 3(0 - 12) Wasting( <- 2 SD), % 0.4 0.45 121 159 221 59

17(12 - 25) 7(4 - 12) 14(10 - 19) 3(0 - 12) Overweight ( > 2 SD), % 0.007 0.03 121 159 221 59

1Values are mean ± SE or % (binomial 95% CI). Sample size is in parentheses below the summary statistics. 2Summary data presented for Emplanada de Chorcha and Soloy is restricted to households for which spatial data was available. 3Density based on households participating in study 4Asset based index, weights derived from the first component of Principle Components Analysis 5Chronic Protozoan Index (CPI) proportion of positive samples provided (for children who provided 4 or more samples of 7 possible sampling periods) 6Chronic Diarrhea Index (CDI), proportion of positive samples provided (for children who provided 4 or more samples of 7 possible sampling periods). 94

Table 4. Predictors of height-for-age Z-score in Panamanian preschool children from a stepwise multiple regression model1.

Total Sample Index Child Raw Standardized Raw Standardized p p Coefficient Coefficient Coefficient Coefficient Age, mo -0.03 -0.03 < 0.001 -0.04 -0.03 <0.001 Water Quality, E.coli cfu /100 mL NE2 - - NE - - Maternal Education, yrs NE - - NE - - Household Wealth Index, HWI3 0.40 0.21 0.01 0.29 0.15 0.02 Wash with soap4 NE - - NE - - Yard Defecation5 NE - - NE - - Protozoan Infection, CPI6 -1.06 -0.20 0.002 -1.41 -0.25 <0.001 Diarrhetic Stool, CDI7 NE - - NE - - 95 Household Density, homes/km2 -0.005 -0.11 0.08 NE - -

Intercept -1.08 - < 0.001 -0.90 - <0.001 Adjusted R2 0.26 0.33

F4,208 19.50 F 3,152 26.44

p > F < 0.001 <0.001

n 213 156

1HAZ available: n=285 children (202 index children), complete data set available: n=213 (156 index children). 2Excluded during stepwise process if p > 0.10 3Asset based index of household wealth, weights derived from the first component of Principle Components Analysis.

4Child washes hands with soap before eating (0 = no, 1 = yes), as reported by the mother. 5Child defecation occurs primarily in the yard around the home (0 = no, 1 = yes), as reported by mother. 6Chronic Protozoan Index (CPI), proportion of positive samples provided (for children who provided 4 or more samples of 7 possible sampling periods). 7Chronic Diarrhea Index (CDI), proportion of positive samples provided (for children who provided 4 or more samples of 7 possible sampling periods). 96

Table 5. Predictors of chronicity of protozoan infection in Panamanian preschool children based on a Zero-inflated Poisson regression model1. Total Sample2 Index Child3 Logistic Portion4 Coefficient P Coefficient p Age, mo -0.04 0.002 -0.04 0.01 Mother’s Education, yrs NE5 - NE - Wash with soap6 NE - NE - Water Quality, cfu E.coli/100 -0.002 0.009 NE - mL Household Wealth Index, NE - NE - 7 HWIYard Defecation 8 -0.85 0.046 -0.86 0.07 Household Density, -0.01 0.06 -0.02 0.05 h /k 2 Intercept 0.94 0.04 0.73 0.15 Poisson Portion9 Coefficient P Coefficient p Age, mo 0.005 < 0.001 0.007 <0.001 Wash with soap -0.088 0.003 -0.21 <0.001 Household Wealth Index, -0.06 0.01 -0.10 <0.001

WaterHWI Quality, cfu E.coli/100 0.0002 <0.001 0.0001 0.008 mL Household Density, 0.002 < 0.001 0.003 <0.001 h /k 2 Mother’s Education, yrs 0.03 < 0.001 0.03 <0.001 Yard Defecation NE - NE - Intercept 3.31 < 0.001 3.23 <0.001 1CPI available: n=292 children (216 index children), complete data set available: n=204 (149 index children). 2 Total Sample Model Statistics: Likelihood Ration X2 = 105.46, p < 0.001, n = 204, zero observations = 55, Vuong statistic of Zero Inflated Poisson vs Poisson z = 10.94, p > z < 0.001. 3 Index Child Model Statistics: Likelihood Ration X2 = 141.51, p < 0.001, n = 149, zero observations = 43, Vuong statistic of Zero Inflated Poisson vs Poisson z = 9.42, p <0.001 4Logistic portion of the model identifies variables associated with likelihood of being uninfected where positive coefficients indicate a greater likelihood of being uninfected 5Excluded during stepwise process if p>0.10 6Child washes hands with soap before eating (0 = no, 1 = yes) as reported by the mother

7Asset based index of household wealth derived using Principle Components Analysis 8Child defecates in the yard around the home (0 = no, 1 = yes) as reported by the mother 9Poisson Portion of the model determines variables associated with chronicity of protozoan infection.

Table 6. Predictors of chronicity of diarrhetic stool in Panamanian preschool children from a Stepwise Poisson regression model1. Total Sample Index Child Raw Standardized Raw Standardized p p Coefficient Coefficient Coefficient Coefficient Age, mo - 0.008 0.99 < 0.001 -0.002 0.97 0.08 Water Quality, E.coli 0.0003 1.10 < 0.001 0.00026 1.09 <0.001 cfu/100 mL Maternal Education, yrs 0.01 1.01 0.01 -0.008 0.97 0.07 Household Wealth -0.10 0.90 < 0.001 NE - - Index2 Wash with soap3 -0.06 0.97 0.05 -0.11 0.95 0.002 Yard Defecation4 0.08 1.04 0.005 0.21 1.10 <0.001 Household Density,

99 5 2 NE - - NE - -

homes/km Intercept 3.51 - < 0.001 3.30 - <0.001 LR X2 197.13 92.82 P value < 0.001 <0.001 N 204 149 1CDI available: n=292 children (216 index children), complete data set available: n=204 (149 index children). 2Asset based index, weights derived from the first component of Principle Components Analysis 3Child washes hands with soap before eating (0 = no, 1 = yes), as reported by the mother 4Child defecation occurs primarily in the yard around the home (0 = no, 1 = yes), as reported by mother 5Excluded during stepwise process if p > 0.10

100

Figure 1. Household density clusters in study area located in western Panamá. Household density clusters, health facilities and schools in the region participating in the study. The capital Panamá City is indicated by a star and dashed lines indicate village. Emplanada de Chorcha: aAlto Caña, bCerro Viejo, cChorchita, dLa Juventud, ePlan de Chorcha, fQuebrada de Lajas; Soloy: gBarrio 19 abril, hBarrio 2000, iBoca de Huso, jBoca de Jebay, kBoca Miel, lIsrael. 101

101

Figure 2. Prevalence of symptomatic diarrhea (a) and protozoan infection (b) over the duration of the study. Prevalence with 95%CI are presented from each sample period for children from dispersed (dark gray) and dense (light gray) households. Household density was classified as “dispersed” if < 50 houses/km2 (n = 178-200 children) or “dense” if > 50 houses/km2 ( n = 45-60 children).

102

Figure 3. Frequency distribution of density of participating households.

Household density was determined using ArcGIS point density estimates with a radius of 250 m around each home (n = 250; mean = 35.2 ± 2.2 houses/km2, minimum of 4.5 and maximum of 128 houses/km2). Households were classified as dispersed (n = 181) or dense (n = 69) according to the natural break in the frequency distribution at 50 houses/km2.

103

18 Frequency Frequency

0 -1 -.5 0 .5 1 1.5 2 Household Wealth Index

Dispersed Dense

Figure 4. Frequency distribution of Household Wealth Index (HWI) by household density group. The HWI is an asset-based index derived from measures of household construction and ownership of durable goods. Household density was classified as “dispersed” if < 50 houses/km2 (n = 172 households) or “dense” if > 50 houses/km2 ( n = 57 households).

104

Figure 5. Mean child health outcomes by household density group and wealth percentile.

Child mean (± SE) values for (a) height-for-age Z-score (n from left: 99, 93, 25, 16, 18, 23) and (b) chronicity of protozoan infection (CPI) (n from left: 107, 87, 28, 14, 19, 26) in the “dispersed” (< 50 houses/km2) and “dense” households ( > 50 houses/km2). The Household Wealth Index (HWI), an asset-based index derived from measures of household construction and ownership of durable goods, was used to define wealth groups (bottom 40%, middle 40% and top 20%) 105

represented by dark to light grey bars. Unique letters indicate significantly different groups (p < 0.05) determined by 2-way ANOVA and post hoc Tukey test (a) or Mann Whitney non-parametric tests (b).

106

Connecting Statement I

In Chapter 3, I developed a new index to examine how frequency of diarrhea and protozoan infection over 16 months relates to child height for age. Using this index, I showed that chronicity of protozoan infection but not diarrhea was related to chronic malnutrition. Hygiene behaviours (yard defecation and washing with soap), and household water quality were influential in both diarrhea and protozoan infection. Using estimates of household density I determined that despite better access to latrines and aqueducts, densely populated households had more chronic protozoan infection than dispersed households. The negative effect of chronic protozoan infection on child HAZ was mitigated by household wealth. In Chapter 4, I further explore the importance of regional traits on gastrointestinal infection by examining the influence of spatial clustering patterns on reinfection dynamics of three STH infections. In addition to regional factors (high prevalence infection cluster), I explore the relative importance of household exposure variables (maternal education, HWI, household density) and individual exposure (defecation habits) and susceptibility variables (age, HAZ) for child reinfection intensity. The research in Chapter 4 allowed me to examine whether stunting increased susceptibility to reinfection. This manuscript was co-authored with my co-supervisors Kristine Koski and Marilyn Scott, our Panamanian co-investigators, Victoria Valdez, and an undergraduate student who was involved in the spatial analysis, Claire Paller

107

CHAPTER 4

Regional, household and individual factors that influence soil transmitted helminth reinfection dynamics in preschool children from rural indigenous Panamá

Carli M. Halpenny1, Claire Paller1, Kristine G. Koski2, Victoria E. Valdés3 and Marilyn E. Scott1

1Institute of Parasitology and McGill School of Environment, 2School of Dietetics and Human Nutrition Macdonald Campus of McGill University, Ste-Anne de Bellevue, Quebec, Canada, 3Escuela de Nutrición y Dietética, Facultad de Medicina, Universidad de Panamá, Ciudad de Panamá, Panamá

Revised submission to PLoS Neglected Tropical Diseases: November, 2012. 108

4.1 Abstract BACKGROUND: Few studies have investigated the relative influence of individual susceptibility versus household exposure factors versus regional clustering of infection on soil transmitted helminth (STH) transmission. The present study examined reinfection dynamics and spatial clustering of Ascaris lumbricoides, Trichuris trichiura and hookworm in an extremely impoverished indigenous setting in rural Panamá over a 16 month period that included 2 treatment and reinfection cycles in preschool children. METHODOLOGY/PRINCIPLE FINDINGS: Spatial cluster analyses were used to identify high prevalence clusters for each nematode. Multivariate models were then used (1) to identify factors that differentiated households within and outside the cluster, and (2) to examine the relative contribution of regional (presence in a high prevalence cluster), household (household density, asset- based household wealth, household crowding, maternal education) and individual (age, sex, pre-treatment eggs per gram (epg) feces, height-for-age, latrine use) factors on preschool child reinfection epgs for each STH. High prevalence spatial clusters were detected for Trichuris and hookworm but not for Ascaris. Theses clusters were characterized by low household density and low household wealth indices (HWI). Reinfection epg of both hookworm and Ascaris was positively associated with pre-treatment epg and was higher in stunted children. Additional individual (latrine use) as well as household variables (HWI, maternal education) entered the reinfection models for Ascaris but not for hookworm. CONCLUSIONS/SIGNIFICANCE: Even within the context of extreme poverty in this remote rural setting, the distinct transmission patterns for hookworm, Trichuris and Ascaris highlight the need for multi-pronged intervention strategies. In addition to poverty reduction, improved sanitation and attention to chronic malnutrition will be key to reducing Ascaris and hookworm transmission.

109

4.2 Introduction The soil transmitted helminth (STH) infections, Ascaris lumbricoides, Trichuris trichiura and the hookworms Ancylostoma duodenale and Necator americanus are estimated to result in the loss of 39 million disability adjusted life years (DALYs) annually [1] and to have long-lasting implications for child physical and cognitive development [2-5]. To control the morbidity associated with these infections for high risk populations living in endemic communities, current public health programs focus on chemotherapy [6] through programs typically delivered to school children [7]. Although there is evidence from theoretical models [8] and community studies [9] that such efforts have spill-over effects that reduce transmission in the untreated portions of the population, ensuring control of infections in preschool children is particularly important given the impact of STH on early growth and development [5,10].

Many studies have shown that the individuals most heavily infected prior to anthelmintic treatment have a high reinfection burden, and thus that some individuals are predisposed to heavy infection [11,12]. Mechanisms used to account for this predisposition include genetic and nutritional components of susceptibility, behavioural patterns that directly promote contact with infective stages and factors that promote egg/larval survival in some domestic and peri- domestic environments. Susceptibility may be related to differences in the genetic regulation of B cell activation and immunoglobulin secretion [13], and to poor nutritional status [14-16] through the impaired immune function that accompanies micro and macronutrient deficiencies [10]. Behaviours including geophagy [17,18], poor hygiene [19] and not wearing shoes [20] reportedly increase an individual’s risk of contact with eggs or larvae. Household level factors associated with poverty [21], such as limited latrine access [16,21,22], within household crowding [23,24] and low maternal education [21,25,26] 110

increase environmental contamination with STH eggs and larvae and contribute to the patterns of household aggregation detected in community-level epidemiological studies [27,28].

It has also been shown that STH infections cluster at the regional and national scales [29,30]. Such clusters have been associated with biophysical and climatic features such as vegetation cover [31,32], soil type [32,33], rainfall, temperature and altitude [29,31,34] that influence egg and larval survival, and also with limited sanitation and hygiene infrastructure that promote transmission [35]. Given the challenge of disentangling the role of individual, household and regional factors in determining risk of STH transmission, spatial analysis is becoming more widely used in epidemiological studies. Three recent investigations that included these methods highlighted the relative importance of household-level socio-economic (SES) factors. In Uganda, a cross-sectional survey found that household variables influencing exposure play a greater role than host genetics in determining the distribution of hookworm infection intensity [36]. In Brazil, a longitudinal study that assessed hookworm post treatment reinfection rates found that child sex and SES variables such as household construction were more influential than regional geographical variables such as rurality, altitude and soil moisture [37]. In Bangladesh, a longitudinal study found that household exposure risk factors accounted for more than half of the variability in household clustering of Ascaris infection and that after accounting for household clustering, individual predisposition to infection was minimal [38]. Thus it is well established that SES variables are influential in determining the spatial distribution and transmission of STH infections [36,38]. However, few studies have the detailed individual data to investigate the relative influence of individual susceptibility due to chronic undernutrition versus household exposure variables on STH transmission within a context of regional clustering of infection in preschool children.

111

Our study was designed to examine the influence of spatial patterns of infection as well as household and individual factors on the transmission dynamics of Ascaris lumbricoides, hookworm, and Trichuris trichiura during two sequential reinfection cycles in preschool children in an impoverished, rural region of Panamá. Specifically, we aimed to: 1) compare reinfection dynamics among the three STH; 2) identify and characterize regional scale spatial clusters of high prevalence of infection; and 3) determine the relative contributions of individual factors (age, sex, height-for-age, pre-treatment intensity, latrine use), household factors (household density, asset-based household wealth index, household crowding, maternal education, preschool sibling infection status) and a regional factor (residence in a high prevalence cluster) on reinfection intensity of STH infections in preschool children.

4.3 Materials and Methods 4.3.1 Ethics Statement.

Ethical approval was obtained from the Instituto Conmemorativo de Gorgas in Panamá and McGill University in Canada. The study was conducted in accordance with the Guía para Realizar Estudios e Investigaciones en los Pueblos Indígenas de Panamá, including initial and result-sharing workshops with participants in each village. Written informed consent was obtained from primary caregivers during a household visit that included an explanation of study significance, of participant requirements and rights as well as an opportunity to ask questions in Spanish and Ngäbere.

4.3.2 Study area.

The comarca Ngäbe-Buglé is a semi-autonomous political region inhabited primarily by the Ngäbe and Buglé indigenous groups (2004 population estimate of 128,978) where over 90% of families live in extreme poverty (< 112

US$1.75/day) [39]. In 2005 and 2007, two forms of the Conditional Transfer (CT) program Red de Oportunidades began in the comarca, providing an additional US$50/mo in either cash (2007) or food vouchers (2005) in exchange for participation in health and education programs. The research reported here is part of a larger study that was conducted in the district of Besiko in two adjacent corregimientos (Soloy and Emplanada de Chorcha), each accessible most of the year by one dirt road. As previously reported [40], household density varied considerably. More densely populated regions (>50 participant households/km2) were closer to a road and had better access to latrines, aqueducts and health facilities and had a higher average asset-based household wealth index (HWI).

4.3.3 Study Design and Protocol. Our study was designed to estimate STH infection and reinfection in 2 treatment and reinfection cycles during a 16 month sample period (Cycle 1: 9 month reinfection period from July 2008 to April 2009; Cycle 2: 6 month reinfection period from April to October 2009). This study recruitment protocol has been described previously [40]. In brief, households with children from 0 - 48 mo of age and living in extreme poverty (defined as having participated in a CT program) from 12 randomly selected villages split evenly between the 2 corregimientos were invited to participate in the study. All but 3 of the 265 eligible households that were approached agreed to participate. Spatial data were unavailable for 12 households (missing or erroneous longitude and latitude coordinates). The household and demographic characteristics of the 12 households for which spatial data were unavailable did not differ from the other 250 participant households. Thus, the present analysis included a total of 250 households and 356 children (153 households with 1 eligible child, 88 with 2, and 9 households with 3 eligible children). Stool samples were collected at 7 household visits, and additional data from questionnaires and anthropometry

113

were provided during 3 additional household visits. Temporary migration for agricultural purposes was common in the participant population and therefore few children were available at all household visits.

Cycle 1 stool samples and treatment: Baseline fecal samples for reinfection Cycle 1 were collected in June 2008 (n=215), after which a single dose of Albendazole (ABZ) was distributed to all available participant children > 12 mo of age (59%) according to Ministry of Health procedures (200 mg: 1-2 yrs, 400 mg: 3-5 yrs). These children were observed during treatment administration to ensure compliance. A post-treatment fecal sample was collected 2 wk later (n=100) from which drug efficacy against Ascaris was assessed from the subsample of 16 children who had been infected and had received treatment. Reinfection was monitored through stool samples collected at 3 mo (n=87) and 9 mo (n=155) (Table 1) from children who had received treatment. From the 9 mo reinfection sample, 122 of the 155 children had also provided baseline samples for Cycle 1. At 9 mo, stool samples were also available from an additional 115 children who had not been treated in Cycle 1 either because they were too young or they were unavailable when the treatment was administered. These data have not been included in Cycle 1 but did form part of the baseline sample for Cycle 2.

Cycle 2 stool samples: The 9 mo reinfection sample (April 2009) also served as the baseline sample for Cycle 2 (n=270), after which a single dose of ABZ was distributed as above to all available participant children > 12 mo of age (78%) (Table 1). Three weeks later, drug efficacy was evaluated as above (Ascaris n=32, hookworm n=25, Trichuris n=7) and those who remained infected with at least one STH (n=18) were given a second dose of ABZ. Reinfection was monitored through stool samples collected at 4 (n=218) and 6 (n=200) mo after ABZ treatment. In Cycle 2, 172 children provided baseline fecal samples, received treatment and provided a 6 mo reinfection sample. Forty eight percent of

114

preschoolers enrolled in the study received treatment in both Cycle 1 and Cycle 2.

4.3.4 Fecal samples. Labelled collection containers and detailed instructions were given to each caregiver during household visits on the day prior to fecal sample collection. Samples were collected from the home the following morning and transported on ice to the Parasitology Laboratory at the Hospital General del Oriente, Chiriqui, Panamá.

The primary outcome in this study was intensity of infection, measured as epg. Hookworms were not identified to species level, Necator americanus is the predominant hookworm in Central America [41]. In Cycle 1, data on epg were obtained only from duplicate Kato Katz preparations [42] whereas in Cycle 2, both Kato Katz and the FLOTAC techniques [43] were used. For each nematode (Ascaris, Trichuris and hookworm), a comparison of the diagnostic ability (presence/absence) between FLOTAC and Kato Katz was conducted using Cohen’s Kappa statistic on samples that were assayed using both techniques. Sensitivity of each method (expressed as a percentage) was also analyzed by dividing the number of positives for a given method by the total number of positives identified by either method. Finally, the correlation of intensity estimates between the two methods was examined using Spearman Rank correlation coefficients.

4.3.5 Characterization of infection risk factors. Individual factors. Age and sex were recorded from child health cards during household interviews. Child latrine use (n: 2008=279, 2009=286) was obtained from questionnaires. Pre-treatment epgs were measured as above, and were

115

used both as an indicator of baseline infection status, and also as an indicator of individual predisposition to reinfection [44,45]. Based on previous observations that child stunting was associated with more rapid reinfection in a nearby population [46], height-for-age Z score (HAZ) was used to assess individual susceptibility to STH reinfection. Height/length of participating children was measured by trained nutritionists using Portable Stadiometers (Seca 214, Birmingham, UK) and Measuring Mats (Seca 210, Birmingham, UK). Child HAZ scores (2008: n= 285, 2009: n = 264) were calculated from WHO growth reference standards using WHO Anthro 3.1 [47]. Children were classified as stunted if their HAZ was < -2SD.

Household factors. The primary caregivers of participating children were interviewed in Spanish or Ngäbere using field tested questionnaires during several household visits to gather information related to factors that directly influence exposure to infection. Years of school attended by the mother (n=240), and number of family members per room (measure of household crowding; n=217) were recorded. Information on 11 variables relating to household construction (dirt floor, absence of walls, solid walls), possessions (radio, cell phone, bicycle, sewing machine, stove, hoe) and access to running water and latrine (n = 229) were used to calculate an asset-based Household Wealth Index (HWI) that describes relative poverty within our population as previously described [40]. Geographic coordinates recorded with a handheld Geographic Positioning System (GPS) (Garmin eTrex Vista HCX, Olathe KS) (n = 250) were used to characterize the spatial dispersion of participant homes [40]. Resulting density estimates for each household described the number of other study participant homes found within a radius of 250m, expressed in homes per square kilometer.

Regional factor. We also conducted spatial analyses across the study area on the location of households with an infected preschool child at baseline of both Cycle 116

1 and Cycle 2. Spatial clusters of “infected” households were detected using SaTScan software (Boston, MA). Kulldorf’s Space-Time Scan Statistic uses a scanning circular window to identify distinct clusters of infection. For each location the circular window varied in size from zero to 50% of the population and the distribution of cases inside the window was compared to that outside the window. Likelihood ratios based on a Bernoulli probability model identified high prevalence clusters as those where the number of households with at least one infected preschool child was greater than expected if the infection was randomly dispersed among households (p<0.05) [48,49]. This approach revealed whether or not regional clusters were present for each STH at baseline of each Cycle, and if present, where they were located.

4.3.6 Statistical analysis.

All statistical comparisons were conducted using STATA 11.1 (College Station, TX). In all cases, the level of significance was set at p < 0.05. Binomial confidence limits (95%) for prevalence data were determined using the Agresti- Coull calculation and comparisons were conducted using contingency tables and X2 tests. Continuous data were reported as the mean ± SEM, unless otherwise stated. Univariate comparisons between households within and outside high prevalence clusters, as well as between infected and uninfected individuals at both the 9 mo (Cycle 1) and 6 mo (Cycle 2) reinfection sample periods, were conducted using non-parametric Mann-Whitney, Kruskal-Wallis and Spearman correlation analyses due to the non-normal distribution of the data (e.g. maternal education, HWI, age, epg). Univariate analyses related to reinfection included only those children who had provided a 9 mo / 6 mo reinfection sample and who had received treatment (Cycle 1, n=155; Cycle 2, n=200).

Treatment efficacy was evaluated in Cycle 1 and Cycle 2 for the subsample of children who provided both baseline and immediate post-

117

treatment stool samples. Egg Reduction Rates (ERR) were calculated as the mean percentage reduction in epg [50], using epgs from the Kato Katz for Cycle 1 and from FLOTAC for Cycle 2.

Step-wise logistic regression models were used to examine which household risk factors were associated with presence in the high prevalence clusters detected by the SaTScan software (hookworm and Trichuris at baseline of Cycle 2). Step-wise negative binomial regression models were also performed to determine the impact of three sets of risk factors on reinfection epgs: 1) regional factor, namely residence in a high infection cluster; 2) household factors, namely HWI, household density, maternal education); and 3) individual factors, namely age, sex, HAZ, predisposition measured by pre-treatment epg, latrine use). These analyses were done for Ascaris reinfection in both Cycle 1 and Cycle 2 and for hookworm only in Cycle 2 when the more reliable FLOTAC epgs were available. No regression models were developed for Trichuris reinfection because of the low drug efficacy. Final models included variables with p<0.10. The Huber estimator for robust standard error estimation was used to account for clustering at the household level. Multivariate analyses were limited to individuals who had received treatment and provided baseline and 9 mo or 6 mo reinfection samples and for whom a complete set of data on risk factors were available (Cycle 1, n=100; Cycle 2, n=140).

4.4 Results 4.4.1 Methodological comparison.

Although there was significant concordance between the methods for detection of all three parasites (Ascaris: k=0.91, p<0.001; Trichuris: k=0.66, p<0.001; hookworm: k=0.52, p<0.001; n=604), FLOTAC was more sensitive than Kato Katz in the detection of hookworm (93% vs 45% of known positive samples, respectively) and Trichuris (97% vs 56% of known positive samples, respectively).

118

Spearman rank correlation analysis of intensity estimates among samples identified as positive by both methods (n=196) revealed a positive correlation between methods for Ascaris epg (r=0.79, p<0.001), hookworm (r=0.57, p<0.001) and Trichuris (r=0.51, p=0.003). Due to the greater sensitivity of FLOTAC to hookworm and Trichuris infection, the similar diagnostic ability of both methods for Ascaris and the larger number of samples examined by FLOTAC in 2009, analysis of infection prevalence and intensity used FLOTAC estimates whenever available.

4.4.2 Household and demographic characteristics of participants.

Average density of participating households was 35 ± 2 houses / km2 and the household wealth index (HWI) was 0.21 ± 0.04. Latrines were available to 31% of households and piped water (aqueducts) to 34% of households. Households had an average of 5.4 ± 0.2 people/room and mothers had 3.8 ± 0.2 yrs of education. The average age of participating children at baseline of Cycle 1 was 31.4 ±0.9, 49% were female, and 72% were stunted. At the beginning of Cycle 2 the average child age was 36.9±0.9, 49% were female and 69% were stunted. Child demographic data presented here relates to the subset of children who received treatment in each Cycle, which differ slightly, but not significantly, from the total population reported elsewhere [40].

4.4.3 Reinfection dynamics.

Although our first objective had been to compare infection and reinfection dynamics among the three STH infections over 2 consecutive reinfection cycles, the absence of FLOTAC data during Cycle 1 limited our analysis only to Ascaris.

Ascaris was detected in 20% of children prior to ABZ treatment in Cycle 1 (Figure 1 A) with relatively low mean intensity (Figure 1 B). Treatment with ABZ

119

had an Ascaris ERR of 100% in the subsample for whom drug efficacy was assessed. Within 3 mo the prevalence of Ascaris in those who had received treatment had increased to 8%, half that recorded at baseline; by 9 mo, 34% of the children were infected and intensity of reinfection was three times higher than baseline values. At the baseline of Cycle 2, 19% of all individuals who provided samples were infected. ABZ efficacy against Ascaris was 97%. Despite the low prevalence of Ascaris at the end of Cycle 2 (11%), the intensity had increased to pre-treatment levels (Figure 1 B).

Hookworm prevalence was 5% at the baseline of Cycle 1 (Figure 1 C), with a low average intensity (Figure 1 D). Drug treatment with ABZ eliminated infection in the 3 infected children. By 3 mo post treatment, hookworm prevalence (Figure 1 C) and intensity (Figure 1 D) were similar to baseline levels and by 9 mo post treatment both metrics had exceeded baseline levels. In Cycle 2, ERR was 89% and both prevalence and intensity had returned to pre- treatment levels within 4 mo of treatment.

Trichuris prevalence was only 1% at the baseline of Cycle 1 (Figure 1 E) and therefore it was not possible to calculate ERR. Three months after treatment, the prevalence remained very low, but by 9 mo post treatment, Trichuris prevalence (Figure 1 E) and intensity (Figure 1 F) reached the highest values detected during the study. A single dose of ABZ at the beginning of Cycle 2 led to only a 40% ERR for Trichuris. The 4 and 6 mo post-treatment prevalence and intensity for Trichuris did not differ from Cycle 2 baseline values.

4.4.4 Factors associated with spatial clusters of infection and with individual reinfection intensity.

Ascaris lumbricoides. No high prevalence clusters were identified for Ascaris infection, indicating the widespread occurrence of infection throughout the region. At the individual child level, univariate comparison between infected 120

and uninfected preschool children at the end of both Cycle 1 and Cycle 2 showed that infected preschool children had higher pre-treatment infection burdens (Table 4). Consistent with this, there were positive correlation between baseline and reinfection epgs (Cycle 1: r=0.35, p<0.001; Cycle 2: r=0.24, p=0.002). No other differences were detected in the univariate analyses.

Composite models emerging from the multivariate analysis of Cycle 1 reinfection epg confirmed that pre-treatment infection burden was a risk factor for higher reinfection intensity and identified two additional individual level factors. Low HAZ was associated with a higher reinfection egp whereas latrine use was associated with a lower reinfection epg. Among the household factors, higher HWI was associated with lower Ascaris reinfection (Table 5). In reinfection Cycle 2, individual factors (higher pre-treatment epg, younger age, female, not using a latrine) were associated with increased reinfection epg. At the household level, reinfection epg was higher in households where mothers had less education, and surprisingly, in households with a greater HWI (Table 5).

Hookworm. A region of high prevalence incorporating 23% of households (Log likelihood ratio 9.97 p=0.01) was detected at baseline of Cycle 2 (Figure 2 A). At 6 months post treatment in Cycle 2, 41% of the homes in the high prevalence cluster had an infected child. The high prevalence cluster was characterized by lower HWI, lower household density, less access to physical infrastructure and lower maternal education compared with households outside the cluster (Table 2). Multivariate models confirmed that low household density, was a risk factor for residing in a high infection cluster (Table 3).

At the individual child level, univariate analysis revealed that those who were infected at the end of Cycle 2 were more likely to live in a high prevalence cluster and had heavier baseline infection burdens than uninfected children (Table 4). Spearman rank correlation confirmed the correlation between

121

baseline and reinfection epgs (r=0.35, p<0.001). After multivariate analysis, at the end of Cycle 2, only individual level factors influenced reinfection epg. Children who had a low HAZ score were most heavily infected and might have had a higher pre treatment epg (p=0.06) (Table 5).

Trichuris trichiura. As with hookworm, a cluster of high Trichuris prevalence (25% of participant households: Log likelihood ratio=11.8928.40 p=0.002) was detected at baseline of Cycle 2 (Figure 2 B). At 6 months post treatment 38% of households in the high prevalence cluster had an infected child. The cluster of high infection prevalence was characterized by lower household density, lower HWI, less access to physical infrastructure and lower maternal education than households outside the high prevalence cluster (Table 2). Multivariate analysis confirmed that the high prevalence cluster was less densely populated and households had a lower HWI (Table 3).

Given the low ABZ efficacy of a single dose of ABZ against Trichuris, and given that only 18 children received a second dose of ABZ, the 4 and 6 mo prevalence and intensity values in Cycle 2 likely reflect continuing Trichuris infection rather than rapid reinfection. Therefore univariate and multivariate analysis of individual reinfection in Cycle 2 was not conducted.

4.5 Discussion Our study builds on past work in this rural indigenous area of Panamá and further characterizes the epidemiology of STH infections. As shown by previous empirical studies [51,52], theoretical studies [8] and a recent meta-analysis [12] of STH infection dynamics, we found that prevalence and intensity of infection returned to baseline levels following treatment. Our use of spatial analysis uncovered distinctions among STH infections. Whereas spatial clusters of Trichuris and hookworm were detected, and overlapped in a region 122

characterized by poor development (as measured by low HWI and low household density) no spatial clustering was detected for Ascaris, indicating its more homogenous dispersion throughout the region. Our multivariate regression models revealed differences between Ascaris and hookworm in the relative impact of household and individual factors in driving reinfection dynamics. Neither regional nor household variables emerged in the model for hookworm, and after controlling for baseline epg, the only individual factor that emerged was HAZ. Reinfection models for Ascaris also included individual factors (baseline epg, HAZ, age, sex) in one or both reinfection cycles, as well as two household factors, maternal education and HWI. Interestingly, in Cycle 1, reinfection rates were higher in households with low HWI whereas the opposite pattern was seen in Cycle 2.

Our first objective was to compare reinfection dynamics among the three STHs. Due to differences in the sensitivity of FLOTAC and Kato Katz for hookworm and Trichuris infection intensity estimates, we have focused our comparison between reinfection cycles primarily on Ascaris, for which both diagnostic methods had similar sensitivity. For Ascaris, the reinfection profile was similar in both years, with epgs reaching pre-treatment levels within 6-9 mo. In contrast, our data indicate that reinfection with hookworm and Trichuris, measured as the change in epg, was more rapid during Cycle 1 between 3 and 9 mo post-treatment than during Cycle 2. Four factors may have contributed to this difference. First, Cycle 1 included the wet season (October – January) whereas Cycle 2 overlapped only briefly with the wet season. Indeed, more wet days [53] and soil wetness [54] are considered conducive to Ascaris transmission and hookworm larval abundance. Second, the reinfection period was longer for Cycle 1 (9 mo) than Cycle 2 (6 mo). Third, the Conditional Transfer programs in the region had incorporated anthelmintic delivery through school based programs for at least 1 year prior to this study. A slower rate of reinfection in

123

subsequent treatment cycles is characteristic of an area with effective control [55,56] even in the segments of the population not targeted by treatment programs [6,9]. Fourth, at least in the case of Trichuris, the low efficacy of ABZ precludes us from assuming that the 6 mo epg at the end of Cycle 2 is driven by reinfection. Over time, epgs reach an equilibrium determined by the Basic Reproduction Ratio of the parasite and the host and environmental characteristics of the region thus Trichuris epgs after ABZ treatment may have been close to this equilibrium. Of further note is that the three STH species did not demonstrate the expected reinfection dynamics. Based on the longevity of the STH species, Ascaris and Trichuris are expected to reach baseline intensity more quickly than hookworm [51] however in our study, hookworm reinfection intensity reached baseline levels in 4 mo compared to the 6 mo that it took for Ascaris and Trichuris infections to reach pre-treatment intensity. We suggest that the lower treatment efficacy of a single dose of ABZ for hookworm (ERR=89%) compared to Ascaris (ERR=97%) and the higher baseline prevalence of hookworm infection (21%) compared to Trichuris (10%) led to increased transmission of hookworm by increasing the number of hookworm infectious stages in the environment relative to Ascaris and hookworm and thus increasing the efficiency of transmission [55].

The low efficacy of ABZ against Trichuris has been noted in other studies [57-60] and, in this Trichuris population, may be related to the genetic polymorphisms associated with benzimidazole resistance that have been detected in Trichuris eggs from our study area [61]. ABZ may not be the most appropriate drug to use for Trichuris in this region, and its continued use, even if directed towards control of Ascaris and hookworm could lead to an increase in the frequency of these polymorphisms and greater evolutionary pressure toward ABZ resistance in Trichuris.

124

Our second objective was to identify and characterize spatial clusters of high STH prevalence. Two intriguing observations emerged. First, the location of hookworm and Trichuris high prevalence clusters significantly overlapped and the clusters shared the common characteristic of low household density. In our study area, we have previously reported that low household density was associated with lower HWI, as well as being farther from roads and health centres [40], and thus we have considered low household density as an indicator of poor regional development or “remoteness”. The relationship between “remoteness” and clusters of high prevalence of infection could also be linked to characteristics of the physical environment [32,33], to high risk activities that occur in these areas [21,62], and to the lack of sanitation and hygiene infrastructure [40]. Surprisingly, residence in a high prevalence cluster did not emerge as a risk factor for hookworm reinfection intensity. This could be explained because the higher isolation of houses in the high prevalence cluster might reduce contact of children with hookworm larvae in neighbouring homes, compared with more densely spaced homes. Alternatively, by controlling for baseline epg, we may have reduced the influence of the regional scale clustering of infection. Second, no high prevalence clusters of Ascaris were detected. Ascaris was the most prevalent parasite in our study area. The long survival and the stickiness of Ascaris eggs [51] may facilitate their dispersion through the environment leading to more evenly distributed egg exposure, and may explain the generalized infection throughout the regions.

Our final objective was to compare the relative contribution of individual, household and regional factors in the transmission dynamics of Ascaris and hookworm. Predisposition to heavy infection is characteristic of STH infection and has been noted throughout the developing world [44,45,63,64]. Our multiple regression analyses on epg of individual children further supported these findings as pre-treatment epg was a strong predictor of reinfection

125

intensity for Ascaris and may have also influenced hookworm reinfection epg. Mechanisms to explain predisposition include the influence of individual traits on susceptibility such as genetic differences in immunity [65] or nutritional status [10], as well as household factors that increase exposure to infection [38].

After controlling for baseline epg, HAZ emerged as an individual factor in models of reinfection intensity for hookworm (Cycle 2) and Ascaris (Cycle 1). Most longitudinal studies examining STH infection and anthropometric outcomes have focused on the potential benefit of anthelmintic treatment on weight or height gain [66,67] rather than the impact of undernutrition on child susceptibility to infection or reinfection. The few studies that have specifically examined the latter relationship have either not controlled for potential confounding factors [14] or have found that the increased rates of reinfection in undernourished children are no longer significant when controlling for maternal literacy, income and latrine access [15,16]. We found, however, that after accounting for poverty related factors (maternal education, HWI, latrine use) that may influence child height-for-age, children who were shorter for their age (a sign of chronic malnutrition) became more heavily reinfected and thus may be more susceptible to hookworm and Ascaris infection. The fact that HAZ was associated with Ascaris reinfection in Cycle 1 but not Cycle 2 was an intriguing finding, as was the finding that younger (not older children) had higher epgs, and that children from less poor homes also had higher epgs. It is possible that the relationship between stunting and reinfection intensity is easier to detect in younger children, given that intensity of infection increases rapidly with age [51,52,68]. Child latrine use was also associated with a lower reinfection burden, demonstrating the importance of sanitation for reducing infection and transmission, as has been shown previously [16,21].

In addition to individual level variables, household factors also contributed to Ascaris reinfection. Household risk factors commonly associated 126

with poverty (low maternal education, low HWI) were related to increased reinfection intensity in either or both reinfection Cycles. Indeed, low maternal education [21,25,26] is commonly associated with greater infection burdens, likely due to poor home sanitation and hygiene practices as well as a reduced use of health services [69]. Interestingly, household poverty was associated with increased infection burden in Cycle 1 but a decreased infection intensity in Cycle 2. Examining the spatial location of infected households at the end of Cycle 2, we determined that the infections were primarily in an area of greater relative wealth. Thus it is possible that risk of infection was greater in that area and it wasn’t HWI per se that influenced Ascaris reinfection dynamics in Cycle 2. Of further note is that although lower household poverty was associated with greater reinfection burden, other poverty related variables that may be more directly linked to transmission (latrine use, maternal education) still demonstrated the expected relationship with infection burden.

It is important to recognize that this study had a few limitations. First, we do not have data on the history of treatment prior to the baseline of Cycle 1 although we know ABZ was available in the area. Second, seasonal work-related migration reduced the number of preschool children available at the end of Cycle 1. Fortunately many children were at home three weeks later. Hence the 3 wk mean epg in Cycle 2 incorporates data from children who had just received Albendazole and children who had not been treated. Also, we were able to provide ABZ treatment at 3 wks to those who were infected but logistics prevented us from confirming the ERR following this treatment. We assume that the available ERR data can be extrapolated to all the children, and therefore that hookworm epgs were reduced significantly after the additional children were treated at 3 wks. Third, a single ABZ treatment did not remove all Trichuris and hookworm parasites. For the reasons noted above, we were unable to determine whether the second dose of ABZ successfully cleared infection.

127

Fourth, we likely underestimated Trichuris and hookworm prevalence and intensity in Cycle 1 because we did not use the FLOTAC technique. Although this limited our ability to compare reinfection dynamics between Cycle 1 and Cycle 2, we were still able to comment on the period of most rapid transmission by verifying the trends observed with Kato Katz data for Cycle 2 (data not shown). The observed greater sensitivity of FLOTAC than duplicate Kato Katz thick smears for low intensity infections characteristic of hookworm and Trichuris has been recorded previously in validation studies [70-72] and is believed to be due to the increased volume of sample used in the FLOTAC method (1g) compared to Kato Katz (41.7mg). Use of both techniques in Cycle 2 alerted us that hookworm and Trichuris were more common in the region than previously recognized and furthermore, highlighted the low ERR for Trichuris after a single treatment with ABZ. In contrast to previous studies, however, Kato Katz and FLOTAC had a similar sensitivity for Ascaris infection. FLOTAC epg estimates were consistently lower than those calculated using Kato Katz for all 3 STH. This has also been noted previously, and could indicate that the Kato Katz technique overestimates epg due to the egg concentration that may occur while sieving the fecal sample [71,72]. A potential consequence is the over estimation of drug efficacy through ERR when using Kato Katz. Previous work that compared ERR values between FLOTAC and Kato Katz in a drug trial [73] also found that FLOTAC estimates resulted in lower ERR values than Kato Katz. An additional contributing factor could be the ability to process diarrhetic samples using FLOTAC but not Kato Katz. However, we did not detect any difference in mean FLOTAC epg between diarrhetic vs non-diarrhetic samples. When validated these findings have implications for the monitoring of preventive control programs, in particular for the perceived success of chemotherapy as well as early detection of drug resistance. Finally, it is likely that the absence of clusters in Cycle 1 was due to the low prevalence of infection in that cycle (Hookworm: Cycle 1=5%, Cycle

128

2=22%; Trichuris: Cycle 1=1%, Cycle 2=10%) which limited our ability to detect clusters.

Our study of STH reinfection dynamics in the comarca Ngäbe Buglé of Western Panamá has emphasized that even within regions of extreme poverty, clusters of STH infections exist and that transmission is related to household level exposure variables as well as individual factors that may influence susceptibility. Our results have specific implications for public health interventions. First, the lower treatment efficacy of ABZ for Trichuris together with high infection levels during Cycle 2 calls attention to the importance of monitoring drug efficacy, especially against Trichuris, as well as the possibility of using multiple treatments or an alternative anthelmintic. Second, improving both the regional and household level sanitation and hygiene environment will be necessary to further reduce STH transmission. This will be especially important for the growth and development of stunted children who may be more susceptible to hookworm and Ascaris infection. Taken together, comprehensive control programs that combine short term morbidity control with the development of long term economic capacity, sanitation infrastructure and improved food security are necessary to make lasting improvements in child health in the comarca Ngäbe Buglé.

4.6 Acknowledgements The authors would like to thank collaborators in the Panamanian Ministry of Health (MINSA) and the Instituto Conmemorativo Gorgas de Estudios de la Salud for logistical support and guidance during field work as well as community interviewers and study participants. FLOTAC fecal analysis equipment was kindly donated by The Regional Center for Monitoring Parasites (CREMOPAR),

129

Department of Pathology and Animal Health, Faculty of Veterinary Medicine, University of Naples Federico II, Naples, Italy.

130

4.7 References 1. Chan MS (1997) The global burden of intestinal nematode infections - Fifty years on. Parasitol Today 13: 438-443.

2. Berkman DS, Lescano AG, Gilman RH, Lopez S, Black MM (2002) Effects of stunting, diarrhoeal disease, and parasitic infection during infancy on cognition in late childhood: a follow-up study. Lancet 359: 564-571.

3. Oberhelman RA, Guerrero ES, Fernandez ML, Silio M, Mercado D, et al. (1998) Correlations between intestinal parasitosis, physical growth, and psychomotor development among infants and children from rural Nicaragua. Am J Trop Med Hyg 58: 470-475.

4. Stephenson LS, Crompton DWT, Latham MC, Schulpen TWJ, Nesheim MC, et al. (1980) Relationships between Ascaris infection and growth of malnourished preschool children in Kenya. Am J Clin Nutr 33: 1165-1172.

5. Stephenson LS, Latham MC, Ottesen EA (2000) Malnutrition and parasitic helminth infections. Parasitology 121: S23-S38.

6. WHO (2006) Preventative chemotherapy in human helminthiasis - Coordinated use of anthelminthic drugs in control interventions: a manual for health professionals and programme managers. Geneva: World Health Organization. 74 p.

7. Montresor A, Crompton DW, Gyorkos T, Savioli L (2002) Helminth control in school-age children: A guide for managers of control programmes. Geneva: World Health Organization. 78 p.

8. Anderson RM, May RM (1982) Population dynamics of human helminth infections: Control by chemotherapy. Nature 297: 557-563.

9. Bundy DAP, Wong MS, Lewis LL, Horton J (1990) Control of geohelminths by delivery of targeted chemotherapy through schools. Trans R Soc Trop Med Hyg 84: 115-120.

10. Koski KG, Scott ME (2001) Gastrointestinal nematodes, nutrition and immunity: Breaking the negative spiral. Annu Rev Nutr 21: 297-321.

11. Holland CV (2009) Predisposition to Ascariasis: Patterns, mechanisms and implications. Parasitology 136: 1537-1547.

12. Jia T-W, S M, Utzinger J, King C, Zhou XN (2012) Soil-transmitted helminth reinfection after drug treatment: A systematic review and meta-analysis. PLoS Neglect Trop Dis 6: e1621.

131

13. Williams-Blangero S, VandeBerg JL, Subedi J, Jha B, Corrêa-Oliveira R, et al. (2008) Localization of multiple quantitative trait loci influencing susceptibility to infection with Ascaris lumbricoides. J Infect Dis 197: 66- 71.

14. Hagel I, Lynch NR, Di Prisco MC, Ṕerez M, Sánchez JE, et al. (1999) Helminthic infection and anthropometric indicators in children from a tropical slum: Ascaris reinfection after anthelmintic treatment. J Trop Pediatr 45: 215- 220.

15. Saldiva SRM, Carvalho HB, Castilho VP, Struchiner CJ, Massad E (2002) Malnutrition and susceptibility to enteroparasites: Reinfection rates after mass chemotherapy. Paediatr Perinat Epidemiol 16: 166-171.

16. Hesham Al-Mekhlafi M, Surin J, Atiya AS, Ariffin WA, Mohammed Mahdy AK, et al. (2008) Pattern and predictors of soil-transmitted helminth reinfection among aboriginal schoolchildren in rural Peninsular Malaysia. Acta Trop 107: 200-204.

17. Luoba AI, Geissler PW, Estambale B, Ouma JH, Alusala D, et al. (2005) Earth- eating and reinfection with intestinal helminths among pregnant and lactating women in western Kenya. Trop Med Int Health 10: 220-227.

18. Saathoff E, Olsen A, Kvalsvig JD, Geissler WP (2002) Geophagy and its association with geohelminth infection in rural schoolchildren from northern KwaZulu-Natal, South Africa. Trans R Soc Trop Med Hyg 96: 485- 490.

19. Bartram J, Cairncross S (2010) Hygiene, sanitation, and water: forgotten foundations of health. PLoS Med 7: e1000367.

20. Chongsuvivatwong V, Pas-Ong S, McNeil D, Geater A, Duerawee M (1996) Predictors for the risk of hookworm infection: Experience from endemic villages in southern Thailand. Trans R Soc Trop Med Hyg 90: 630-633.

21. Traub RJ, Robertson ID, Irwin P, Mencke N, Thompson RCA (2004) The prevalence, intensities and risk factors associated with geohelminth infection in tea-growing communities of Assam, India. Trop Med Int Health 9: 688-701.

22. Stothard JR, Imison E, French MD, Sousa-Figueiredo JC, Khamis IS, et al. (2008) Soil-transmitted helminthiasis among mothers and their preschool children on Unguja Island, Zanzibar with emphasis upon ascariasis. Parasitology 135: 1447-1455.

132

23. Pullan RL, Bethony JM, Geiger SM, Cundill B, Correa-Oliveira R, et al. (2008) Human helminth co-infection: Analysis of spatial patterns and risk factors in a Brazilian community. PLoS Neglect Trop Dis 2: e352.

24. De Souza EA, Da Silva-Nunes M, Dos Santos Malafronte R, Muniz PT, Cardoso MA, et al. (2007) Prevalence and spatial distribution of intestinal parasitic infections in a rural Amazonian settlement, Acre State, Brazil. Cad Saude Publica 23: 427-434.

25. Naish S, McCarthy J, Williams GM (2004) Prevalence, intensity and risk factors for soil-transmitted helminth infection in a South Indian fishing village. Acta Trop 91: 177-187.

26. Holland CV, Taren DL, Crompton DWT, Nesheim MC, Sanjur D, et al. (1988) Intestinal helminthiases in relation to the socioeconomic environment of Panamanian children. Soc Sci Med 26: 209-213.

27. Forrester JE, Scott ME, Bundy DAP, Golden MHN (1988) Clustering of Ascaris lumbricoides and Trichuris trichiura infections within households. Trans R Soc Trop Med Hyg 82: 282-288.

28. Chan L, Bundy DAP, Kan SP (1994) Aggregation and predisposition to Ascaris lumbricoides and Trichuris trichiura at the familial level. Trans R Soc Trop Med Hyg 88: 46-48.

29. Brooker S, Kabatereine NB, Tukahebwa EM, Kazibwe F (2004) Spatial analysis of the distribution of intestinal nematode infections in Uganda. Epidemiol Infect 132: 1065-1071.

30. Koroma JB, Peterson J, Gbakima AA, Nylander FE, Sahr F, et al. (2010) Geographical distribution of intestinal schistosomiasis and soil- transmitted helminthiasis and preventive chemotherapy strategies in Sierra Leone. PLoS Neglect Trop Dis 4: e891.

31. Raso G, Vounatsou P, Gosoniu L, Tanner M, Goran E, et al. (2006) Risk factors and spatial patterns of hookworm infection among schoolchildren in a rural area of western Cote d'Ivoire. Int J Parasitol 36: 201-210.

32. Saathoff E, Olsen A, Kvalsvig JD, Appleton CC, Sharp B, et al. (2005) Ecological covariates of Ascaris lumbricoides infection in schoolchildren from rural KwaZulu-Natal, South Africa. Trop Med Int Health 10: 412-422.

33. Knopp S, Mohammed KA, Khamis IS, Mgeni AF, Stothard JR, et al. (2008) Spatial distribution of soil-transmitted helminths, including Strongyloides stercoralis, among children in Zanzibar. Geospat Health 3: 47-56.

133

34. Steinmann P, Zhou X-N, Li Y-L, Li H-J, Chen S-R, et al. (2007) Helminth infections and risk factor analysis among residents in Eryuan county, Yunnan province, China. Acta Trop 104: 38-51.

35. Knopp S, Stothard JR, Rollinson D, Mohammed KA, Khamis IS, et al. (2011) From morbidity control to transmission control: time to change tactics against helminths on Unguja Island, Zanzibar. Acta Trop: 10.1016/j.actatropica.2011.1004.1010.

36. Pullan RL, Kabatereine NB, Quinnell RJ, Brooker S (2010) Spatial and genetic epidemiology of hookworm in a rural community in Uganda. PLoS Neglect Trop Dis 4: e713.

37. Cundill B, Alexander N, Bethony JM, Diemert D, Pullan RL, et al. (2011) Rates and intensity of re-infection with human helminths after treatment and the influence of individual, household, and environmental factors in a Brazilian community. Parasitology 138: 1406-1416.

38. Walker M, Hall A, Basáñez MG (2011) Individual predisposition, household clustering and risk factors for human infection with Ascaris lumbricoides: New epidemiological insights. PLoS Neglect Trop Dis 5: e1047

39. Ministerio de Economía y Finanzas (2010) La Distribución del Ingreso en Los Hogares de Panamá:Encuesta de Niveles de Vida 2008. Panamá: Ministerio de Economía y Finanzas. 34 p.

40. Halpenny CM, Koski KG, Valdés VE, Scott ME (2012) Prediction of child health by household density and asset-based indices in impoverished Indigenous villages in rural Panamá. Am J Trop Med Hyg 86: 280-291.

41. Hotez PJ, De Silva N, Brooker S, Bethony J (2003) Soil transmitted helminth infections: The nature, causes and burden of the condition, Working Paper No. 3, Disease Control Priorities Project. Bethesda, Maryland: Fogarty International Center, National Institutes of Health.

42. Montresor A, Crompton DW, Bundy D, Hall A, Savioli L (1998) Guidelines for the evaluation of soil transmitted helminthiasis and schistosomiasis at community level. . Geneva: WHO. 49 p.

43. Cringoli G (2006) FLOTAC, a novel apparatus for a multivalent faecal egg count technique. Parassitologia 48: 381-384.

44. Forrester JE, Scott ME, Bundy DAP, Golden MHN (1990) Predisposition of individuals and families in Mexico to heavy infection with Ascaris lumbricoides and Trichuris trichiura. Trans R Soc Trop Med Hyg 84: 272- 276. 134

45. Kightlinger LK, Seed JR, Kightlinger MB (1995) The epidemiology of Ascaris lumbricoides, Trichuris trichiura, and hookworm in children in the Ranomafana rainforest, Madagascar. J Parasitol 81: 159-169.

46. Payne L, Koski KG, Ortega-Barria E, Scott ME (2007) Benefit of vitamin A supplementation on Ascaris re-infection is less evident in stunted children. J Nutr 137: 1455-1459.

47. WHO Multicentre Growth Reference Study Group (2006) WHO Child Growth Standards: Methods and development: Length/height-for-age, weight- for-age, weight-for-length, weight-for-height and body mass index-for- age: Methods and development. Geneva:

48. Kuldorff M, Nagarwalla N (1995) Spatial disease clusters: detection and inference. Stat Med 14: 799-819.

49. Kuldorff M (2011) SaTScan v9.0: Software for the spatial and space-time scan statistics. Information Management Services, Inc. http://www.satscan.org

50. Sacko M, De Clercq D, Behnke JM, Gilbert FS, Dorny P, et al. (1999) Comparison of the efficacy of mebendazole, albendazole and pyrantel in treatment of human hookworm infections in the Southern Region of Mali, West Africa. Trans R Soc Trop Med Hyg 93: 195-203.

51. Anderson RM, May RM (1991) Infectious diseases of Humans: Dynamics and Control. Oxford: Oxford University Press. 757 p.

52. Bundy DAP, Thompson DE, Cooper ES, Golden MHN, Anderson RM (1985) Population dynamics and chemotherapeutic control of Trichuris trichiura infection of children in Jamaica and St. Lucia. Trans R Soc Trop Med Hyg 79: 759-764.

53. Gunawardena GSA, Karunaweera ND, Ismail MM (2004) Wet-days: Are they better indicators of Ascaris infection levels? J Helminthol 78: 305-310.

54. Nwosu ABC, Anya AO (1980) Seasonality in human hookworm infection in an endemic area of Nigeria, and its relationship to rainfall. Tropenmed Parasitol 31: 201-208.

55. Gabrielli AF, Montresor A, Chitsulo L, Engels D, Savioli L (2011) Preventive chemotherapy in human helminthiasis: Theoretical and operational aspects. Trans R Soc Trop Med Hyg 105: 683-693.

56. Albonico M, Stoltzfus RJ, Savioli L, Chwaya HM, D'Harcourt E, et al. (1999) A controlled evaluation of two school based anthelminthic chemotherapy

135

regimens on intensity of intestinal helminth infections. Int J Epidemiol 28: 591-596.

57. Albonico M, Allen H, Chitsulo L, Engels D, Gabrielli AF, et al. (2008) Controlling soil-transmitted helminthiasis in pre-school-age children through preventive chemotherapy. PLoS Neglect Trop Dis 2: e126.

58. Keiser J, Utzinger J (2008) Efficacy of current drugs against soil-transmitted helminth infections: Systematic review and meta-analysis. J Am Med Ass 299: 1937-1948.

59. Bennett A, Guyatt H (2000) Reducing intestinal nematode infection: Efficacy of albendazole and mebendazole. Parasitol Today 16: 71-74.

60. Vercruysse J, Behnke JM, Albonico M, Ame SM, Angebault C, et al. (2011) Assessment of the anthelmintic efficacy of albendazole in school children in seven countries where soil-transmitted helminths are endemic. PLoS Neglect Trop Dis 5: e948.

61. Diawara A, Drake LJ, Suswillo RR, Kihara J, Bundy DAP, et al. (2009) Assays to detect β-tubulin codon 200 polymorphism in Trichuris trichiura and Ascaris lumbricoides. PLoS Neglect Trop Dis 3: e397.

62. Matthys B, Tschannen AB, Tian-Bi NT, Comoé H, Diabaté S, et al. (2007) Risk factors for Schistosoma mansoni and hookworm in urban farming communities in western Côte d'Ivoire. Trop Med Int Health 12: 709-723.

63. Stothard JR, French MD, Khamis IS, Basáñez MG, Rollinson D (2009) The epidemiology and control of urinary schistosomiasis and soil-transmitted helminthiasis in schoolchildren on Unguja Island, Zanzibar. Trans R Soc Trop Med Hyg 103: 1031-1044.

64. Quinnell RJ, Pullan RL, Breitling LP, Geiger SM, Cundill B, et al. (2010) Genetic and household determinants of predisposition to human hookworm infectionin a Brazilian community. J Infect Dis 202: 954-961.

65. Quinnell RJ (2003) Genetics of susceptibility to human helminth infection. Int J Parasitol 33: 1219-1231.

66. Stephenson LS, Latham MC, Adams EJ, Kinoti SN, Pertet A (1993) Weight gain of Kenyan school children infected with hookworm, Trichuris trichiura and Ascaris lumbricoides is improved following once- or twice- yearly treatment with albendazole. J Nutr 123: 656-665.

67. Hlaing T, Toe T, Saw T, Kyin ML, Lwin M (1991) A controlled chemotherapeutic intervention trial on the relationship between Ascaris 136

lumbricoides infection and malnutrition in children. Trans R Soc Trop Med Hyg 85: 523-528.

68. Bundy DAP, Cooper ES, Thompson DE, Didier JM, Simmons I (1987) Epidemiology and population dynamics of Ascaris lumbricoides and Trichuris trichuria infection in the same community. Trans R Soc Trop Med Hyg 81: 987-993.

69. LeVine RA, Rowe ML (2009) Maternal literacy and child health in less- developed countries: evidence, processes, and limitations. J Dev Behav Pediatr 30: 340-349.

70. Knopp S, Glinz D, Rinaldi L, Mohammed KA, N'Goran EK, et al. (2009) FLOTAC: A promising technique for detecting helminth eggs in human faeces. Trans R Soc Trop Med Hyg 103: 1190-1194.

71. Utzinger J, Rinaldi L, Lohourignon LK, Rohner F, Zimmermann MB, et al. (2008) FLOTAC: a new sensitive technique for the diagnosis of hookworm infections in humans. Trans R Soc Trop Med Hyg 102: 84-90.

72. Glinz D, Silué KD, Knopp S, Lohourignon LK, Yao KP, et al. (2010) Comparing diagnostic accuracy of Kato-Katz, Koga Agar Plate, Ether-Concentration, and FLOTAC for Schistosoma mansoni and soil-transmitted helminths. PLoS Neglect Trop Dis 4: e754

73. Knopp S, Speich B, Hattendorf J, Rinaldi L, Mohammed KA, et al. (2011) Diagnostic accuracy of Kato-Katz and FLOTAC for assessing anthelmintic drug efficacy. PLoS Neglect Trop Dis 5: e1036.

137

Table 1. Sample sizes for two STH reinfection cycles among Panamanian preschool children.

Baseline Received 3 mo/4 mo 9 mo/6 mo 2wk/3wk epg epg Treatment reinfection epg reinfection epg

Cycle 1 215 209 100 87 155

Cycle 2 270 279 222 218 200

Reinfection epg sample sizes only include individuals who also received treatment. Cycle 1 Baseline, 2 wk and 3 mo epg estimates were calculated using Kato Katz methodology. Cycle 1 9 mo reinfection epg and all Cycle 2 epg estimates were calculated using FLOTAC. 138

138

Table 2. Comparison of characteristics between households within and outside the high prevalence clusters.1

Hookworm Trichuris

In Out In Out

(n=58) (n=195) (n=63) (n=190)

Household Factors

Household Density, 13±1 41±3** 13±1 42±3** 2 Wealthh /k Index, HWI 2 -0.04±0.06 0.27±0.05* -0.08±0.06 0.30±0.05* Latrine Access, % 18(10-30) 35 (28-42)* 16(9-27) 36(29-43)* Aqueduct Access, % 16(8-27) 40(33-47) * 14(7-25) 41(34-48)** * *

139 Mother's Education, yrs 2.9±0.4 4.1±0.3 2.8±0.4 4.2±0.3

# People / Room 4.7±0.3 5.6±0.2 5.2±0.3 5.4±0.2 1 Summary statistics presented are mean ±SEM or % (95% CL). 2 Asset based index, weights derived from the first component of Principle Components Analysis. *p<0.05; **p<0.001

139

Table 3. Final multiple logistic regression models predicting household presence within Cycle 2 high prevalence clusters.1,2 Hookworm Trichuris

Household Factors

Household Density, 0.95 (0.93 – 0.97) 0.96 (0.93-0.98) houses/km2

Wealth Index, HWI3 NE4 0.53 (0.28-0.98)

Model Statistics

n 231 231

Χ2 38.84 47.40

P <0.001 <0.001

1 Odds ratios ± 95% CI 2 Latrine access and maternal education did not enter any model. 3Asset based index, weights derived from the first component of Principle Components Analysis 4 NE=Variable was excluded during the stepwise process (p>0.10).

140

Table 4. Comparison of regional, household and individual factors influencing STH reinfection of Panamanian preschool children.1

Cycle 1 Cycle 2

Ascaris Ascaris Hookworm n Infected Uninfected n Infected Uninfected Infected Uninfected Regional Factors

Residence in high prevalence - - 144 - - 41(25-58) 20(14-29)* cluster, % Household Factors

Household Density, km2 131 45±9 31±3 144 31±9 35±3 39±7 33±3 Wealth Index, HWI2 128 0.17±0.13 0.20±0.07 140 -0.06±0.16 0.16±0.06 0.06±0.1 0.16±0.06 Mother's Education, yrs 131 3.1±0.7 3.5±0.3 143 2.1±0.8 3.7±0.33 3.5±0.7 3.5±0.3 14 1 # People/Room 123 5.3±0.6 5.2±0.3 143 5.5±0.6 5.5±0.2 5.8±0.5 5.5±0.3

Individual Factors

Cycle 1 Baseline Infection, epg 122 10463±3119 1737±478** 172 2895±1549 612±193* 91±39 4±2** Age, mo 155 32±2 31±1 200 22±3 27±1 29±2 25±1 Female, % 155 48(33-65) 52(44-61) 200 45(27-65) 52(45-59) 47(33-61) 53(45-61) Height for age

Z score 130 -2.8±0.2 -2.6±0.1 170 -2.6±0.2 -2.4±0.1 -2.6±0.2 -2.4±0.08 Stunting, % 130 84(67-93) 70(60-78) 170 78(54-92) 70(63-77) 77(61-88) 70(61-77) Latrine use, % 135 9(2-24) 18(11-26) 185 10(2-31) 15(10-21) 12(5-26) 15(10-22) 1 Summary statistics presented are mean ±SEM or % (95% CL). 2 Asset based index, weights derived from the first component of Principle Components Analysis. *p<0.05; **p<0.001

141

Table 5 – Negative binomial regression models of Ascaris and hookworm reinfection intensity in Panamanian preschool children.

Ascaris Ascaris Hookworm Cycle 1 Cycle 2 Cycle 2 9 mo 6 mo 6 mo IRR1 IRR IRR

Regional Factors In High Cluster, baseline Cycle 22 NA3 NA NE

Household Factors Household density, homes/km2 NE4 0.96 (0.93 – 1.001) NE Household wealth, HWI 0.19 (0.08-0.48)** 145 (7-3012)* NE Mother's education, yrs NE 0.48 (0.28-0.80)* NE Individual Factors

Age, mo NE 0.93 (0.88-0.99)* NE Sex5 NE 27 (2-303)* NE 142 1.0001 1.0003 1.004 Baseline Infection, epg (1.0000-1.0002)** (1.0001-1.0006)* (1.00-1.009) Height-for-Age, Z score 0.15 (0.07-0.32)** NE 0.49 (0.29-0.84)* 0.0001 Latrine use6 0.007 (0.001-0.04)** NE (6.9 e-8 – 0.05)* Model Statistics N 100 140 140 P <0.001 <0.001 <0.001 Wald chi2 104.74 55.95 10.49 Alpha7 28.3(18.0 -44.5) 88.4 (50.9-153.4) 40.59 (25.89-63.6)

142

1Estimated rate of increase in dependent variable for 1 unit increase in the independent variable. 2Household within identified high prevalence cluster of infection by SaTScan Spatial Scan (0=no, 1=yes). 3 NA = Not applicable 4NE = Variable was excluded during the stepwise process (p>0.10). 50=boy, 1=girl 60=no, 1=yes 7Alpha statistic indicates the degree of overdispersion in the data. *p<0.05, **p<0.001 Variables that did not enter any model: Household crowding. 143

143

Figure 1. Prevalence and intensity for Ascaris, hookworm and Trichuris in Panamanian preschool children.

Prevalence (A, C, E) and intensity (B, D, F) for Ascaris (A,B), hookworm (C,D) and Trichuris (E,F) were assessed by Kato Katz (dark grey bar – left vertical axis) and FLOTAC (light grey bar – right vertical axis) during two reinfection cycles. Different letters indicate significant differences (P < 0.05) among sample periods (lower case Kato Katz comparisons; upper case FLOTAC comparisons). Solid vertical bar separates Kato Katz from FLOTAC data. A single dose of Albendazole (200 mg: 1-2 yrs, 400 mg: 3-5 yrs) was delivered after the baseline samples in each cycle and children who remained infected after treatment in Cycle 2 received a second dose. Data presented consider each cycle independently resulting in a different sample group for the 9 mo reinfection point of Cycle 1 and the baseline of Cycle 2. Analyses include children who received treatment at the baseline of both or either reinfection cycle.

144

145

Figure 2. Spatial clusters of households with high prevalence of hookworm and Trichuris infection.

High prevalence clusters (dotted line) for hookworm (A) and Trichuris (B) infection detected at Baseline of Cycle 2. Uninfected (small grey dots) and infected households (large dark dots) based on data from Cycle 2 at 6 mo (hookworm and Trichuris).

145

Connecting Statement II

In Chapter 4, I demonstrated that regional clustering of STH infections in preschool children was primarily driven by poverty and remoteness whereas reinfection was related to regional, household and individual traits that presumably influenced exposure and susceptibility. Furthermore, there were species specific differences in the risk factors for reinfection. Trichuris reinfection was more strongly related to regional prevalence of infection and household exposure variables, whereas stunted children were more susceptible to Ascaris and hookworm reinfection. Thus in Chapters 3 and 4 I have shown how chronic protozoan infection increases stunting which in turn increases susceptibility to Ascaris and hookworm reinfection. In Chapter 5, I consider whether food and macronutrient consumption may further influence preschool child linear growth as well as weight gain. Furthermore, I examine how type of conditional transfer program (food voucher or cash transfer) influenced preschool child food consumption patterns and whether the predictors of child growth differed between these two program variants

146

CHAPTER 5

Comparison of preschool child diet and anthropometry between a cash transfer and food voucher conditional transfer programs in rural Panamá

Carli M. Halpenny1, Marilyn E. Scott1, Kristine G. Koski2

1Institute of Parasitology and McGill School of Environment, 2School of Dietetics and Human Nutrition Macdonald Campus of McGill University, Ste-Anne de Bellevue, Quebec, Canada,

147

5.1 Abstract Background: Poverty reduction strategies such as Conditional Transfer Programs (CTP) aim to improve child nutritional status through health and education programs while improving the economic capacity of participants through regular cash transfers or food vouchers. CTPs have improved the quality of household food acquisition and child nutritional status but few studies have compared the cash transfer (CT) and food voucher (FV) variants of these programs and none have examined the interrelationship of food consumption and anthropometry within these programs while controlling for gastrointestinal infections. Objective: To identify, characterize and compare macronutrient intakes and food patterns in preschool children between the FV and CT programs; to determine the relative contribution of foods / food groups and of macronutrients to preschool child growth over a 9 mo period from 2008 to 2009. Methods: The study was conducted in a rural indigenous region of Panamá where over 90% of households live in extreme poverty and where 60% of preschool children are stunted. Anthropometric measures and one or two 24 hr recalls were conducted in both years for 220 preschool children. Food group and macronutrient consumption as well as dietary patterns revealed through principal component analysis were compared for preschool children from CT and FV regions. Macronutrient and food group predictors of child growth were identified using multiple linear regression models of change in height-for-age (HAZ) and weight-for-age (WAZ). Results: The basic diet of legumes, rice, sugar and coffee was supplemented in the FV region by more fruits and vegetables as well as eggs and legumes whereas, in the CT region, market products were more common. Two diet patterns were identified. The “Beans & Rice” diet was consumed throughout the 2 regions whereas the “Market” diet was concentrated in the CT region, especially close to the access road. Linear growth (change in HAZ) was similar in the FV

148

(0.18±0.10 SD) and CT (0.06±0.07 SD) regions. Carbohydrates, especially sweets and root vegetables, were detrimental to height for age (HAZ), even after controlling for infection and socio-economic status. Although the FV region had lower asset-based wealth scores and were more remote compared with CT region, preschool child weight gain (change in WAZ) was greater in the FV region (0.17±0.16 SD) than in the CT region (-0.17±0.11 SD) where market snacks (chips and sweets) were detrimental to child weight gain. Conclusion: Nutrition messages that promote the inclusion of meat and fish but reduce sweets, chips and root vegetables in preschool child diet will increase the intake of the nutrients needed for growth.

149

5.2 Introduction Child undernutrition is responsible for 35% of the deaths in children under the age of 5 and 11% of the total global disease burden [1]. Importantly, undernutrition not only contributes to child mortality, but it also has consequences for the physical [2-4] and cognitive [5,6] development, negatively affecting educational and earning potential which ultimately results in decreased human capital [7]. Thus, effective nutrition interventions have the potential not only to reduce child mortality but also to reduce poverty in affected areas.

For these reasons, poverty reduction programs have become increasingly multi-dimensional in nature, targeting maternal-child health at the same time as improving the economic capacity of participants. Conditional transfer programs in particular provide a short term economic incentive in return for participation in education and health programs that are believed to give recipients the capabilities and resources to effectively escape extreme poverty [8,9]. Aspects of conditional transfer programs that could improve child nutrition are the provision of micronutrient fortified food and supplements, increasing maternal knowledge through preventive health and nutrition education and improving food security if increased household income is used to purchase more nutritious food [10].

The effects of five conditional cash transfer programs on child nutrition were recently reviewed by Leroy et al [10]. Overall, CCT programs were found to have a strong impact on child anthropometry, especially height-for- age (HAZ). However, despite including the delivery of fortified supplements, there was little effect on child micronutrient status. Evaluation of child diets found an increased intake of animal source foods and vegetables in Colombia but in Mexico the increased quality of food bought by the household did not result in dietary improvements for preschool children. In addition to cash 150

transfers, conditional transfer programs also provide aid in the form of food vouchers however there are few reported comparisons between food-based and cash-based transfer programs. In Mexico, households participating in a food-based in-kind transfer program consumed more calories, animal source foods, cereals and legumes than those who received a cash transfer (reviewed in [10]). To the best of our knowledge, studies of diet and anthropometry outcomes within conditional transfer programs focus on household level dietary analysis, thus they have been unable to examine the links between diet and child anthropometry at the individual level among program participants.

In rural regions of Panamá, two types of conditional transfer programs were implemented under the Red de Oportunidades. The Food Voucher (FV) program provides $50/month in food vouchers and the Cash Transfer (CT) program provides $50/month in cash, both in exchange for regular visits to the health centre, attendance at educational sessions and regular school attendance for school age children. Our longitudinal study in the comarca Ngäbe Buglé of Western Panamá gave us the opportunity to examine preschool child diet and anthropometry in an impoverished, rural area where these two types of Conditional Transfer Programs had been in place for a few years. Our objectives were to: 1) identify, characterize and compare macronutrient intakes and food patterns in preschool children between the FV and CT programs; and 2) determine the relative contribution of foods / food groups and of macronutrients to preschool child growth over a 9 mo period from 2008 to 2009.

5.3 Materials and Methods 5.3.1 Study area and population. The comarca Ngäbe-Buglé is a semi-autonomous political region inhabited primarily by the Ngäbe and Buglé indigenous groups (2004 population estimate 151

of 128,978) of whom 91% live in extreme poverty (< US$ 1.75/day) [11]. The Conditional Transfer Program Red de Oportunidades provides an additional US$ 50/month in either cash or food vouchers to the over 90% of families in extreme poverty in the comarca in exchange for participation in health and education programs. Through the Food Voucher (FV) program implemented in Emplanada de Chorcha (western region of our study area) beginning in October 2005, a designated woman per household is provided with $50/mo in vouchers for purchase of a select group of foods (chicken, canned meat, canned fish, eggs, milk, legumes, rice, porridges, flour, noodles, tomato sauce, oil, salt, sugar, coffee) as well as soap and matches at the local stores. In exchange, a member of the household must participate in agricultural training, women must be up-to- date on their health checks (pregnancy and reproductive health), and children must attend school and be current on their vaccinations. Through the Cash Transfer (CT) program operating in Soloy (eastern region of our study area) since April 2006, a selected woman in each household is given $50/mo cash in exchange for similar health and education commitments. The training component of the CT program focuses primarily on preventive health and financial management education. As described previously [12], the study area was accessible through two dirt roads, the western road in the FV region having limited access during the rainy season (July—November). Household density varied considerably throughout the study area, with more densely populated regions being closer to a road and having better access to latrines, aqueducts and health facilities as well as a greater average asset based wealth score. Mothers had completed less than 4 years of education, the average child age was 26 months and half the children were female. A large proportion of the preschool children were stunted (61%) and the most common parasitic infections, Giardia sp., Ascaris, Trichuris and hookworm, had a peak prevalence of 11% - 34% over the 16 month study period [12,13]. 152

5.3.2 Ethical considerations.

Ethical approval was obtained from the Instituto Conmemorativo de Gorgas in Panamá and McGill University in Canada. Community interactions were established in accordance with the Guía para Realizar Estudios e Investigaciones en los Pueblos Indígenas de Panamá, which included participation in introductory and results workshops in each village. Written informed consent was obtained from primary caregivers during a household visit that included an explanation of study significance, of participant requirements and rights as well as an opportunity to ask questions in Spanish and Ngäbere.

5.3.3 Study participants. In each program region, six villages were randomly selected to participate in the study from a list of villages with a population between 25-100 and that were within 2 hours walk of a health facility. In these villages, all mothers or primary caregivers from households that had ever participated in a FV or CT program and had children from 0-4 years of age were invited to participate in the study. A total of 340 children were recruited for the study who ranged in age from 1 -48 months. Of these children, 67 were 4-12 mo of age, 85 were 13-24 mo, 83 were 25-36 mo, and 105 were 37-48 mo at the time of enrolment. Multiple mothers per household (n=15) and multiple children per mother (n=62) were included in the study.

153

5.3.4 Study design and protocol. This study was part of a large collaborative investigation of child health conducted by the Panamanian Ministry of Health (MINSA), the University of Panamá and McGill University [12,13]. The outcomes of primary interest for this research (longitudinal measures of child anthropometry and diet) were collected during two 10-wk sampling periods (July – September 2008 and April - June 2009) through household visits that alternated between FV and CT regions at bi- weekly intervals to control for any seasonal variation in diet. Additional outcomes such as socio-economic status (SES), spatial data and data on intestinal protozoan infections were collected as part of the larger study during additional household visits (for details see [12]).

5.3.5 Household and child characterization. As reported previously [13], the primary caregivers of participating children were interviewed in Spanish or Ngäbere using field tested questionnaires to gather information related to household and child demographics. Specifically, information on 11 variables relating to household construction (dirt floor, absence of walls, solid walls), possessions (radio, cell phone, bicycle, sewing machine, stove, hoe) and access to running water and latrine were used to calculate our asset-based Household Wealth Index (HWI) described in Halpenny et al[12]. Years of school attended by the mother was also recorded during household interviews. A Chronic Protozoan Index (CPI) was calculated as the number of fecal samples positive for protozoan infection divided by the total number of samples for children who provided between 4 and 7 samples over a 16 mo period (for details, see [12]).

5.3.6 Anthropometry. Preschool child height/length was measured using Portable Stadiometers (Seca 214, Birmingham, UK) and Measuring Mats (Seca 210, Birmingham, UK)). 154

Child age and sex were recorded from the child’s health card. Child height-for- age Z scores (HAZ), weight-for-age Z scores (WAZ) and weight-for-height Z scores WHZ) were calculated from WHO growth reference standards [14] using WHO Anthro 3.1. (2008: n= 285; 2009: n = 264). Children were classified as stunted or underweight if their HAZ or WAZ scores respectively were ≤ -2SD. Children with a WHZ ≥2SD were classified as overweight. Change in HAZ or WAZ scores was calculated as the 2009 Z score – 2008 Z score and was used to assess child linear growth and change in weight for those children for whom anthropometric measures were made in both years (n=221).

5.3.7 Diet analysis. Trained MINSA nutritionists conducted two 24 hr recalls 1 day apart in each year (2008: n=272; 2009: n=267). Since longitudinal measures of diet were of primary interest, analysis was limited to children with at least one 24 hour recall in each year and who had not been exclusively breast fed in 24 hr recalls in either year (n=220). Of these children 175 had all 4 recalls, 42 provided 3 recalls and 3 children had only 1 recall in each year. Mothers were asked to recall all food and drink consumed by each participant child who was not exclusively breastfed in the last 24 hrs. Serving sizes were estimated using photographic aids or were measured in the home. For diet summaries, all of the 125 foods recorded in the household interviews were grouped into 54 specific food groups. Of these 54 food groups, 26 were consumed by at least 10% of the children (Table 1A) and 47 were consumed by at least 3% of the children (Table 1A, B). Dietary pattern analysis used the 47 food groups consumed by at least 3% of the population and food group multiple regression models were conducted using a simplified set of food groups outlined in Table 1 A, B.

Trained MINSA nutritionists also conducted semi-quantitative Food Frequency Questionnaires (FFQ) with the primary care giver to quantify the 155

number of days one child from the household consumed each food group. The source of the food was also recorded (ie. Bought vs grown/collected).

Dietary patterns were derived using Principal Components Analysis (PCA) [15,16] on the mean g/d intake of the 47 food groups consumed by at least 3% of the population. Scree plots identified 2 dominant diet patterns which explained 8% (first principle component) and 6% (second principle component) of the total variance in the food group variables respectively (Figure 1). Foods with component weights ≥0.15 or ≤ -0.15 were considered characteristic of that dietary pattern and used to name the pattern. The first principal component was characterized by the high consumption of fish as well as a variety of market products – pasta, breads, oil, sweets, tomato sauce, and juice (Table 2). High consumption of porridges and nance but low Nutricrema consumption were also associated with this pattern (Table 2). For reference, this pattern was named “Market”. The second principal component was characterized by the high consumption of legumes, rice, squash, avocado, sugar, and coffee as well as the low consumption of chips, meat broth, soda and juice crystals (Table 2). This pattern was named “Beans & Rice”. For each diet pattern, an individual’s score was calculated as the sum of the mean daily intake (g) of each of the 37 food groups consumed over the course of the study multiplied by the factor weight from the respective PCA component. Thus, a score was calculated for each child for both dietary patterns; higher scores indicated a diet more characteristic of a particular pattern. Energy and macronutrient intakes from complementary foods but not breast feeding were calculated for all children using the food database of the Instituto de Nutrición de Centro America y Panama (INCAP). For foods that were not included in the INCAP database, energy and nutrient values were taken from the USDA National Nutrient Database for Standard Reference (Release 24), the Handbook of Energy Crops [17] and Medicinal Plants of China [18]. Nutrient 156

information for the nutritional supplement Nutricrema was provided by MINSA. Macronutrient consumption was standardized per 1000 kcal to control for energy intake. To understand the influence of season on the subsistence components of the diet in the study regions, key informant interviews were conducted with farmers in multiple communities from each region. A calendar of growing seasons and harvest periods was created from information gathered during these interviews.

5.3.8 Spatial analysis. Geographical longitude and latitude points were recorded at each household in the larger study (n = 250) and access roads were recorded as tracks with a handheld Geographic Positioning System (GPS) (Garmin eTrex Vista HCX, Olathe KS). Geographic coordinates were converted to metric coordinates using the Universal Transverse Mercator (UTM) 17N projection and analyzed in ArcGIS Map 9.3.1. Household geographic coordinates were used to characterize the spatial dispersion of participant homes as well as the distance from access roads, as described by Halpenny et al [12]. Density estimates for each household describe the number of other study participant homes found within this circular area around their home (250m), expressed in homes per square kilometer. We also conducted spatial analyses on preschool child diet pattern scores to detect spatial clusters of individuals who consumed a diet more typical of each diet pattern (higher diet pattern score), using SaTScan software (Boston, MA). Kulldorf’s Spatial Scan Statistic uses a scanning circular window to identify distinct clusters of high diet pattern scores. For each location the circular base varies in size from zero to 50% of the population and the distribution of cases inside the window was compared to that outside the window. Likelihood ratios based on a Norm probability model for positive or negative continuous variables identified high score clusters as those where the scores within the cluster were 157

greater than expected if the diet pattern was randomly dispersed among households (p<0.05) [19].

5.3.9 Statistical analysis.

Regression models were developed for 2009 HAZ and WAZ, and for change in HAZ and WAZ from 2008 to 2009 within both the FV and CT regions. In all cases, model development used a step-wise process to build a multiple linear regression model. Two distinct sets of variables were used to create models for anthropometric indices: food group models included 20 specific food groups consumed by at least 10% of the children (Table 1) whereas the macronutrient model included energy, carbohydrate, protein and fat. Both also considered household SES and child demographic and protozoan infection variables. Age was included in all models to account for age related patterns in food group or macronutrient consumption and infection. Final models included those variables with p<0.10 and the Huber estimator for robust standard error estimation was used to account for clustering at the household level.

158

5.4 Results 5.4.1 Study population description.

Approximately 40% of study participants were in the FV program (41% households, 43% children) whereas 60% in CT program (59% households, 57% children). Households in the FV region were more impoverished (lower HWI), had less access to aqueducts, and were more remote (further from access roads and were less densely spaced) than households in the CT region (Table 3).

For children included in the longitudinal study (n=220), child demographics were similar in both the FV and CT regions (Table 3). All these children were eating solid foods and the majority of children in both regions were not breast fed in either year of the study (FV: 64%, CT: 70%):(Table 3). With respect to child height, the prevalence of stunting was 67% in both years. On average, HAZ improved by 0.11±0.06 SD between 2008 and 2009 (HAZ 2008: - 2.50±0.08; HAZ 2009: -2.40±0.07). HAZ metrics did not differ between children in the FV and CT programs. With respect to weight, about one quarter of the preschool children were underweight in both years (WAZ 2008: -1.03±0.08; WAZ 2009: -1.05±0.08). On average, WAZ scores increased in the FV region between 2008 and 2009 whereas WAZ scores decreased in the CT region over the same period (Table 3). With respect to overweight, the overall prevalence of overweight was consistent at 11% throughout the study period and did not differ between program regions (Table 3).

Mean energy consumption was low throughout the study area however there was a large range of caloric intake among study participants (Table 3). Mean carbohydrate consumption was higher in the FV region but both regions had similar intake of protein and fat (Table 3).

159

5.4.2 Diet description.

The 26 foods consumed by at least 10% of the population are displayed in Figure 2 A, B and show the strong similarities in the basic diet of preschool children in both the FV and CT region. The foods most commonly consumed (>50% of preschool children) were legumes, rice, sugar and coffee.

Program and annual comparison of food group consumption. Preschool child diet differed between programs primarily with respect to foods that complemented the basic diet (Table 4). Regardless of year, children from the FV region consumed more green bananas, eggs, citrus, peach palm (Bactris gasipaes) and legumes as well as wild fruits and vegetables whereas the consumption of breads, oil, chips and juice was higher in the CT region (Table 4)

Differences between 2008 and 2009 were observed for fruit and vegetable intake (Table 4). The consumption of citrus fruits, squash, peach palm (Bactris gasipaes), nance (Byrsonima crassifolia), wild fruit and vegetables was higher in 2008 whereas mango intake was higher in 2009. Intake of legumes, rice, sugar, breads, chips, sweets and juice was also higher in 2009. Avocado intake was highest in the FV region, but only in 2009 (Table 4).

To explore the influence of seasonality on foods available in the region, key informant interviews were conducted throughout the study area to create a timeline of the growing and harvest seasons of food crops and wild foods available in the region (Table 5). The diet staples of rice, corn and beans were available from subsistence sources (direct from harvest or stored) at the beginning of the 2008 diet analysis period. Rice and corn were also available at the end of the 2009 diet analysis period. Root vegetable consumption was captured in the 2008 sample period and local fruits and vegetables were available throughout both sample periods.

160

Data on the source of each food group was used to explore how season may have impacted diet. Two of the primary foods in the diet, rice and coffee, were purchased in over 95% of the households throughout the study (Table 6). Beans, another staple, were purchased in over half of the households in both the FV and CT regions. The majority of tubers, fruits, vegetables and eggs were collected or grown in both program regions whereas meat was primarily purchased in the CT region but provided through home agriculture in the FV region. The source of corn products was subsistence crops in the FV region but was split evenly between subsistence and purchased in the CT region. All bread products consumed were purchased.

5.4.3 Dietary pattern description.

The “Beans & Rice” diet pattern was commonly consumed in both the FV and CT regions, as indicated by the high average scores for each region. The “Market” diet pattern, however, was more commonly consumed in the CT region (Table 3). Spatial analysis confirmed that both diet patterns were widely dispersed throughout both the FV and CT regions. A single small spatial cluster of households with children who had high “Market” scores (51.5) was detected around the access road in the CT region.

5.4.4 Predictors of preschool linear growth.

Multiple regression models based on food groups for 2009 HAZ scores revealed differences between regions (Table 7). In the FV region, meat, bread and juice crystal consumption was positively associated whereas consumption of sweets was negatively associated with 2009 HAZ (Table 7A). In the CT region, children with higher HAZ consumed more pasta and corn products but less root vegetables (Table 7A). Furthermore, the change in HAZ between baseline and 2009 in each region was predicted by the same foods (Table 7B).

161

In contrast to the food group models, macronutrient models for both regions revealed that carbohydrate consumption was associated with poor 2009 HAZ scores (Table 7A) and poor linear growth between baseline and 2009 (Table 7B). Both the food group and the macronutrient models had similar explanatory power (Table 7A, B).

Children who were shorter for their age in 2008 also had lower HAZ in 2009 (Table 7A), but had a larger improvement in HAZ over 9 mo (Table 7B), regardless of region or model.

5.4.5 Predictors of preschool child weight gain.

Multiple regression models based on food groups for 2009 WAZ scores revealed differences between regions (Table 8A, B). In the FV region, eggs, fish, bread, coffee and juice crystal consumption was positively associated whereas consumption of Nutricrema was negatively associated with 2009 WAZ (Table 8A). In the CT region, children with higher WAZ consumed more milk products and Nutricrema but less sweets and chips (Table 8A). The change in WAZ between baseline and 2009 in each region was predicted by the same foods except eggs did not enter the model in the FV region (Table 8B).

The macronutrient models also differed between regions. In the FV region, children who consumed more calories and protein had a higher 2009 WAZ scores (Table 8A) and better weight gain (Table 8B). In the CT region, no macronutrients were associated with child WAZ (Table 8A).

Children who weighed less for their age in 2008 also had lower WAZ in 2009, but had a larger improvement in WAZ over 9 mo (Table 8 A,B), regardless of region or model.

162

5.5 Discussion This study provided an opportunity to compare diet and growth of preschool children between two regions of extreme poverty, one where households received monthly food vouchers and one where they received an equivalent amount in cash. In addition to the type of conditional transfer program, the two regions differed in average household asset scores, spatial density of homes, average distance to a road and access to aqueducts. Therefore, these factors were taken into consideration for all regression analyses. The basic diet throughout the region included legumes (locally referred to as “menestras”), rice, sugar and coffee; however, additional components of the diet differed between regions. Children in the FV region consumed more fruits and vegetables as well as eggs and legumes whereas market products were more common in the CT region. In both regions, high carbohydrate intake was detrimental to HAZ but not all carbohydrate source foods had a negative influence on HAZ. Consumption of root vegetables and sweets was associated with lower HAZ whereas breads, corn products and pasta were positively associated with HAZ. Unlike linear growth, preschool child weight gain differed between regions. In the FV region, despite being from poorer and more remote households, children gained more weight over the course of the study than in the CT region where market snacks (chips and sweets) were detrimental to child weight gain.

The high prevalence of stunting detected among preschool children in our study is comparable to other studies in the comarca Ngäbe Buglé where 60% of preschool children were short for their age [20,21], well below the national average of 18% [22]. Our results in combination with previous work with the Ngäbe Buglé highlight the multi-factorial causes of stunting [20]. In the current paper we have shown that high carbohydrate consumption was a factor associated with poor linear growth. This is consistent with literature showing

163

that diets high in plant based starches lack and/or inhibit the absorption of micronutrients important for growth [23,24]. Interestingly, in our food group analysis, the type of food providing carbohydrates in the diet was important. Root vegetables and sweets had a negative impact on linear growth but bread, pasta and corn products were associated with better linear growth. The average daily consumption of the root vegetables suggests that, when this food was included in the diet, it made up a larger proportion of the calories (37 kcal/day) than those from other carbohydrate sources. Specifically, the average caloric intake per day from breads, pasta and corn products ranged from 8 kcal/day to 13 kcal/day, thus it is likely that for children consuming breads, pasta and corn, their diet may have included multiple other foods and thus been more diverse than the diet of children consuming root vegetables and green bananas. Indeed, low dietary diversity has been linked to stunting in a community studies in Africa [25] and Bangladesh [26] as well as national health surveys [27]. Low diversity diets are lower in dietary quality, especially micronutrient composition [28,29] and thus may not provide the nutrients essential for growth. Our finding that meat consumption (red meat, chicken, pork) predicted better linear growth support the evidence that animal source foods are important for child growth. Animal source proteins (meat and milk) have been linked to better growth [23,30] and cognition as well as increased physical activity [31] due to the greater amount and availability of micronutrients in these foods [32]. In particular, zinc, calcium and iron as well as vitamins A and B12, found in high amounts in animal source foods, have been linked to child growth [23,33].

Although meat (red meat, chicken, pork) entered the food group model for HAZ, protein did not enter the macronutrient model for linear growth. We suggest that this further supports the importance of micronutrients from animal source foods for child growth, as the majority of the protein consumed in our study area was from plant origin. Indeed, plant based diets are low in the

164

micronutrients required for growth [34] and the inclusion of even a small amount of animal source foods in a primarily vegetarian diet has proven beneficial for child growth [23,32,33].

The present analyses built on our previous model of stunting by including diet variables in addition to socio-economic and infection variables. Previously we reported that CPI and HWI were the dominant predictors of preschool child HAZ in 2008 without considering diet [12]; however, the models presented here demonstrate a link between HAZ and diet. The current analyses differ from our previous analyses in that we specifically examined growth by including baseline HAZ (2008) and by creating models for change in HAZ over the 9 mo study period. Thus, by controlling for baseline Z score in the models of growth we were able to account for the influence of HWI and CPI on prior nutritional status and determine the specific food groups related to growth. These combined results are in agreement with stunting literature. As mentioned previously, specific micronutrients as well as protein and calorie sufficiency are necessary for child growth but in addition, chronic infection has negative consequences for growth. Anorexia and impaired nutrient absorption that result from gastrointestinal infections can directly affect nutritional status and this is particularly damaging in areas where child diet is already of low quality [35,36].

Furthermore, the host immune response caused by chronic infections can negatively affect the bone formation necessary for growth [37]. Importantly, diet has the potential to not only improve growth but also increase resistance against infection. We have shown previously that stunted children had higher reinfection intensities with Ascaris and hookworm [12] and furthermore that they did not benefit from Vitamin A supplementation that reduced intensity of Ascaris reinfection in children of normal height [20]. Our findings are in agreement with previous work that showed an association between stunting and susceptibility to diarrheal infection [38] and nematode reinfection [39-41]. Thus, 165

if the increased spending power provided by the CT programs results in improved child diet, these programs have the potential to markedly improve growth which also increases resistance against gastro-intestinal infections.

Underweight was of less concern than stunting in our study area, with a prevalence of 19%, similar to levels found previously among the Ngäbe (15%) [20] but much higher than the national average of 4% [11]. Our results from the WAZ models in the FV region demonstrated that protein consumption and not just calorie consumption were important for weight gain. This supports recent findings from supplementation research that has shown the importance of animal protein for weight gain [23]. In the CT region, macronutrients did not enter the models for change in WAZ; however, Nutricrema, a government supplied supplement, was related to weight gain suggesting that the supplement is having a positive effect on preschool child nutritional status. Interestingly, in the FV region, Nutricrema intake was negatively associated with weight gain. Children in the FV region who consumed Nutricrema were from homes with a lower HWI than those who did not consume Nutricrema thus it is possible that the household socio-economic environment influenced weight gain. Another possibility is that children who consumed Nutricrema and were from the poorer homes had a higher chronicity of protozoan infection and thus would gain less weight. We have shown previously that HWI was protective against CPI, due in part to the fact that homes with aqueduct access had higher HWI scores. If the Nutricrema fed to the children in poorer homes was made from water taken from a groundwater source and not properly boiled it is possible that the Nutricrema acted to increase exposure to protozoan infection. Overall models for WAZ support the hypothesis that socio-economic factors influenced weight gain in the FV region. Specifically, WAZ was associated with greater household asset-based wealth in the FV but not CT regions. The base level of poverty was

166

greater in the FV region therefore we suggest that the benefit of increased household wealth was more pronounced in the FV region than in the CT region.

In the CT region, the consumption of chips and sweets was also associated with a lower change in WAZ score for children. This is in contrast to recent literature that highlights association between market foods and obesity [42,43], especially in Latin America [44,45]. Our study area was in a remote region of Panamá with high levels of extreme poverty and the introduction of market type foods was relatively recent. The consumption of fat and sugar dense foods is still relatively low in our study area, approximately 2.5g/day (chips and sweets), suggesting that it has yet to experience the “nutrition transition”. Thus, we suggest that the low weight gain associated with these foods is explained by limited financial resources being spent on “unhealthy” market snacks rather than foods that provide high quality nutrients needed for child weight gain.

The use of dietary pattern analysis (DPA) to describe preschool child diet in developing countries is rare. Instead, the majority of studies of preschool children that use DPA focus on food patterns related to obesity in the developed world [46-48]. In our study, DPA proved useful for identifying the influence of socio-economic and geographic variables on dietary choices. Specifically, we were able to use DPA to identify unique combinations of food consumed by different subsets of the population. This demonstrated the utility of this method to capture cultural consumption patterns that differ greatly from dietary guidelines. Despite these positive implications, our analysis of dietary patterns had several limitations. First, we used four 24-hour recalls to generate data for the DPA rather than FFQ data. Although 24 hr recalls have been used in the generation of diet patterns [49-52], FFQ are more frequently used because they cover a longer period of time and are thus considered more representative of usual intake. The majority of studies where 24 hr recalls are used for DPA rely on 1 or 2 recalls and thus are not likely to capture infrequently consumed 167

components of the diet. In our study we were able to conduct four 24-hr recalls over a 6 month period thus we believe that we were able to capture sufficient variability in the diet while taking advantage of the greater accuracy of 24 hr recalls for assessing dietary intake [53-55].

Another related limitation of our study was the lack of dietary data between the months of October – March. Although the food consumption patterns identified are only representative of the six months studied, our data on seasonal variation of food availability as well as the source of consumed foods allowed us to discuss how preschool child diet may be affected during the 6 months not covered in our diet assessment. The base diet of beans, rice, coffee and sugar was comprised of foods that were primarily purchased and available year round from local stores (personal observation). Similarly, the 2 identified diet patterns “Market” and “Beans and Rice” were distinguished by foods that were primarily purchased leading us to hypothesize that these patterns would be present from October to March as well. The few exceptions are the fruits and vegetables found in the “Market” (nance) and “Beans and Rice” (squash and avocado) diet patterns. Although fruit and vegetable availability had the greatest seasonal fluctuations, our diet interviews were conducted during the harvest period for most items. Thus, we have described the period of most variety in the fruits and vegetables available which provided a comprehensive analysis of foods influential to growth in this region. Indeed, the fact that our analysis covered the period of root vegetable availability enabled us to detect the detrimental effect of root vegetable consumption on linear growth in the CT region. Besides root vegetables, the majority of foods influential to child growth were bought or are not expected to demonstrate a seasonal pattern (eg. eggs). Thus, although our current analysis may be insufficient to describe a typical preschool child diet throughout the year, by capturing the period when the most diverse groups of foods were available we were able to assess how these

168

seasonally available foods were incorporated into the diet and furthermore whether or not they were influential for growth.

CT and FV programs both target poverty by increasing purchasing power. Furthermore the required participation in health and education programs is expected to improve prenatal, postnatal and infant health, as well as child education and empowerment. Conditional transfer programs in Mexico, Nicaragua and Honduras have been shown to improve household dietary consumption [56] and child nutritional status [56,57]; however, our analysis of diets allowed us to explore whether the type of transfer program affected food choices. We found that mothers in the FV region fed more eggs, fruits and vegetables as well as green bananas to their children whereas mothers in the CT region provided their children with more breads, oil, chips/crackers and juices. This is in agreement with Leroy et al [10] who found that fruit and vegetable consumption was greater in households that received a food basket program compared to a cash transfer. Although our design did not allow us to explore diets before introduction of the conditional transfer programs, our available information suggests that in the FV region, higher meat intake promotes linear growth of the children and that provision of eggs and fish to the children improves their weight gain. This finding that animal source foods are beneficial for child growth is in agreement with nutrition literature [23,32,33]. Positive choices made in the CT region include pasta and corn products, associated with better linear growth and milk products and Nutricrema associated with weight gain. Indeed, fortified milk products and formulated nutritional supplements are commonly used to promote weight gain in infants [58] as well as young children [59-61].

Another component of the transfer programs that may have influenced food choices is the focus of the training programs. In the FV region, training targeted agricultural themes whereas in the CT region preventive health and 169

financial management were the primary topics. Thus, the focus on agriculture in the FV region may have increased the likelihood that mothers would choose to grow, collect or purchase agricultural products such as meat, eggs, fruits and vegetables as evidenced by what they feed their children. Indeed, in the FV region, fruits, vegetables and eggs were primarily grown or collected which may have been influenced by the training programs. Meat was also a subsistence product in this region but to a lesser extent, especially in the second year. A recent study in sub Saharan Africa highlighted the importance of the local context for complementary feeding choices and furthermore emphasized the importance of considering this context in nutrition education interventions. Specifically, in areas where coconut agriculture and access to fish was high, these foods played an important nutritional role in the diet, although the introduction of supplements reduced consumption of these foods [62].

It is important to note that our interpretation of diets and nutritional status of preschool children between the FV and CT programs is limited by the socio- economic differences between the regions where these programs were delivered. In fact, regional development in the area may influence diet more than type of transfer program. Specifically, the FV region was further from access roads, had fewer stores, and households were more impoverished. Thus households in the FV region may have a greater reliance on home grown and collected foods which could explain the higher fruit and vegetable consumption. Indeed, wild foods have been shown to be an important source of energy and protein as well as vitamins and minerals for households in Mali, especially in the rural areas [63]. Children from the CT region lived closer to access roads and consumed a diet more characteristic of the “Market” pattern as well as foods such as breads, chips and juice. Diets in areas with increased commercial activity tend to have more processed foods [45,64] as well as a higher intake of fats and sugars which is a growing health concern in the developing world [65-67].

170

Despite these differences, many households within the CT region were similarly remote and impoverished to those within the FV region [12]. An additional factor to consider during interpretation is the fact that the FV program was initiated 6 months prior to the CT program. Thus, if the rate of improvement in growth increases with time in the program than it is possible that preschool child weight gain in the CT region improved in the months after the study.

Our study has several implications for public health interventions in the comarca Ngäbe Buglé. Specifically, in an area where the majority of children consume a diet based on beans and rice, health messages to decrease the amount of carbohydrate in the diet, especially from root vegetables and sweets, and increase meat and fish consumption will improve child growth and are also likely to help promote resistance against infections common in the area. Furthermore, encouraging the substitution of sweets and chips with foods that contain nutrients for growth will improve child weight gain, especially in the CT region where these foods are more readily available.

5.6 Acknowledgements The authors would like to thank collaborators in the Panamanian Ministry of Health (MINSA) and the Universidad de Panama for logistical support and guidance during field work as well as community interviewers and study participants. We are also grateful to the assistance of Nilofar Hariri, Dianna Mohid and Sara Wing in the nutrient database construction and data entry.

171

5.7 References 1. Black R, E., Allen L, H., Bhutta Z, A., Caulfield L, E. , de Onis M, et al. (2008) Maternal and child undernutrition: global and regional exposures and health consequences. Lancet 371: 243-260.

2. Olney DK, Pollitt E, Kariger PK, Khalfan SS, Ali NS, et al. (2007) Young Zanzibari children with iron deficiency, iron deficiency anemia, stunting, or malaria have lower motor activity scores and spend less time in locomotion. J Nutr 137: 2756-2762.

3. Siegel EH, Stoltzfus RJ, Kariger PK, Katz J, Khatry SK, et al. (2005) Growth indices, anemia, and diet independently predict motor milestone acquisition of infants in South Central Nepal. J Nutr 135: 2840-2844.

4. Kuklina EV, Ramakrishnan U, Stein AD, Barnhart HH, Martorell R (2004) Growth and diet quality are associated with the attainment of walking in rural Guatemalan infants. J Nutr 134: 3296-3300.

5. Kar BR, Rao SL, Chandramouli BA (2008) Cognitive development in children with chronic protein energy malnutrition. Behav Brain Funct 4: art. no. 31. doi: 10.1186/1744-9081-1184-1131.

6. Crookston BT, Dearden KA, Alder SC, Porucznik CA, Stanford JB, et al. (2011) Impact of early and concurrent stunting on cognition. Matern Child Nutr 7: 397-409.

7. Victora CG, Adair L, Fall C, Hallal PC, Martorell R, et al. (2008) Maternal and child undernutrition: consequences for adult health and human capital. The Lancet 371: 340-357.

8. World Bank (2010) Country partnership strategy for the republic of Panamá. Washington, DC: World Bank. 97 p.

9. Fiszbein A, Schady N (2009) Conditional cash transfers: reducing present and future poverty. Washington, DC: World Bank. 383 p.

10. Leroy JL, Gadsden P, Rodríguez-Ramírez S, de Cossío TG (2010) Cash and In- Kind Transfers in Poor Rural Communities in Mexico Increase Household Fruit, Vegetable, and Micronutrient Consumption but Also Lead to Excess Energy Consumption. J Nutr 140: 612-617.

172

11. Ministerio de Economía y Finanzas (2010) La Distribución del Ingreso en Los Hogares de Panamá:Encuesta de Niveles de Vida 2008. Panamá: Ministerio de Economía y Finanzas. 34 p.

12. Halpenny CM, Koski KG, Valdés VE, Scott ME (2012) Prediction of child health by household density and asset-based indices in impoverished Indigenous villages in rural Panamá. Am J Trop Med Hyg 86: 280-291.

13. Halpenny CM, Paller C, Koski KG, Valdés VE, Scott ME (2012) A spatio- temporal analysis of soil transmitted helminth reinfection dynamics in preschool children from rural indigenous Panamá. PLoS Neglect Trop D: submitted.

14. WHO Multicentre Growth Reference Study Group (2006) WHO Child Growth Standards: Methods and development: Length/height-for-age, weight- for-age, weight-for-length, weight-for-height and body mass index-for- age: Methods and development. Geneva:

15. Hu FB (2002) Dietary pattern analysis: A new direction in nutritional epidemiology. Curr Opin Lipidol 13: 3-9.

16. North K, Emmett P (2000) Multivariate analysis of diet among three-year-old children and associations with socio-demographic characteristics. Eur J Clin Nutr 54: 73-80.

17. Duke JA (1983) Handbook of Energy Crops. Electronic Publication: www.hort.purdue.edu/newcrop/duke_energy/dukeindex.html.

18. Duke JA, Ayensu ES (1984) Medicinal Plants of China. Algonac, Michigan: Reference Publications Inc.

19. Kuldorff M, Nagarwalla N (1995) Spatial disease clusters: detection and inference. Stat Med 14: 799-819.

20. Payne L, Koski KG, Ortega-Barria E, Scott ME (2007) Benefit of vitamin A supplementation on Ascaris re-infection is less evident in stunted children. J Nutr 137: 1455-1459.

21. Ortega A, Fontes F, Sinisterra O, Valdés V (2006) Evaluación nutricional en pre-escolares y escolares de los distritos de Mironó, comarca Ngobe- Buglé y Santa Fe, Veraguas. Panama City, Panama: Secretaría del Plan Alimentario Nacional (SENAPAN). 46 p.

173

22. UNICEF (2008) The State of the World's Children 2009: Maternal and Newborn Health. New York: UNICEF. 168 p.

23. Grillenberger M, Neumann CG, Murphy SP, Bwibo NO, Weiss RE, et al. (2006) Intake of micronutrients high in animal-source foods is associated with better growth in rural Kenyan school children. Br J Nutr 95: 379-390.

24. Hotz C, Gibson RS (2001) Complementary feeding practices and dietary intakes from complementary foods amongst weanlings in rural Malawi. Eur J Clin Nutr 55: 841-849.

25. Hatløy A, Hallund J, Diarra MM, Oshaug A (2000) Food variety, socioeconomic status and nutritional status in urban and rural areas in Koutiala (Mali). Public Health Nutr 3: 57-65.

26. Rah JH, Akhter N, Semba RD, Pee SD, Bloem MW, et al. (2010) Low dietary diversity is a predictor of child stunting in rural Bangladesh. Eur J Clin Nutr 64: 1393-1398.

27. Arimond M, Ruel MT (2004) Dietary diversity is associated with child nutritional status: Evidence from 11 Demographic and Health Surveys. J Nutr 134: 2579-2585.

28. Hatloy A, Torheim LE, Oshaug A (1998) Food variety - a good indicator of nutritional adequacy of the diet? A case study from an urban area in Mali, West Africa. Eur J Clin Nutr 52: 891-898.

29. Kennedy GL, Pedro MR, Seghieri C, Nantel G, Brouwer I (2007) Dietary diversity score is a useful indicator of micronutrient intake in non-breast- feeding Filipino children. J Nutr 137: 472-477.

30. Leonard WR, Dewalt KM, Stansbury JP, McCaston MK (2000) Influence of dietary quality on the growth of highland and coastal Ecuadorian children. Am J Hum Biol 12: 825-837.

31. Neumann CG, Murphy SP, Gewa C, Grillenberger M, Bwibo NO (2007) Meat supplementation improves growth, cognitive, and behavioral outcomes in Kenyan children. J Nutr 137: 1119-1123.

32. Dror DK, Allen LH (2011) The importance of milk and other animal-source foods for children in low-income countries. Food Nutr Bull 32: 227-243.

174

33. Rivera JA, Hotz C, González-Cossío T, Neufeld L, García-Guerra A (2003) The Effect of Micronutrient Deficiencies on Child Growth: A Review of Results from Community-Based Supplementation Trials. J Nutr 133: 4010S-4020S.

34. Murphy SP, Allen LH (2003) Nutritional importance of animal source foods. J Nutr 133: 3932S-3935S.

35. Branca F, Ferrari M (2002) Impact of micronutrient deficiencies on growth: The stunting syndrome. Ann Nutr Metab 46: 8-17.

36. McKay S, Gaudier E, Campbell DI, Prentice AM, Albers R (2010) Environmental enteropathy: New targets for nutritional interventions. International Health 2: 172-180.

37. Stephensen CB (1999) Burden of infection on growth failure. J Nutr 129: 534S-538S.

38. Tomkins AM, Dunn DT, Hayes RJ (1989) Nutritional status and risk of morbidity among young Gambian children allowing for social and environmental factors. Trans R Soc Trop Med Hyg 83: 282-287.

39. Hagel I, Lynch NR, Di Prisco MC, Ṕerez M, Sánchez JE, et al. (1999) Helminthic infection and anthropometric indicators in children from a tropical slum: Ascaris reinfection after anthelmintic treatment. J Trop Pediatr 45: 215- 220.

40. Saldiva SRM, Carvalho HB, Castilho VP, Struchiner CJ, Massad E (2002) Malnutrition and susceptibility to enteroparasites: Reinfection rates after mass chemotherapy. Paediatr Perinat Epidemiol 16: 166-171.

41. Hesham Al-Mekhlafi M, Surin J, Atiya AS, Ariffin WA, Mohammed Mahdy AK, et al. (2008) Pattern and predictors of soil-transmitted helminth reinfection among aboriginal schoolchildren in rural Peninsular Malaysia. Acta Trop 107: 200-204.

42. Caballero B (2005) A nutrition paradox - Underweight and obesity in developing countries. N Engl J Med 352: 1514-1516.

43. Popkin BM (2004) The nutrition transition: An overview of world patterns of change. Nutr Rev 62.

44. Bermudez OI, Tucker KL (2003) Trends in dietary patterns of Latin American populations. Cad Saude Publica 19: S87-S99. 175

45. Aguirre-Arenas J, Escobar-Pérez M, Chávez-Villasana A (1998) Evaluation of the food consumption and nutrition in 4 rural communities. Salud Publica Mex 40: 398-407.

46. Shin KO, Oh SY, Park HS (2007) Empirically derived major dietary patterns and their associations with overweight in Korean preschool children. Br J Nutr 98: 416-421.

47. Robinson S, Marriott L, Poole J, Crozier S, Borland S, et al. (2007) Dietary patterns in infancy: The importance of maternal and family influences on feeding practice. Br J Nutr 98: 1029-1037.

48. Friedman LS, Lukyanova EM, Serdiuk A, et al (2009) Social-environmental factors associated with elevated body mass indexin Ukranian cohort of children. Int J Pediatr Obes 4: 81-90.

49. LaRowe TL, Moeller SM, Adams AK (2007) Beverage Patterns, Diet Quality, and Body Mass Index of US Preschool and School-Aged Children. J Am Diet Assoc 107: 1124-1133.

50. Knol LL, Haughton B, Fitzhugh EC (2005) Dietary Patterns of Young, Low- Income US Children. J Am Diet Assoc 105: 1765-1773.

51. Keding GB, Msuya JM, Maass BL, Krawinkel MB (2011) Dietary patterns and nutritional health of women: The nutrition transition in rural Tanzania. Food Nutr Bull 32: 218-226.

52. Kim J, Jo I, Joung H (2012) A rice-based traditional dietary pattern is associated with obesity in Korean adults. Journal of the Academy of Nutrition and Dietetics 112: 246-253.

53. Thompson FE, Subar AF (2001) Dietary assessment methodology. Nutrition in the Prevention and Treatment of Disease 2.

54. Burrows TL, Martin RJ, Collins CE (2010) A Systematic Review of the Validity of Dietary Assessment Methods in Children when Compared with the Method of Doubly Labeled Water. J Am Diet Assoc 110: 1501-1510.

55. McPherson RS, Hoelscher DM, Alexander M, Scanlon KS, Serdula MK (2000) Dietary assessment methods among school-aged children: Validity and reliability. Prev Med 31: S11-S33.

176

56. Hoddinott J, Wiesmann D (2008) The impact of conditional cash transfer programs on food consumption in Honduras, Mexico, and Nicaragua. Washington, DC: International Food Policy Institute.

57. Leroy JL, Ruel M, Verhofstadt E (2009) The impact of conditional cash transfer programmes on child nutrition: a review of evidence using a programme theory framework. Journal of Development Effectiveness 1: 103-129.

58. Miller J, Makrides M, Gibson RA, McPhee AJ, Stanford TE, et al. (2012) Effect of increasing protein content of human milk fortifier on growth in preterm infants born at <31 wk gestation: a randomized controlled trial. Am J Clin Nutr 95: 648-655.

59. Malcolm LA (1970) Growth retardation in a New Guinea boarding school and its response to supplementary feeding. Br J Nutr 24: 297-305.

60. Walker SP, Powell CA, Grantham-McGregor SM, Himes JH, Chang SM (1991) Nutritional supplementation, psychosocial stimulation, and growth of stunted children: the Jamaican study. Am J Clin Nutr 54: 642-648.

61. Grillenberger M, Neumann CG, Murphy SP, Bwibo NO, van't Veer P, et al. (2003) Food Supplements Have a Positive Impact on Weight Gain and the Addition of Animal Source Foods Increases Lean Body Mass of Kenyan Schoolchildren. J Nutr 133: 3957S-3964S.

62. Paul KH, Muti M, Khalfan SS, Humphrey JH, Caffarella R, et al. (2011) Beyond food insecurity: How context can improve complementary feeding interventions. Food Nutr Bull 32: 244-253.

63. Nordeide MB, Hatloy A, Folling M, Lied E, Oshaug A (1996) Nutrient composition and nutritional importance of green leaves and wild food resources in an agricultural district, Koutiala, in Southern Mali. Int J Food Sci Nutr 47: 455-468.

64. Romaguera D, Samman N, Rossi A, Miranda C, Pons A, et al. (2008) Dietary patterns of the Andean population of Puna and Quebrada of Humahuaca, Jujuy, Argentina. Br J Nutr 99: 390-397.

65. Drewnowski A, Popkin BM (1997) The Nutrition Transition: New Trends in the Global Diet. Nutr Rev 55: 31-43.

66. Popkin BM, Adair LS, Ng SW (2012) Global nutrition transition and the pandemic of obesity in developing countries. Nutr Rev 70: 3-21. 177

67. Poskitt EME (2009) Countries in transition: Underweight to obesity non-stop? Ann Trop Paediatr 29: 1-11.

178

A Food Food Group Group Food Food (PCA) (PCA) Red Beef, Full Cut, Boiled Mango6 Mango, mature Meat1 Beef, Full Cut, Fried Mango, green Byrsonima Goat, Meat Nance6 crassifolia Rabbit, Meat Ripe Banana6 Banana, Datil Chicken breast, meat Chicken1 Banana, Mature only, stewed Chicken meat, meat Plantain, Mature only broiled Chicken meat, meat Avocado6 Avocado only fried Chicken meat, meat Root Cassava, root only stewed Vegetables7 Chicken, meat & skin, Xanthosoma sp, raw fried Chicken, meat & skin, Taro, cooked w/o

stewed salt Chicken, thigh, meat Yams, boiled, w/o

only, fried salt Chicken, thigh, meat Potatoes, boiled w/o

only, stewed skin Chicken, thigh, meat Potatoes, boiled w/

& skin, fried skin Chicken, thigh, meat Green Bananas8 Banana, immature & skin, stewed Chicken wing, meat Plantain, immature only, stewed Chicken wing, meat Green banana,

only, fried cuadrado Chicken wing, meat & Corn yellow, Corn Products9 skin, stewed degermed Fish2 Croaker Green corn, 179

immature Largetooth sawfish Green corn, yellow Mojarra Tortilla, corn yellow Rice, white, Robalo Rice10 unenriched Bread Bread, wheat, Snapper Products11 French-type Canned Sardine canned w/ oil Hojaldre Fish2 Sardine canned w/ Wheat flour, all

tomato purpose Tuna, canned w/ oil Porridges12 Barley cream flour Tuna, canned w/ Corn cream flour water Milk Milk, dry, whole Rice cream flour Products3 Milk, evaporated, Sugar13 Sugar, white, canned granulated Milk, powdered, Sugar, brown, crude fortified Egg4 Egg, yolk Oil14 Oil, Vegetable, Soybean Egg, boiled Soup15 Canned soup, vegetable Egg, fried Soup dehydrated, chicken noodle Legumes5 Beans, red Soup dehydrated, beef vegetable Beans, chiricano Sweets16 Candies, hard Beans, runner Sugar Lentils Vanilla, cream filling Pigeon peas Sweet bread (Panamá) Beans, all varieties Chips/Crackers17 Crackers, saltines Citrus Mandarin Snacks, corn twists, Fruit6 cheese Lime Snacks, potato chips, plain Lemon Juice Crystals18 Drink flavour, powder Mamon, Spanish lime Coffee19 Coffee grain, toasted, powder Sweet orange Nutricrema20 Nutricrema 180

Superscript numbers 1-22 represent the food groups used in the food group multiple linear regression models for preschool child anthropometry. 1=Meat, 2=Fish, 3=Milk Products, 4=Eggs, 5=Legumes, 6=Fruits & Vegetables, 7=Root Vegetables, 8=Green Bananas,9=Corn Products, 10=Rice, 11=Bread Products, 12=Porridges, 13=Sugar, 14=Oil, 15=Soup, 16=Sweets, 17=Chips/Crackers, 18=Juice Crystals, 19=Coffee, 20=Nutricrema.

181

Food Group Food Group B Food Food (PCA) (PCA) Processed Frankfurt, Beef Peach Palm Bactris gasipaes Meat & Pork Luncheon Meat, Tomato Tomato, red, raw Canned Chicken liver, Pumpkin, Leaves & Chicken Liver Green Leafy cooked shoots Chicken Chicken gizzard, Phytolacca

Extras cooked americana Chicken feet, Pumpkin, mature Squash boiled and yellow Chicken neck, Spices Celery, stalk, raw simmered Crab, Blue, Crustaceans Garlic bulb, raw cooked Shrimp, cooked Onions bulb, raw Pork, meat, pan- Peppers, sweet Pork fried red, raw Pork, meat, Pasta, commercial, Pasta cooked unenriched Pork, skin fried Meat Broth Broth, chicken Cheese, pasteurized, Cheese Broth, beef processed, american Sauce, tomato, Papaya Papaya, pulp Tomato Sauce canned Watermelon, Juice, nectar, Watermelon Juice raw peach Wild Fruit & Inga edulis Juice, nectar, pear Vegetables Sweet orange, Zapote natural juice Nutritional Cashew, fruit Nutritional Biscuits Biscuits Cabbage/ Carbonated Cabbage, raw Soda Chayote beverage, cola Chayote, Carbonated

cooked, no salt beverage, non cola

182

Table 2. Factor weights of the 2 Principle Component Analysis derived diet patterns detected among Panamanian preschool children.1

Market Beans & Rice Weight2 Weight PROTEIN Fish 0.3524 Legumes 0.3097 FRUIT & VEGETABLES Nance 0.1725 Avocado 0.1633 Squash 0.1735 STARCHES Porridges 0.3314 Rice 0.3594 Breads 0.277 Pasta 0.3685 MARKET EXTRAS Sugar 0.3037 Oil 0.3316 Sweets 0.1822 Chips/Crackers -0.346 Meat Broth -0.177 Tomato Sauce 0.3552 SUPPLEMENTS & BEVERAGES Nutricrema -0.157 Juice 0.2835 Soda -0.282 Coffee 0.4107 Juice Crystals -0.16 1 Characteristic food groups had factor weights ≥0.15 or ≤ -0.15

2Weights determined as the variable loading from the 1st (Market) and 2nd (Beans & Rice) components of Principal Components Analysis that explained 8% and 6% of the variance in food group consumption respectively.

183

Table 3. Summary statistics of household and child variables comparing Food Voucher (FV) and Cash Transfer (CT) regions1. FV Region CT Region

n Mean±SEM n Mean±SEM p

Household2

Wealth Index, HWI3 68 -0.11±0.05 96 0.34±0.07 <0.001 Household Density, 70 14.6±2 94 41.9±4 <0.001 houses/km2 Distance to Road, km 70 2.0±0.2 94 1.3±0.2 <0.001 Mother's education, 71 2.9±0.4 97 3.6±0.4 0.34 yrs Latrine Ownership, % 70 20(12-31) 97 33(24-43) 0.06 Aqueduct Access, % 71 11(6-21) 97 42(33-52) <0.001 Child

2008 Age, mo 94 29.4±1.5 126 28.2±1.2 0.56 1 years 7 7% 15 12%

2 years 22 23% 33 26%

3 years 26 28% 28 22%

4 years 39 41% 50 40%

Sex, % female 94 51(41-61) 126 57(48-65) 0.37 Breast Fed4

Neither year 60 64% 88 70%

2008 22 23% 23 18%

2008 & 2009 9 10% 15 12%

Height for Age

Stunted 2008, % 5 86 63(52-72) 122 70(61-77) 0.30 Z score 2008 86 -2.46±0.12 122 -2.52±0.11 0.34 Stunted 2009, % 80 61(50-71) 115 73(64-80) 0.08 Z score 2009 80 -2.27±0.11 115 -2.48±0.09 0.08 Change HAZ 73 0.18±0.10 111 0.06±0.07 0.15 Weight for Age

Underweight 2008, % 87 23(15-33) 122 22(16-30) 0.88 5 Z score 2008 87 -1.05±0.14 122 -1.02±0.10 0.43 Underweight 2009, % 80 16(10-26) 116 22(15-30) 0.36 Z score 2009 80 -0.80±0.14 116 -1.22±0.10 0.007 Change WAZ 74 0.17±0.16 112 -0.17±0.11 0.04 Weight for Height

Overweight 2008, %6 87 11(6-18) 122 11(5-19) 0.97 Overweight 2009, % 80 14(8-23) 116 9(5-16) 0.35 Macronutrient

184

Consumption Mean Energy, kcal 94 706±30 126 707±28 0.99 Min, Max 94 139, 1598 126 66, 1792

Mean Carbohydrates, 94 213±2 126 205±2 0.03 g/1000kcal Min, Max 94 172, 247 126 125, 247

Mean Protein, 94 21.3±0.6 126 23.5±0.7 0.08 g/1000kcal Min, Max 94 8.8, 38.1 126 8.0, 59.8

Mean Fat, g/1000kcal 94 9.8±0.6 126 11.4±0.8 0.47 Min, Max 94 1.9, 27.1 126 1.3, 48.1

Diet Patterns Market Score 94 2.6±1.4 126 13.5±2.1 <0.001 Beans & Rice Score 94 57.8±2.9 126 52.0±2.4 0.29 1 Summary statistics presented are Mean ±SEM or % (95% CL). 2 Household characteristic data were collected once in 2008. 3Asset based index, weights derived from the first component of Principle Components Analysis. 4Percentage of children that were breast fed in neither year, 2008 only or both years. All children included in dietary analysis were eating solid foods. 5< -2SD below the mean. 6> 2SD above the mean.

185

FV Region (n=94) CT Region (n=126) ANOVA1 2008 2009 2008 2009 Mean±SEM Mean±SEM Mean±SEM Mean±SEM Program Year PROTEIN

Eggs 5.0±1 5.8±1 1.9±0.6 3.4±0.9 FV>CT Legumes 7.7±1 14.4±2 6.1±0.9 9.7±0.8 FV>CT 2009>2008 FRUITS&VEGETABLES

Mango 0±0 40.4±7 1.8±1 40.7±8 2009> 2008 186 Citrus fruit 13.2±4 2.8±1 3.3±2 1.8±1 FV>CT 2008> 2009 Avocado 0±0 26.3±5 2.1±1 7.9±3 FV>CT2 2009>20082 Squash 3.7±2 0±0 1.2±0.8 0±0 2008>2009

Wild Fruit 12.4±5 3.3±1 0.7±0.7 0.3±0.2 FV>CT 2008>2009 &Veg Pixbae 6.5±3 1.0±0.7 1.5±1 0±0 FV>CT 2008>2009 Nance 5.0±2 0±0 4.8±1 0.8±0.4 2008>2009

STARCHES

Rice 71.4±6 90.9±6 64.0±5 99.1±6 2009>2008

Breads 0.4±0.3 1.1±0.6 2.7±0.8 7.3±2 CT>FV 2009>2008 Green 57.0±12 67.8±7 52.6±8 30.5±4 FV>CT Banana

186

MARKET EXTRAS

Sugar 28.4±2 34.2±3 30.3±3 37.7±2 2009>2008

Oil 0.9±0.3 0.4±0.1 1.8±0.6 1.5±0.3 CT>FV Chips/ 0.1±0.1 0.8±0.4 0.5±0.2 1.7±0.4 CT>FV 2009>2008 Crackers Sweets 0.2±0.1 0.5±0.2 0.2±0.1 1.6±0.5 2009>2008

SUPPLEMENTS & BEVERAGES

Juice 0±0 2.7±2 3.5±3 12.8±4 CT>FV 2009>2008 1Two way ANOVA (program, year) for mean amount of food group consumed. 2 The only significant interaction term was for avocados, where intake was higher in the FV region in 2009. 187

187

Jan Feb Mar Apr May Jun Jul Aug Sept Oct Nov Dec Data 2008 2009 Collection Rice X Corn X Legumes (bush) Legumes (vine) Vegetables Taro Cassava 188 Xanthosoma sp.

Yams Mango Avocado Oranges Banana Lemon Peach Palm Zapote Guava Nance Jiraca Water apple Mamon

188

Table 6. Percentage of households who bought each food group for those households in which the food was consumed.

2008 2009 FV region CT region FV region CT region n % (95% CL) n % (95% CL) n % (95% CL) n % (95% CL) Rice 71 99 (92-100) 132 96(91-99) 77 96 (89-99) 123 92 (86-96) Legumes 60 60 (47-71) 99 77 (67-84) 74 55 (44-66) 117 72 (63-79) 66 114 69 110 100 (96- Coffee 98(91-100) 96 (91-99) 96 (87-99) 100) Tubers 38 29 (17-45) 60 37 (26-49) 50 10 (4-22) 77 35 (25-46) Corn 21 34 19 35 10(1-30) 50 (34-66) 32 (15-54) 49(33-64) Products Bread 19 100 (80- 71 100 (94- 24 100 (84- 82 100 (95- Products 100) 100) 100) 100) Fruit 38 0 (0-11) 34 12 (4-27) 55 2 (0-11) 72 4 (1-12) Vegetables 22 27(13-48) 35 34 (21-51) 45 9(3-21) 63 43 (31-55) Eggs 31 23 (11-40) 46 43 (30-58) 45 4 (0-16) 64 39 (28-51) Meat 37 24 (13-40) 72 81 (70-88) 59 49(37-62) 93 75(66-83)

189

Table 7. Hierarchical multiple linear regression models of HAZ scores (A) and change in HAZ scores (B) of Panamanian preschool children by Conditional Transfer region1.

A) HAZ 2009 Food Group Model Macronutrient Model

FV Region CT Region FV Region CT Region

β2 p β p β p β p

Age, mo 0.12 0.18 0.13 0.03 0.12 0.21 0.15 0.02 2008 Z score 0.77 <0.001 0.78 <0.001 0.66 <0.001 0.80 <0.001 Wealth Index, HWI3 0.09 0.08 NE4 NE Chronic Protozoan Infection, CPI5 NE NE -0.15 0.06 NE Meat, g/1000 kcal 0.11 0.05 Pasta, g/1000 kcal NE 0.19 0.01 Bread, g/1000 kcal 0.34 <0.001 NE

190 Corn Products, g/1000 kcal NE 0.12 <0.001 Root Vegetables, g/1000 kcal NE -0.11 0.02

Green Bananas, g/1000 kcal NE -0.11 0.06 Juice Crystals, g/1000 kcal 0.32 <0.001 NE Sweets, g/1000 kcal -0.13 0.04 NE Fat, g/1000 kcal -0.25 0.06 NE Carbohydrates, g/1000 kcal -0.37 0.04 -0.12 0.05 n 64 96 64 96 p <0.001 <0.001 <0.001 <0.001 F 29.68 39.96 30.04 71.32 R2 0.73 0.71 0.57 0.65

190

B) Change in HAZ (2008-2009) Food Group Model Macronutrient Model

FV Region CT Region FV Region CT Region

β p β p β p β p

Age, mo 0.14 0.18 0.17 0.03 0.14 0.21 0.20 0.02 2008 Z score -0.44 <0.001 -0.57 <0.001 -0.57 <0.001 -0.54 <0.001 Wealth Index, HWI3 0.11 0.08 NE NE Chronic Protozoan NE NE -0.18 0.06 NE Infection, CPI5 Meat, g/1000 kcal 0.14 0.05 NE Pasta, g/1000 kcal NE 0.25 0.01 Bread, g/1000 kcal 0.41 <0.001 NE Corn Products, g/1000 kcal NE 0.17 <0.001 Root Vegetables, g/1000 NE -0.14 0.02 kcal Green Bananas, g/1000 NE -0.15 0.06 kcal Juice Crystals, g/1000 kcal 0.40 <0.001 NE Sweets, g/1000 kcal -0.16 0.04 NE 191 Fat, g/1000 kcal -0.31 0.06 NE Carbohydrates, g/1000 kcal -0.45 0.04 -0.17 0.05 n 64 96 64 96 p <0.001 <0.001 <0.001 <0.001 F 28.19 17.38 8.13 18.02 R2 0.59 0.48 0.57 0.65

191

1 Variables that did not enter any model: Sex, maternal education, and household density. Eggs, fish, milk products, fruits and vegetables, legumes, rice, porridge, Nutricrema, oil, sugar, coffee, juice, soda and chips/crackers did not enter the Food Groups models. Kilocalories and protein did not enter the macronutrient models. 2Standardized regression coefficient 3Asset based index, weights derived from the first component of Principle Components Analysis. 4 NE = Variable was excluded during the stepwise process (p>0.10). 5Chronic Protozoan Index (CPI), proportion of positive stool samples for children who provided stool samples at 4 or more of 7 sampling periods.

192

Table 8. Hierarchical multiple linear regression models of WAZ scores (A) and change in WAZ scores (B) of Panamanian preschool children by Conditional Transfer region.1 A) WAZ 2009 Food Group Model Macronutrient Model

FV Region CT Region FV Region CT Region

β2 p β p β p β p

Age, mo 0.10 0.30 -0.10 0.19 0.02 0.85 -0.06 0.45 2008 Z score 0.45 <0.001 0.58 <0.001 0.36 <0.001 0.52 <0.001 Sex -0.23 0.03 NE3 NE NE Wealth Index, HWI4 0.17 0.04 0.20 0.06 House Density, homes/km2 -0.24 0.008 NE Milk Product, g/1000 kcal 0.14 0.03 Eggs, g/1000 kcal 0.23 0.03 NE Fish, g/1000 kcal 0.23 0.01 NE Bread, g/1000 kcal 0.20 <0.001 NE Pasta, g/1000 kcal -0.06 0.07 NE Nutricrema, g/1000 kcal -0.42 <0.001 0.18 0.004 193 Coffee, g/1000 kcal 0.35 0.002 NE

Juice Crystals, g/1000 kcal 0.24 0.002 NE Sweets, g/1000 kcal NE -0.17 0.01 Chips, g/1000 kcal NE -0.28 <0.001 Protein, g/1000 kcal 0.24 0.02 NE Energy, kcal 0.28 0.01 NE n 65 97 65 97 F 24.56 8.94 7.60 13.64 P <0.001 <0.001 <0.001 <0.001 R2 0.58 0.43 0.35 0.29

193

B) Change in WAZ (2008-2009) Food Group Model Macronutrient Model

FVRegion CT Region FV Region CT Region

β p β p β p β P

Age, mo 0.08 0.28 -0.11 0.19 0.002 0.99 -0.06 0.45 2008 Z score -0.50 <0.001 -0.42 <0.001 -0.60 <0.001 -0.49 <0.001 Sex -0.18 0.04 NE -0.16 0.08 NE Wealth Index, HWI4 0.14 0.07 0.20 0.06 NE NE House Density, homes/km2 -0.19 0.02 NE NE NE Milk Product, g/1000 kcal NE 0.15 0.03 Eggs, g/1000 kcal NE NE Fish, g/1000 kcal 0.21 0.02 NE Bread, g/1000 kcal 0.16 <0.001 NE Pasta, g/1000 kcal NE NE Nutricrema, g/1000 kcal -0.28 0.001 0.18 0.004

194 Coffee, g/1000 kcal 0.31 0.001 NE Juice Crystals, g/1000 kcal 0.25 0.001 NE

Sweets, g/1000 kcal NE -0.17 0.01 Chips, g/1000 kcal NE -0.29 <0.001 Protein, g/1000 kcal 0.19 0.06 NE Energy, kcal 0.31 0.003 NE n 65 97 65 97 F 27.67 11.61 13.40 10.32 P <0.001 <0.001 <0.001 <0.001 R2 0.65 0.38 0.46 0.23

194

1Variables that did not enter any model: Maternal education, Chronic Protozoan Index, (CPI). Meat, legumes, fruits and vegetables, rice, porridge, corn products, root vegetables, green bananas, oil, sugar, juice and soda did not enter the Food Group models. Fat and carbohydrate did not enter the Macronutrients models.2Standardized regression coefficient3 NE = Variable was excluded during the stepwise process (p>0.10).4Asset based index, weights derived from the first component of Principle Components Analysis.

195

4 3 2 Eigenvalues 1 0

0 10 20 30 40 Component

Figure 1 – Scree plot of eigenvalues after Principle Component Analysis (PCA) of dietary intake data.

196

Food Voucher, % Consumption A 0 20 40 60 80 100 Red Meat 2008 Fowl 2009 Fish

PROTEI Canned Fish Eggs 1

Legumes Ripe Banana Mango Citrus Fruit Avocado FRUIT & VEG Nance Porridges Rice Root Veg Corn Product

STARCH Breads Green Banana Sugar

Oil Chips/Crackers EXTRA

MARKET Sweets Soup

Nutricrema Milk Products Coffee Juice Crystals BEVERAGE & SUPPLEMENT SUPPLEMENT

197

Cash Transfer, % Consumption 0 20 40 60 80 100 B Red Meat 2008

Fowl 2009 Fish

PROTEIN Canned Fish Eggs

Legumes Ripe Banana Mango Citrus Fruit Avocado FRUIT & VEG Nance Porridges Rice Root Veg Corn Product

STARCH Breads Green Banana Sugar

Oil Chips/Crackers EXTRA

MARKET Sweets Soup

Nutricrema Milk Products Coffee Juice Crystals BEVERAGE & SUPPLEMENT SUPPLEMENT

Figure 1. Percentage of Panamanian preschool children in the Food Voucher (A) and Cash Transfer (B) programs that consumed specific foods or food groups in 2008 and 2009 consumed by at least 10% of the population in 2008 or 2009.

1 Legumes included red beans, chiricano beans, lentils and pigeon peas.

198

Connecting Statement III

In Chapter 5, I have shown that type of conditional transfer program is another important contextual determinant of preschool child health. Specifically, I report differences between the food voucher (FV) and cash transfer (CT) regions with respect to food consumption patterns, rate of weight gain over a 9 mo period, as well as the predictors of weight gain and linear growth. Although carbohydrate consumption was detrimental to child linear growth in both regions, the types of carbohydrate dense foods related to lower growth differed by program region. Furthermore, meat consumption improved linear growth, but only in the FV region. Regional differences were also detected in the dietary predictors of weight gain. Fish and eggs improved weight gain in the FV region whereas milk products were beneficial but sweets and chips were detrimental for weight gain in the CT region. Results from Chapters 3-5 have focused on quantitative results that highlight the complexity of regional, household and individual factors that determine child health outcomes. In Chapter 6, I examine perceptions of local health priorities in the FV and CT regions through a series of participatory methodologies with small groups in each of the communities that informed the quantitative study design and the combined quantitative and qualitative analysis identified public health targets for the region. I also explore whether the priority health concern most commonly identified (diarrhea) and local understanding of causal and preventative pathways are supported by data from fecal samples and water samples. The participatory approach highlighted unique components to be included in public health interventions that would not have been identified without participant suggestions.

199

CHAPTER 6

A mixed methods approach reveals understanding of the interrelationships between diarrhea and water quality among the Panamanian Ngäbe Buglé

Carli M. Halpenny1, Kristine G. Koski2, Marilyn E. Scott1

1Institute of Parasitology and McGill School of Environment, 2School of Dietetics and Human Nutrition Macdonald Campus of McGill University, Ste-Anne de Bellevue, Quebec, Canada,

200

6.1 Abstract Conditional Transfer Programs (CTP) are a widely used form of social support that aims to reduce poverty by providing recipients with a short term economic incentive that is conditional upon participation in education and health programs. Despite a specific focus on health and nutrition, CTPs have not recorded the expected improvements in health and nutrition outcomes, possibly due to a lack of consideration for the socio-cultural context. This study identified and compared participant-defined health priorities between two variants of a CTP in rural Panamá and developed key targets to inform health promotion in the regions. Mixed methods combined influence diagrams for health priorities created by 68 small groups with household measures of water quality and water purification knowledge and practice and child estimates of diarrhea,. Further small group work identified attitudes towards and barriers against water purification. The high prevalence of diarrhetic stool samples over a 16 mo period (18 - 50%) confirmed the finding that diarrhea was the top health concern in 28 of 64 small groups E.coli counts from household water samples revealed lower counts among households with access to aqueducts (75±33) than those who used ground water (270±43), and supported small group influence diagrams showing that access to an aqueduct and water purification could reduce diarrhea by improving water quality. Despite their knowledge of the link between water quality and diarrhea, less than 10% of the 210 caregivers interviewed reported purifying water. Focus group discussions identified the poor taste of purified water, increased use of firewood and lack of confidence in knowledge of chlorination as the primary reasons for the gap between knowledge and practice of water purification methods. Four public health targets were identified: 1) messages about the use of bleach for household water chlorination should be more clear; 2) the taste of purified water and extra firewood use associated with boiling water need to be addressed to increase the uptake of water treatment

201

messages; 3) public health education needs to promote action rather than just distribute information; and 4) education in aqueduct management for community aqueduct committees would improve the quality and service of community aqueducts.

202

6.2 Introduction Preschool child health is considered crucial to community development and poverty alleviation. With the development of the United Nations multi- dimensional definition of poverty, poverty reduction strategies are addressing the social, environmental and economic forces that contribute to poverty (United Nations, 2006, resolution 1, annex II). Conditional transfer programs in particular are believed to provide recipients with the capabilities and resources to escape poverty by providing a short term economic incentive that is conditional upon participation in education and health programs [1,2].

These types of interventions, however, are not experiencing the benefits that could be expected in health and nutrition outcomes, possibly because they do not take into account the socio-cultural context influencing health care decisions [3]. In addition to encouraging health care use, conditional transfer programs often have a training component that addresses topics such as health promotion, agricultural skills and financial management. The design and execution of these training programs is one way in which conditional transfer programs could impact on community perspectives of health needs/concerns/initiatives. Indeed, health promotion strategies that incorporate community involvement in identifying priorities as well as designing solutions have been shown to improve behaviour change [4] especially when informed by social psychology theories [5-7]. Participatory methodologies focus on involving local communities in the research process and are known for their capacity to characterize multiple contexts and processes that surround the research interests [8-10].

Mixed methods research combines quantitative and qualitative inquiry to give rich context to testable patterns of association [11-13]. Thus, mixed methods are considered particularly beneficial in applied fields due to their usefulness in addressing ‘complex’ health environments [14,15]. The majority of 203

mixed methods study designs can be described as either sequential, when quantitative and qualitative data are collected in succession, or concurrent, when data collection occurs simultaneously [11,13,16]. A sequential study design allows one phase of data collection to inform the next and thus has the potential for a more integrated research approach, a goal of mixed methods research [17,18]. Depending on the nature of the mode of inquiry used in the first phase of research sequential study designs can be exploratory (qualitative first) or explanatory (quantitative first) [16]. Despite the strength of a sequential mixed method design, the significant time needed for data collection in these studies is often a limiting factor in their adoption [13,16]. Our 16 month study in rural Panamá provided an opportunity to take a sequential mixed methods approach to examine health priorities in two regions of the comarca Ngäbe Buglé that differed in their level of development and the type of conditional transfer program in place. Specifically, we were able to combine an exploratory and explanatory sequential design to: 1) identify, characterize and compare the major health concerns in the two regions; 2) relate the perceived top health concern (diarrhea) with frequency of diarrhea and water quality measures in the regions; and 3) identify barriers to water preventive health behaviours that will inform health promotion activities in the regions. Through our participatory workshops, it became clear that the most common top health concern was diarrhea. Influence diagrams revealed that the participants understood this to be strongly linked with water quality. Therefore this paper focuses primarily on interactions between diarrhea and water quality in an Ecohealth framework.

6.3 Materials and Methods 6.3.1 Study area and population. The comarca Ngäbe-Buglé is a semi-autonomous political region inhabited primarily by the Ngäbe and Buglé indigenous groups (2004 population estimate of 128,978) of whom 91% live in extreme poverty (< US$ 1.75/day) [19]. The

204

Conditional Transfer Program Red de Oportunidades provides an additional US$ 50/month in either cash or food vouchers to the over 90% of families in extreme poverty in the comarca in exchange for participation in health and education programs. The study population has been described previously [20]. In brief, the study was conducted in the district of Besiko in two adjacent corregimientos or political regions (Soloy and Emplanada de Chorcha), each accessible most of the year by one dirt road. In Soloy the Cash Transfer (CT) version of the Red de Oportunidades is in operation whereas in Chorcha it is the Food Voucher (FV) version. Approximately 1/3 of houses had access to aqueducts and latrines. The study regions were accessible through two dirt roads, the western road having limited access during the rainy season (July—November). Household density varied considerably throughout the study area, with more densely populated regions being closer to a road and having better access to latrines, aqueducts and health facilities as well as a greater average asset based wealth score. Mothers had completed less than 4 years of education, the average child age was 26 months and half the children were female.

6.3.2 Study design.

This study was part of a large collaborative investigation of preschool child health conducted by the Panamanian Ministry of Health (MINSA), the University of Panamá and McGill University to help inform public health initiatives for MINSA and the Red de Oportunidades programs [20,21]. The current study was conducted in 3 phases and adopted a sequential mixed methods design with exploratory and explanatory components [16] that was integrated by connecting the results of previous phases with the design of subsequent phases. Phase I consisted of workshops to identify priority health issues in each of the 12 villages that were randomly selected (6 from each study region). Phase II consisted of household surveys to quantify the priority health

205

concerns identified in Phase I workshops and for the top health concern, diarrhea, additional information was sought to describe related behaviours and practices. Thus, Phase I and II of the study were exploratory in nature whereby the factors identified through qualitative methodologies in Phase I were explored through collection of quantitative data in Phase II. In Phase III, the explanatory phase, follow up community workshops were conducted to assist in interpreting the quantitative findings from Phase II and to identify target areas for public health initiatives. In line with Participatory research theory, the specific methods used for each Phase of research were chosen based on their local acceptance and validity in addition to their ability to address the research objectives [8].

6.3.3 Phase I Workshops.

In Phase I, day long workshops were conducted in all 12 communities (6 in the CT region and 6 in FV region) which were open to all members of each community and used methods from Participatory Rural Assessment [8] to identify the health priorities of the communities in the region. Participation ranged from 20 – 40 participants and included men, women and children. The first part of the workshop consisted of introductions between the community members and researchers to gain familiarity with each other and the proposed project. Next, participants worked in self-assigned small groups (4-8 participants) to brainstorm about health issues that occurred in their community. This involved a total of 34 small groups in each region. Each health issue was written on a cue card that was then stacked in order of priority (lowest priority on the bottom, highest on the top) separately for each small group. In the afternoon, each small group created influence diagrams [22] that identified factors that they felt caused or prevented their priority health issue. Influence diagrams were presented to the rest of the small groups and themes repeated in the influence diagrams were discussed. 206

Qualitative data from workshops was recorded on flip charts and in field notebooks. The top priority health issue from each small group from the cue card exercise was recorded. Further analysis was done only on the most commonly identified top priority (diarrhea). Factors included in the 28 diarrhea influence diagrams were grouped into the following thematic categories: water quality, hygiene, food safety, health facilities, parasites, weather, nutrition, traditional medicine, poverty (Table 2). In each thematic category, factors were recorded as either preventive or causative depending on how the small group depicted the relationship between that factor and the health outcome. Tallies were kept of the number of times each thematic category was mentioned as either preventive or causative on the influence diagrams.

6.3.4 Phase II Surveys

In the 12 villages where Phase I workshops were conducted, households with children under 4 yrs of age that had participated in a Red de Oportunidades program were invited to participate in the household surveys (for details, see [20]). Information from Phase I workshops was used to inform survey development and all surveys were pilot tested prior to use. In particular, the priority concern regarding diarrhea led us to determine the frequency of diarrhea in the area, the quality of drinking water, and knowledge and behaviours regarding water treatment.

Water quality, knowledge and behaviour. In the wet and dry season, water samples taken from the point of consumption in each household (CT region: n=125, FV region n=85) were analyzed within 6 hours using Colilert/Quanti--tray® according to manufacturer’s instructions (IDEXX, Westbrook, ME). Quanti-Tray uses a modified most probable number assay to estimate the concentration of Escherichia coli (cfu/100 mL) in samples from each home. Primary caregiver knowledge of water treatment and diarrhea as well as

207

household water treatment practices were assessed using questionnaires conducted at the time of water sample collection (CT regions: n=125, FV region: n=85).

Child infection status. Stool samples were collected from preschool children in participating households. Samples were identified as diarrhetic (binary classification) macroscopically by trained MINSA laboratory technicians. In addition to overall prevalence of diarrhea, a Chronic Diarrhea Index (CDI) was calculated for each child (CT region: n=170, FV region: n=122) as the number of fecal samples positive for protozoan infection divided by the total number of samples provided for children who provided at least 4 samples [21].

6.3.5 Phase III Workshops

Results of the Phase I workshop and the Phase II surveys were presented to community members. These provide an overview of regional results in addition to community specific details. Phase III workshops were conducted in 11 of the 12 communities (due to an unforeseen conflict precluded a workshop in the 12th community) and participation ranged from 15-40 participants, again a mix of men, women and children. Given the focus on diarrhea, presentations of results served to initiate activities designed to investigate knowledge of fecal/oral pathogen transmission and discuss barriers to preventive actions. Specifically, participants worked in small groups to develop “stories” of families or children who became infected with a gastrointestinal pathogen using cut-outs on a felt board. The objective of the first exercise with the felt boards was for participants to tell a “story” of how the characters became infected in order to examine their understanding of transmission pathways for gastrointestinal pathogens. These “stories” were then presented to the larger group and initiated a brainstorming session in which groups identified how the characters in their “stories” could have prevented the infection. For example, if the infection 208

occurred when the person drank water taken from an untreated ground water source, the group identified boiling water before drinking it. After the dominant preventive actions were identified, participants discussed barriers that prevented these actions from occurring and brainstormed ways to enable these actions. The majority of Phase III workshop participants had participated in Phase I and Phase II of the study but additional community members were also welcome.

Analysis of the results of the Phase III workshops involved examination of field notes for themes that were summarized to identify the dominant issues.

6.3.6 Ethical considerations.

Ethical approval was obtained from the Instituto Conmemorativo de Gorgas in Panamá and McGill University in Canada. Community interactions were established in accordance with the Guía para Realizar Estudios e Investigaciones en los Pueblos Indígenas de Panamá, which included participation in introductory and results workshops in each village. A lunch was provided for participants at Phase I and III workshops although no other remuneration was provided for their participation. For the household surveys, written informed consent was obtained from primary caregivers during a household visit that included an explanation of study significance, of participant requirements and rights as well as an opportunity to ask questions in Spanish and Ngäbere. All study information was coded to preserve participant confidentiality.

209

6.3.7 Statistical analysis.

Phase II quantitative survey data on knowledge, behaviours, water quality and CDI were compiled and compared between regions statistically using STATA 11.1 (College Station, TX). In all cases, the level of significance was set at p < 0.05. Binomial confidence limits (95%) for prevalence data were determined using the Agresti-Coull calculation and comparisons were conducted using contingency tables and X2 tests. Continuous data were reported as the mean ± SEM, unless otherwise stated and were compared using ANOVA (water quality) or Mann- Whitney tests due to the non-normal distribution of the data (eg. CDI).

6.4 Results 6.4.1 Phase I workshop – Community health perceptions.

Results from the cue card exercise in Phase I workshops in both regions showed that 41% of the 68 small groups identified diarrhea as the top concern for the health of all individuals in communities from the CT region (13/34 groups) and the FV region (15/34 groups) (Table 1). In the CT region this was followed by vomiting, identified by 9 of 34 small groups. Tuberculosis, colds and vomiting were also notable health issues in FV region.

Influence diagrams were created in all 68 small groups and addressed all the health issues listed in Table 1. Given that diarrhea was the predominant concern, the remainder of the results focus on this topic.

Figure 1 provides examples of two influence diagrams created by small groups who identified diarrhea as their top concern. In Fig 1a, the small group showed that diarrhea is caused by a lack of potable water, eating too much fat, certain types of foods, parasites, bad nutrition, lack of hygiene, very sweet foods. Preventive factors for diarrhea were boiling water, washing water storage tanks, medicinal plants, doctor’s care and money. The small group that created the

210

influence diagram in Fig 1b identified drinking rain water, bad food because of poverty, worms, and dirty fruits as causative factors of diarrhea. Boiled water, vegetables and medicine from the health centre were noted as preventive factors.

Based on the set of 28 influence diagrams for diarrhea, factors were grouped into 9 thematic categories. Table 2 shows these thematic categories and provides examples of phrases used on the influence diagrams. Each factor was represented as preventative or causative of diarrhea. Based on how the factor was phrased and the colour of the arrow, the understanding of a preventative or causative relationship emerged (Table 2). In both regions factors associated with the thematic categories of water quality and hygiene were the most commonly cited, followed by factors associated with nutrition in the FV region and with health facilities and parasites in the CT region (Table 2).

Water quality: Groups indicated that poor water quality was a cause of diarrhea (causative relationship) and that boiling water reduced diarrhea (preventative relationship). In the CT region, references to water quality were evenly split between causative and preventive relationships whereas in the FV region, water quality was mentioned more frequently as a preventive factor (ie. Boiling water/treating water would prevent diarrhea) (Table 2).

Hygiene: Hygiene included personal hygiene such as hand washing before eating as well as household cleanliness. These factors were evenly split between causative (ie. Bad hygiene) and preventive (ie. Wash hands before eating) in the CT region but were mentioned more commonly as causative factors in the FV region.

Nutrition: In the FV region, poor nutrition was also considered causative of diarrhea mentioned primarily as a “lack of good food”. In both regions

211

parasites were seen as causative for diarrhea whereas access to health facilities or visiting the doctor was seen as preventive.

Food safety: Food safety was viewed as primarily causative in the CT region (ie not properly cooking food, leaving food uncovered) but in the FV region both causative and preventive components of food safety were identified.

Health facilities: Having access to health facilities was seen as primarily preventive in both regions (ie. Close to the health centre, medicine, seeing the doctor regularly)

Others: Traditional medicine and poverty were only mentioned in the FV region viewed as preventive and causative respectively. Weather was identified as causative in the CT region and the FV region (ie rain) although in one case it was identified as positive in the CT region.

6.4.2 Phase II survey data.

The identification of diarrhea as the top health priority in Phase I workshops led to collection of water quality data and the analysis of longitudinal stool samples to observe patterns of diarrhea prevalence in both regions (Figure 2). Diarrhea prevalence was 18% at the beginning of the study, increasing to just over 40% one month later. In both regions, prevalence remained at approximately 50% until April 2009 before dropping in May 2009, especially in the FV region. Prevalence remained low (13-20%) until the end of the study in October 2009. Chronicity of diarrhea was also similar in both regions; the CDI scores were 0.32 ± 0.02 for children in Soloy and 0.31 ± 0.02 for children in the FV region indicating that approximately 1/3 of fecal samples provided were diarrhetic over the 16 mo period during which fecal samples were collected.

212

Given the high prevalence of diarrhea among preschool children in both regions and the health concern about diarrhea that emerged during Phase I workshops, questionnaires administered to primary caregivers included questions about access to water and knowledge and behaviours concerning diarrhea and water treatment.

Access to aqueducts was significantly higher among study households in the CT region (37%) than in the FV region (4%) (p < 0.001) In the CT region, water samples collected at homes with access to an aqueduct had a significantly lower concentration of E.coli (75±33) than samples obtained from homes that used a ground water source (270±43). Conversely, in the FV region household water quality did not differ between sources likely due to the large variance in E.coli concentration in aqueduct samples (Figure 3). E.coli concentration was similar in both regions for those households relying on ground water sources (Figure 3).

Three specific questions were posed regarding knowledge and practices related to diarrhea and water treatment (Table 3). The majority of participants believed that water treatment would reduce diarrhea (63% in the CT region and 74% in the FV region); however, 36% of respondents in the CT region and 24% in the FV region believed that water treatment would not have an impact on their child’s diarrhea. Three quarters of primary caregivers in the CT region reported not knowing how to use any method of water treatment whereas compared with 41% in the FV region. Of caregivers who reported knowing how to use water treatment methods, caregivers reported that they knew about boiling water before consumption and adding drops of bleach as methods to treat water. More people (p<0.001) reported knowing these methods in the FV region compared with the CT region (Table 3). Despite knowledge of treatment methods, implementation of any water treatment method was rare; 94% and

213

100% of participants replied that they did not treat water before drinking in the CT region and the FV region respectively (Table 3).

6.4.3 Phase III workshop – Barriers to water treatment.

Phase III workshops allowed researchers to discuss Phase II results with regard to diarrhea and water quality with community participants. Workshops were again open to the communities and included a mix of participants from Phase I workshops as well as some who responded to Phase II surveys but had not attended Phase I workshops. General Phase II results on overall prevalence and seasonal changes in diarrhea, on E. coli concentrations in water from aqueducts and ground water were presented at each workshop to provide a regional overview. In addition, community specific data were presented. Examples of felt board stories were: 1) an infected person defecating in the field then rain falling and washing the pathogen into the ground water source, a person then drinks the water and gets infected; or 2) an infected person defecates in the field then a pig rolls in the grass and the parasite sticks to the pig. When the person touches the pig they get the pathogen on their hands and when they eat the pathogen is ingested. Examples of preventive actions discussed after these stories were: 1) purifying the water before drinking; 2) using a latrine and 3) washing hands before eating. The primary preventive actions identified by the small groups were boiling water before drinking or purifying water with chlorine. Discussions that followed focused on when participants treat their water and why participants do not treat their water. Based on notes taken throughout the discussion, participants identified water treatment as “medicine” rather than preventive. A common comment was that participants treated the drinking water when their child was sick. In several communities, participants also reported treating water when it “looked dirty” and said this happened after heavy rain. The most common reasons given for not treating water were: 214

• “It doesn’t taste the same” • “The children won’t drink it” • “Uses too much firewood” • “Bleach is like poison...we don’t know the recipe” • “They do it in the aqueduct” • Cost of bleach

6.5 Discussion In general, the sequential integration of participatory and survey methods demonstrated that community perceptions of health issues and the linkages with environmental factors were in agreement with survey results on frequency of diarrhea and poor water quality. The importance of diarrhea as a health concern in the CT region and the FV region was confirmed by the high prevalence of diarrhea (up to 50%) and the fact that 1/3 of all fecal samples provided over the 16 month study period were diarrhetic In the CT region, water quality was indeed better in homes with aqueduct access which had implications for diarrhea when considered in combination with our previous work that showed higher E.coli concentration in household water increased the risk of chronic diarrhea [20].

Despite community and household awareness of the link between diarrhea and water quality, the majority of participants in the CT region reported that they did not know any methods to treat water. In the FV region, more participants reported knowing how to treat water but nearly half of the respondents reported that they were not aware of any methods to treat water. Knowledge of water treatment methods did not translate into practice in either region. This was especially evident in the FV region where 59% of individuals reported knowledge of water treatment practices but none of the respondents 215

reported treating household water. In the CT region, fewer knew how to treat their water (25%) but still, only 6% actually treated their water before drinking it.

Focus group discussion around water treatment revealed that water treatment was seen more as a curative practice to be used like “medicine” rather than a daily preventive action. This is a very important observation, and one that had not been anticipated in advance of the Phase III workshops. This view may be in part because the increased use of resources (firewood, money) and the unpleasant taste do not make this a desirable daily activity. Furthermore, lack of knowledge on exact methods for treating water with bleach and the perception that bleach is “like poison” highlight areas that should be targeted for health education programs. Thus, qualitative research was particularly beneficial in describing the disconnect between knowledge and practice of water treatment in the study area, in justifying the association that we made between poor water quality and diarrheal diseases, and in highlighting key health education objectives. Specific public health targets that emerged from this study were: 1) messages about household water chlorination should enable individuals to feel confident in their abilities to implement this method correctly; 2) the taste of purified water and extra firewood use associated with boiling water need to be addressed in order to increase the uptake of water treatment messages; 3) public health education needs to promote action rather than just distribute information; and 4) education in aqueduct management for community aqueduct committees would improve the quality and service of community aqueducts. Importantly, the qualitative component of our research was based on participatory methodology in order to involve participants throughout the research process and to provide a more in depth understanding of local health perceptions. The incorporation of participatory methodology into mixed methods research is more common in the fields of geography and the natural resource sciences [10,23,24] than in the health [13,14] and social sciences [11]

216

which have relied primarily on interviews for qualitative perspectives. We believe that the participatory nature of our study design was fundamental to identifying locally appropriate solutions to the health issue that was of most immediate concern to participants.

Household water quality measures were added to the original quantitative study design due to the concern about diarrheal diseases evident in the Phase I workshops. The inclusion of water quality measures proved to be beneficial in understanding the link between environmental factors and preschool child health in our study region. Previously published results based on pooled data from across both the CT region and the FV region have shown that a higher concentration of E.coli in household water samples was associated with a greater chronicity of diarrhea and protozoan infection in preschool children [20]. Furthermore, chronic protozoan infection had negative implications for child anthropometry [20]. The current study uncovered additional important subtleties in relation to water quality and aqueduct access. First, results from the Phase I workshops indicated that participants believed that water quality was better from the aqueduct than from ground water sources and therefore that use of water from the aqueduct would help to prevent diarrhea. Although E.coli data supported the belief that water quality was better from the aqueduct than from groundwater sources, E.coli concentrations in samples from homes with aqueduct access were still well above what is considered safe for drinking water [25]. Second, results from the Phase III workshops indicated that one of the reasons for not treating water at the household level was the belief that water is already “treated” in the aqueduct. Importantly, the aqueducts that provide water to both regions are community managed, above ground, gravity fed water distribution systems that often have exposed and damaged tubes. Water was rarely chlorinated (Personal interviews and field notes). Thus, although water from the aqueduct was less contaminated than water from

217

groundwater sources, the poor quality of water from rural aqueducts needs to be addressed. Otherwise, the prevalence of preschool child diarrhea will remain high in these regions.

Regional comparisons also suggest a link between aqueduct access and attitudes towards water treatment. In the FV region, where less than 5% of the homes had access to an aqueduct, mothers were more likely to believe that water treatment would reduce child diarrhea and to report knowing methods of water purification. We can suggest several possible explanations for these regional differences. First, due to very low aqueduct access in the FV region (<5% of households), average water quality in the FV region (assessed by E.coli concentrations) was worse than in the CT region which may have made the benefits of water treatment for child health more apparent. Second, given the predominant use of ground water in the FV region and the higher percentage of participants who knew how to treat water, it is possible that health messages in the FV region have focused on household strategies for water treatment, although we do not have confirmation of this. In addition to awareness of household treatment strategies as revealed in Phase II questionnaires, communities in the FV region appeared to focus more on the preventive factors of water quality (ie boiling water, treating water), identifying them in their influence diagrams more frequently than causative water quality factors. This may be a result of the type or delivery of health education programs in the FV region and thus could help inform the development of future programs in the comarca.

Our study has several implications for mixed methods research. First, we demonstrated the value of mixed methods research for developing public health interventions by providing insight on the inconsistencies between knowledge and practice of water treatment methods. Specifically, influence diagrams from community workshops and household survey results showed that the majority of 218

participants/respondents believed that water treatment would reduce child diarrhea in their homes and communities. Despite this belief, less than 10% of households reported treating their water before drinking it. Results from household surveys and Phase III workshops suggested that participants were not confident in their understanding of how to implement water treatment, especially adding bleach. Workshop discussions also identified that the change in taste because of chlorination and boiling as well as increased use of firewood when boiling were barriers to the regular treatment of household drinking water. Research from the positive deviance field may provide insights to identify locally feasible and acceptable solutions to these problems. By identifying individuals within the community that practice healthy behaviours despite a similar socio- economic setting, positive deviance based interventions have led to improvements in a range of health problems from child malnutrition [26] to safe sex practices [27]. In addition to identifying methods that are locally feasible, the belief that the solution came from within the community has been shown to provide the social mobilization that facilitates acceptance of the intervention [4]. An initial suggestion that emerged from a Phase III workshop was to add citrus fruit (grown locally throughout the year) to improve the taste.

Second, sequential study designs are rare in mixed methods research. This design allowed us to incorporate multiple phases of qualitative and quantitative work leading to a more integrated understanding of the context surrounding child diarrhea and water quality. In a recent review of mixed methods studies, sequential study designs accounted for only 30% of the 187 studies identified [18] likely due to the significant temporal and financial investment for studies using a sequential design [13,16] . Thus we have contributed to an area in need of expansion in mixed method research. The majority of mixed methods sequential studies begin with a qualitative phase to inform the development of an instrument, questionnaire, hypothesis or program

219

based on the experiences, perceptions and cultures of participants [18,28] and may be most useful when researchers have a low pre-existing knowledge of a research questions or setting [28]. This was similar to the first two phases of our study where the qualitative development of ranked health concerns and corresponding influence diagrams helped to identify the local importance of diarrhea and furthermore the link to water quality. By incorporating an additional qualitative phase we added an explanatory component to our study. The third phase of research helped to explain some of the barriers to preventive health behaviours (water treatment), providing a rich context to our quantitative results that identified locally appropriate targets for public health interventions. Although less common than the exploratory design, the results of our explanatory component add to other health studies that have been useful in explaining barriers to the development of private practice for rural dietitians [29] as well as understanding inadequate post-operative pain management [30] and dietary risk factors for cancer [31].

Third, our study adds to the growing body of literature that includes participatory methodologies within a mixed method design. The integration of qualitative and quantitative data is considered a current focus in mixed methods research and takes three different forms: merging, connecting and embedding data [13]. By identifying and tallying themes in qualitative data we were able to compare local perceptions of health priorities with quantitative measures of diarrhea prevalence. Furthermore, our use of prior datasets to inform subsequent phases of the study connected the qualitative and quantitative analyses. This allowed us to explore local perceptions of health, quantify the occurrence of disease outcomes and related health behaviours and finally gain an in-depth understanding of motivations behind certain health behaviours. Although we were able to integrate our data in multiple ways, the ultimate goal in integrated mixed methods (IMM) research is methodological parallelism [17].

220

In an IMM study integration begins with a unified conceptualization of “research evidence” in the study development phase that gives equal weight to quantitative and qualitative data, during analysis data from both forms of inquiry are transformed into comparable forms and the research process ends with the recontextualization of quantitative data into the original qualitative context [17]. Thus, although we were able to add to the small number of sequential mixed methods studies, our study would have benefitted from more a priori integration of qualitative and quantitative methodologies in order to contribute to the emerging paradigms of integrative mixed methods research.

Additional valuable contextual information emerged during the study through workshops and discussion with study participants and community leaders. Rural aqueducts in the study region are gravity fed systems that often have insufficient water supply in the dry season, causing participants to revert to ground water sources. The rural aqueducts are managed by a volunteer rural aqueduct committee that is in charge of maintenance and administration. MINSA is responsible for providing training and administrative support but this seldom happens [32]. In this context, a final target for public health initiatives was identified. For long term improvements to the quality of drinking water in the region, the low coverage (5% in the FV region) and poor quality of the community water distribution systems must improve. As rural water systems in Panamá are managed by the communities themselves, education in aqueduct maintenance and administration as well as improved communication between MINSA and the community aqueduct committees were identified as areas that would assist this process. A training program has been initiated by Peace Corp volunteers on the north side of the comarca [32,33]; however, it has yet to be implemented in the study area. Future collaboration between Peace Corp or other organizations and MINSA may assist in improving the content and frequency of MINSA community aqueduct committee training.

221

In addition to identifying locally acceptable content for public health interventions, the way that the messages are developed and delivered is important for encouraging behaviour change rather than just providing information [34,35]. Lessons from behaviour change theory, especially social cognitive theory have identified ‘self-efficacy’, the belief of one’s ability to control one’s own health outcomes, as particularly important for the adoption of behaviour change interventions [5,35-37]. Indeed, there is increasing evidence that public health interventions based in social and behavioural theory are more effective than those without a theoretical grounding [7]. Social support, provided through community groups or public health workers has been successful in nutrition and water quality interventions [38-40].

There are several existing opportunities for delivering the identified education messages in the comarca Ngäbe Buglé. The Red de Oportunidades program that is present in both study regions requires attendance at a bimonthly training program that addresses topics ranging from agriculture to money management and preventive health. In addition, the Red de Oportunidades also encourages regular visits to the health facilities where MINSA currently delivers preventive health talks. Thus, results from our initial community workshops, as well as insights obtained during our study could facilitate the development of more appropriate health interventions on water treatment methods as well as community aqueduct committee training that could be delivered through the existing Red de Oportunidades or MINSA programs.

Our study in the comarca Ngäbe- Buglé combined quantitative and qualitative methods to provide a more detailed and comprehensive understanding of the priority health issue in the region. Importantly, this mixed methods study highlighted the disconnect between knowledge and practice related to water treatment. We suggest that future interventions involve community members in the process of identifying and implementing solutions 222

that include self-efficacy building components in order to ensure that preventive health knowledge is translated into health behaviours.

6.6 Acknowledgements

The authors would like to thank the communities and households in the study area for their continued participation over 16 months. Collaborators with Ministry of Health were also invaluable to this project. We are also grateful to the Peace Corps organization and comarca volunteers for their support and discussion of water quality issues in the comarca.

223

6.7 References

1. Fiszbein A, Schady N (2009) Conditional cash transfers: reducing present and future poverty. Washington, DC: World Bank. 383 p.

2. United Nations (2009) Rethinking Poverty: Report on the World Social Situation 2010. Geneva: United Nations.

3. Adato M, Roopnaraine T, Becker E (2011) Understanding use of health services in conditional cash transfer programs: Insights from qualitative research in Latin America and Turkey. Soc Sci Med 72: 1921-1929.

4. Marsh DR, Schroeder DG, Dearden KA, Sternin J, Sternin M (2004) The power of positive deviance. Br Med J 329: 1177-1179.

5. Bandura A (2004) Health promotion by social cognitive means. Health Educ Behav 31: 143-164.

6. Richards CR, Tucker CM, Brozyna A, Ferdinand LA, Shapiro MA (2009) Social and cognitive factors associated with preventative health care behaviors of culturally diverse adolescents. J Natl Med Assoc 101: 236-242.

7. Glanz K, Bishop DB (2010) The role of behavioral science theory in development and implementation of public health interventions. Annu Rev Public Health 31: 399-418.

8. Rennie JK, Singh NC (1996) Participatory Research for Sustainable Livelihoods. Winnipeg, Canada: International Institute of Sustainable Development. 122 p.

9. Chan L, Bundy DAP, Kan SP (1994) Aggregation and predisposition to Ascaris lumbricoides and Trichuris trichiura at the familial level. Trans R Soc Trop Med Hyg 88: 46-48.

10. Parkes M, Panelli R (2001) Integrating catchment ecosystems and community health: The value of participatory action research. Ecosystem Health 7: 85-106. 11. Bryman A (2006) Integrating quantitative and qualitative research: How is it done? Qualitative Research 6: 97-113.

224

12. Johnson RB, Onwuegbuzie AJ, Turner LA (2007) Toward a Definition of Mixed Methods Research. Journal of Mixed Methods Research 1: 112-133.

13. Creswell JW, Klassen AC, Plano Clark VL, Clegg Smith K (2011) Best practices for mixed methods research in the health sciences. National Institutes of Health. Retrieved Nov. 8, 2012: http://obssr.od.nih.gov/mixed_methods_research. .

14. O'Cathain A, Murphy E, Nicholl J (2007) Why, and how, mixed methods research is undertaken in health services research in England: A mixed methods study. BMC Health Services Research 7.

15. Shaw JA, Connelly DM, Zecevic AA (2010) Pragmatism in practice: Mixed methods research for physiotherapy. Physiother Theory Pract 26: 510- 518.

16. Creswell JW (2008) Research Design: Qualitative, quantitative and mixed methods approaches. Nerbraska: Sage Publications.

17. Castro FG, Kellison JG, Boyd SJ, Kopak A (2010) A methodology for conducting integrative mixed methods research and data analyses. Journal of Mixed Methods Research 4: 342-360.

18. Clark VLP (2010) The adoption and practice of mixed methods: U.S. trends in federally funded health-related research. Qualitative Inquiry 16: 428-440.

19. Ministerio de Economía y Finanzas (2010) La Distribución del Ingreso en Los Hogares de Panamá:Encuesta de Niveles de Vida 2008. Panamá: Ministerio de Economía y Finanzas. 34 p.

20. Halpenny CM, Koski KG, Valdés VE, Scott ME (2012) Prediction of child health by household density and asset-based indices in impoverished Indigenous villages in rural Panamá. Am J Trop Med Hyg 86: 280-291.

21. Halpenny CM, Paller C, Koski KG, Valdés VE, Scott ME (2012) A spatio- temporal analysis of soil transmitted helminth reinfection dynamics in preschool children from rural indigenous Panamá. PLoS Neglect Trop D pay: submitted.

22. Waltner-Toews D (2004) Ecosystem sustainability and health: A practical approach. Cambridge: Cambridge University Press. 138 p.

225

23. Johnson NL, Lilja N, Ashby JA (2003) Measuring the impact of user participation in agricultural and natural resource management research. Agricultural Systems 78: 287-306.

24. Adato M, Lund F, Mhlongo P (2007) Methodological Innovations in Research on the Dynamics of Poverty: A Longitudinal Study in KwaZulu-Natal, South Africa. World Dev 35: 247-263.

25. World Health Organization (WHO) (2008) Guidelines for Drinking Water Quality. Geneva: World Health Organization.

26. Marsh DR, Pachón H, Schroeder DG, Ha TT, Dearden K, et al. (2002) Design of a prospective, randomized evaluation of an integrated nutrition program in rural Viet Nam. Food Nutr Bull 23: 36-47.

27. Babalola S, Awasum D, Quenum-Renaud B (2002) The Correlates of Safe Sex Practices among Rwandan Youth: A Positive Deviance Approach. AJAR 1: 11-21.

28. Crookston BT, Dearden KA, Alder SC, Porucznik CA, Stanford JB, et al. (2011) Impact of early and concurrent stunting on cognition. Matern Child Nutr 7: 397-409.

29. Brown LJ, Mitchell LJ, Williams LT, Macdonald-Wicks L, Capra S (2011) Private practice in rural areas: An untapped opportunity for dietitians. Aust J Rural Health 19: 191-196.

30. Carr ECJ (2009) Understanding inadequate pain management in the clinical setting: The value of the sequential explanatory mixed method study. J Clin Nurs 18: 124-131.

31. Klassen AC, Smith KC, Black MM, Caulfield LE (2009) Mixed method approaches to understanding cancer-related dietary risk reduction among public housing residents. J Urban Health 86: 624-640.

32. Suzuki R (2010) Post project assessment and follow-up support for community managed rural water systems in Panama. . MSc Thesis. Houghton, MI: Michigan Technological University. 87 p.

33. Braithwaite B (2009) Training water committees in Bocas del Toro, Panama: A case study of Peace Corps volunteers' initiative to improve rural water system management. MSc Thesis. Houghton, MI: Michigan Technological University. 80 p. 226

34. Valente TW, Paredes P, Poppe PR (1998) Matching the message to the process: The relative ordering of knowledge, attitudes, and practices in behavior change research. Human Communication Research 24: 366-385.

35. Rimal RN (2000) Closing the knowledge-behavior gap in health promotion: The mediating role of self-efficacy. Health Communication 12: X-235.

36. Parker Fiebelkorn A, Person B, Quick RE, Vindigni SM, Jhung M, et al. (2012) Systematic review of behavior change research on point-of-use water treatment interventions in countries categorized as low- to medium- development on the human development index. Soc Sci Med 75: 622- 633.

37. Walker LO, Kim S, Sterling BS, Latimer L (2010) Developing health promotion interventions: A multisource method applied to weight loss among low- income postpartum women. Public Health Nurs 27: 188-195.

38. Ammerman AS, Lindquist CH, Lohr KN, Hersey J (2002) The efficacy of behavioral interventions to modify dietary fat and fruit and vegetable intake: A review of the evidence. Prev Med 35: 25-41.

39. Quick RE, Kimura A, Thevos A, Tembo M, Shamputa I, et al. (2002) Diarrhea prevention through household-level water disinfection and safe storage in Zambia. Am J Trop Med Hyg 66: 584-589.

40. Thevos AK, Quick RE, Yanduli V (2000) Motivational Interviewing enhances the adoption of water disinfection practices in Zambia. Health Promot Int 15: 207-214.

227

Table 1. Summary of the top health priority identified by each of 68 small groups in Soloy and Chorcha through the cue card exercise.

CT region FV region Number of Groups Number of Groups Diarrhea 13 15 Vomiting 9 4 Tuberculosis 2 5 Cold 1 5 Bronchitis 2 2 Parasites 2 0 Malnutrition 0 2 Fever 2 0 Kidney problems 2 0 Head Aches 1 1 TOTAL 34 34

228

Table 2. Summary of 28 influence diagrams (13 from the CT region, 15 from the FV region) on diarrhea developed by small groups during Phase I workshops. Factors included in each influence diagram were grouped into nine thematic categories. For each relationship on the influence diagram, the phrase used by the small group and the colour of the arrow were used to interpret their intention preventative or causative.

Phrases interpreted as Phrases interpreted as preventative (number on Thematic causative(number on influence influence diagrams in the Category diagrams in the CT and FV CT and FV regions, regions, respectively) respectively)

Aqueduct “Dirty” water Water Quality Water treatment (8, 8) (8, 17) Hand washing Lack of hygiene Hygiene Covering food (9, 13) (7, 4) Vitamins Malnutrition Nutrition Vegetables Poor food (2,2) (2,12) Washing food Leaving food uncovered Food Safety Cooking food well (2,12) (2,2) Seeing doctor regularly Health Being far from health centre Being close to health centre Facilities (1,0) (6,10) Worms Parasites (0,0) Parasites (7,10) Change in climate Rain Weather (1,0) (1,2)

Traditional Traditional plants Medicine/Own Homemade rehydration (0,0) Medicine (0,5) Lack of resources Poverty (0,0) Poverty (0,3)

229

Table 3. Summary of water treatment knowledge and practices and link between water treatment and diarrhea symptoms based on questionnaires administered to primary care giver of preschool children.

CT region FV region (n=126) (n=81) If you treat your water does your child have ______diarrhea? More 1% 2% Equal 36% 24% Less 63% 74% Water Treatment Knowledge – “What water treatment method do you know how to use?” None 75% 41% Boiling 10% 26% Bleach 12% 21% Both 3% 12% Water Treatment Practice – “What do you do to your water before drinking it?” Nothing 94% 100% Boil 1% 0% Bleach 5% 0%

230

Figure 1. Examples of cause and prevention influence diagrams for diarrhea from community workshops. Red arrows indicate causes and green arrows indicate preventive factors.

231

60 CT region FV region 50

30 231

20 Prevalence (95% (95% CI) Prevalence

0 Jul-2008 Aug-2008 Oct-2008 Apr-2009 May-2009 Aug-2009 Oct-2009

Figure 2. Longitudinal prevalence with 95% CI of diarrhetic samples collected from preschool children in the CT and FV region (CT region: n=152-175, FV region: n=101-116). .

232

350 CT region b A 300 FV region

250 SE ) 200 concentration

150 A

E.coli a (cfu/100 mL ±

100 Average

0 Aqueduct Ground Water

Figure 3. Average E.coli concentration in water samples collected from households who used aqueduct or ground water sources in the CT and FV regions (CT region: Aqueduct n=46, Groundwater n=79; FV region: Aqueduct n=4, Groundwater n=81).

233

Figure 4. Examples of felt board “stories” created by participants during Phase III workshops.

234

CHAPTER 7

General Discussion

7.1 Major Findings The overall objective of this thesis was to examine the interrelationships between preschool child stunting and gastrointestinal infections within the biophysical, social and spatial context of extreme poverty among the Ngäbe in Western Panama where conditional food voucher (FV) and cash transfer (CT) programs occurred. Incorporating quantitative and qualitative methodologies from parasitology, nutrition and geography provided a more comprehensive perspective on the complex social, spatial and biophysical environments that influenced preschool child health. First, I was able to demonstrate that chronic protozoan infection over the 16 month study period increased stunting and that children from areas with more densely spaced households had more frequent protozoan infection, despite better access to sanitation and hygiene infrastructure. Importantly, HWI counteracted the negative effect of CPI on child HAZ, especially in the more densely populated regions. Second, regional clustering patterns in STH infection were driven by poverty and remoteness whereas reinfection patterns were related to regional, household and individual level factors and differed among STH species. Specifically, Trichuris reinfection was more strongly related to regional prevalence of infection and household exposure variables, whereas stunted children were more susceptible to Ascaris and hookworm reinfection. Third, I demonstrated how type of conditional transfer program (FV vs CT) was a contextual determinant of preschool child food consumption patterns, rate of weight gain and predictors of weight gain and linear growth. Although carbohydrate consumption was detrimental to child 235

linear growth in both regions, the types of carbohydrate dense foods related to lower growth differed by program region. Furthermore, meat consumption improved linear growth, but only in the FV region. Dietary predictors of weight gain also differed by region. Fish and eggs improved weight gain in the FV region whereas milk products were beneficial but sweets and chips were detrimental for weight gain in the CT region. Fourth, quantitative data confirmed community perceptions that diarrhea was a major health issue in the area and furthermore supported the local understanding that household water quality and chid hygiene behaviours influenced diarrhea. The inclusion of participatory methods in this mixed methods approach identified unique aspects to be included in public health interventions that would not have been identified without participant suggestions.

Another strength of this thesis was the comprehensive analysis of preschool child stunting. It is well recognized that the interaction between infection and nutrition is central to child health [1-3]; however, few field studies are able to take an integrated approach that considers diet, anthropometry and multiple types of infection in addition to socio-spatial context variables. The transdisciplinary nature of this investigation allowed us to examine the relationship between stunting and two types of GI infection as well as diet while considering multiple social and spatial factors. Furthermore, due to the longitudinal nature of the study we could examine how duration of infection as well as the dietary factors related to linear growth. These results again highlighted the multiplicity of factors influencing child health as well as the interactions between these factors. Specifically, children with more chronic protozoan infections were more likely to be stunted but household wealth reduced the likelihood of stunting. Importantly, household wealth was particularly beneficial to child health in the dense areas where it was protective against protozoan infection and stunting. By controlling for baseline HAZ status

236

we were able to account for the effect of chronic protozoan infection and HWI on HAZ and examine the unique contribution of diet to linear growth.

In general, carbohydrates that made up a small proportion of the diet (breads and pasta) were beneficial for child growth; however, in the CT region, diets based predominantly on root vegetables were associated with less linear growth. Furthermore, in the FV region meat consumption was related to better growth. Thus, although infection and diet were influential for stunting there were regional differences in how these factors influenced child height.

7.2 Methodological Advances One of the major findings that emerged from this research is the high degree of heterogeneity in health outcomes and health determinants in an area where over 90% of the population lives in extreme poverty. Several methodological tools were instrumental in describing these heterogeneities. First, the development of an asset-based household wealth index allowed us to characterize relative levels of poverty among participating households in a population living in extreme poverty with few possessions. This also allowed me to identify that household wealth could counteract the negative effect of chronic protozoan infection on child HAZ and that household poverty was greater in regional clusters of STH infections than surrounding areas. This adds to the growing body of literature that has demonstrated the utility of asset-based scores in describing health outcomes [4-7] as well as health care service provision [8]. Furthermore, this study demonstrates the ability of asset-based wealth indices to discriminate among households with a limited number of possessions in a way that is relevant to multiple health outcomes. Second, spatial mapping allowed us to explore different ways of defining larger regions that may be influential to health. Spatial cluster analysis has been particularly useful in identifying small-scale trends that differ from larger generalized patterns [9,10]. This was particularly useful in our study to identify areas that 237

had a high prevalence of STH infection and show how this was influential in Trichuris reinfection dynamics. Studies that consider the multi-dimensional determinants of health need to identify a “context” within which to base their investigation. Ultimately this is done by defining a region or “neighbourhood” which should be based on the proposed mechanism of interest [11,12]. For this reason we used household density estimates to examine regional risk of exposure to protozoan and STH infection as well as characterizing levels of infrastructural development in our study area but considered type of conditional transfer program in our investigations of diet. Importantly, in the diet manuscript (Chapter 5) we were also able to use spatial mapping data (ie. density estimates and cluster analysis) and household wealth indices to control for regional differences in poverty and development while investigating for program level effects. Thus, by examining the effect of multiple, relevant regional contexts influential to preschool child health we were able to provide a more complete picture of the multiple pathways influencing child health outcomes.

The qualitative component of our mixed methods approach was particularly beneficial in ensuring that the research results had practical relevance. Specifically, prior to community workshops measures of diarrhea and water quality were not included in the research design. It was only after diarrhea was highlighted as a ubiquitous health priority for the communities that it was included as a quantitative measure. Furthermore, discussions that began in community workshops and continued informally during the field work highlighted the importance of investigating water treatment knowledge and practice as well as perceptions of the link between water treatment and child diarrhea. By including water quality measures in the quantitative study we were able to demonstrate the poor quality of household water in the study area (aqueduct and ground water sources) and identify the increased risk of diarrhea

238

and protozoan infection associated with contaminated water. Furthermore, through follow-up in household questionnaires and Phase III workshops we identified the gap between knowledge of water treatment practices and the application of this knowledge. We also identified reasons that the locally recommended water purification methods were not implemented. Future research into locally acceptable methods for reducing firewood consumption while boiling water and improving the taste of chlorinated water would further help to improve child health in the region. Importantly, our study used participatory methodologies during the qualitative components of our study in order to ensure local participation throughout the research process and to identify locally appropriate solutions to priority health issues.

7.3 Reflections on the Methodological Approach 7.3.1 Transdisciplinarity

The study of the interrelationships among gastrointestinal infection and malnutrition lends itself well to the adoption of an approach that considers more than one discipline. Even at the individual level, studying this interrelationship requires an understanding of parasitology, nutrition and immunology. In order to move beyond multi-disciplinary into a transdisciplinary approach, this study sought to consider how individual through to societal level factors influenced the gastrointestinal infection – malnutrition relationship. By examining multiple levels we were able to detect the different effects of individual, household and regional level factors on preschool child infection and malnutrition outcomes. More importantly for a transdisciplinary perspective, we were able to identify interactions between these influential factors [13,14].

Another distinguishing characteristic of a transdisciplinary study is the creation of a unified conceptual framework that is informed by multiple 239

disciplines [15]. The development of the conceptual framework for this thesis required extensive reading across disciplines. Throughout the process, concepts and methodologies from specific fields were integrated around the central theme of identifying the interactions between biophysical, social and spatial factors that influence preschool child nutrition and GI parasite reinfection. Although this process was conducted primarily as an individual endeavour, discussion with researchers and students from other fields was integral to creating a more comprehensive framework. Indeed, as noted by Kessel and Rosenfield [13], institutional support for transdisciplinary training, a lack of barriers between the disciplines and productive communication are necessary for effective transdisciplinary research. Two specific facets of my research experience standout as particularly beneficial in developing a transdisciplinary research framework. First, I was part of the McGill School of Environment Graduate Option which brought together students from a broad range of disciplines to discuss environmental issues and our research projects from multiple perspectives. Consequently, I developed communication skills that facilitated transdisciplinary communication and engaged in discussion with researchers from other fields about our research projects. Second, my involvement with the Community of Practice in EcoHealth (CoPEH) was beneficial in creating partnerships and collaborations with transdisciplinary researchers as well as identifying examples of Ecohealth projects, two commonly identified barriers to transdisciplinary [13] and Ecohealth [16] research.

7.3.2 Participation

Transdisciplinary research, especially Ecohealth research, emphasizes the importance of including multiple perspectives in order to capture the rich complexities of health processes and outcomes [13,17,18]. Importantly, local perspectives, and thus local participation in the research process, are considered central to a project that takes on an Ecohealth approach. For this reason 240

Panamanian perspectives that ranged from the national governmental level to the community participant level were integral to the research process. The value of this participation was evident in the development of a funding proposal that was readily accepted by the national research funding agency of Panamanian because it touched upon locally relevant health priorities. The National Ministry of Health priority on child malnutrition, community concerns over diarrhea and water quality and the desire of Ministry of Social Development to examine their Conditional Transfer programs helped shape the research project and ensure collaboration among government and community members alike. Importantly, the value of this participation goes beyond the development of a locally relevant project. Specifically, partnerships and collaborations initiated during this process have already led to future joint projects. This was particularly evident in the relationship between Ministry of Health officials and community health promoters. Initial distrust of the ability to accomplish this research in indigenous communities was eventually replaced with a respectful working relationship. The importance of relationships for the practice of Ecohealth was recently identified among public health employees in Ontario [16]. Similar to the Panamanian project, formal and informal relationships from the grassroots to government levels helped to overcome challenges of implementing a “system perspective” and aided in working towards a longer-term vision of health promotion. At the community level this has been shown to improve the abilities of groups to take a proactive stance on health [16], to improve their capacity to respond to future challenges [19] and to promote social justice [20].

Central to productive participation throughout this thesis research was the need to value “multiple ways of knowing” [14]. A recognized struggle in Ecohealth research [21], determining how to not only understand but also value different epistemologies was present throughout the research process. The inclusion of mixed methodology in this thesis research aided in giving structure

241

to the research design and consequently helped in the valuation of quantitative and qualitative modes of inquiry. Although this is a commonly recognized value of mixed methods research [22,23] the majority of studies that employ this methodology rely on simple interview techniques and see it as a way of overcoming academic disciplinary differences in knowledge generation [24,25]. In addition to the gap between the social and medical sciences, there is a growing recognition of the importance of community based participatory research especially in the health sciences [18,26,27]. This thesis research adds to the body of work that uses participatory research in an effort to include culturally distinct views; however, it is recognized that there is still a long way to go in overcoming the disconnect between different cultural epistemologies [28].

7.4 Limitations It is important to recognize that this study has several limitations. First, our inclusion criteria of having ever participated in the Red de Oportunidades program may have introduced a selection bias. It is believed however, that the study population is representative of the region since over 90% of households are eligible to receive the CT program in the study area. Unfortunately, we were not able to compare our sample population to the overall population as the larger census that was used to identify eligible households did not record information from those without preschool children or who had not participated in CT.

The inclusion of multiple children per household could have inflated the influence of household level variables in our statistical models. To adjust for this fact, we conducted total population models in addition to index child models for the first manuscript (Chapter 3) and we used the Huber estimator for robust standard error estimation to account for clustering at the household level in the subsequent chapters.

242

Our estimates of household density were limited to participant households thus reflecting the density only of households with children <5 years of age. Although this led to an underestimation of the actual density of households, we believe it is of specific health relevance as the proximity of preschool children has implications for child to child transmission.

Seasonal work-related migration also influenced sample size throughout the study and with particular consequences for STH manuscript (Chapter 4). Specifically, it reduced the number of preschool children available at the end of reinfection Cycle 1 but fortunately many children were at home three weeks later at which time they received treatment and were included in reinfection Cycle 2. Hence we were not able to verify efficacy of the single dose of Albendazole in these children. It was also of note that a single Albendazole treatment did not remove all Trichuris and hookworm parasites. Although we were able to give a second dose of Albendazole we were unable to assess whether the second dose of Albendazole successfully cleared infection and thus we cannot be certain that the reinfection epg estimates for Trichuris and hookworm measure reinfection rather than existing infection.

Our use of a novel chronicity index for diarrhea and protozoan infection differs from the measures of duration and frequency of symptomatic episodes used in child health literature. Despite differing from conventional methods, our easily compiled indices based on MINSA field diagnostic procedures proved valuable in assessing the developmental impact of the frequency of illness over a 16 month period.

Our comparison of conditional transfer programs may be limited by several factors. Although many households within the CT region were similarly remote and impoverished to those within the FV region [29], there were socio-economic differences between the regions that may have influenced diet and

243

anthropometric outcomes. We accounted for these factors by including household density and household wealth indices in our models of HAZ and WAZ. An additional factor to consider during interpretation of the nutrition results is the fact that the FV program was initiated 6 months prior to the CT program. Thus, if the rate of improvement in growth increases with time in the program than it is possible that preschool child weight gain in the CT region improved in the months after the study to a level that was similar to the FV region after a similar duration of participation in the CT program.

Without a control region where either program had not yet started or without baseline measures in the CT and FV regions we were unable to comment on the degree of improvement in nutrition and infection outcomes related to the FV and CT programs. Unfortunately, baseline measures were not taken prior to program initiation and upon arriving in Panamá it became clear that there were no regions in the corregimiento of Besiko that were not enrolled in either conditional transfer program.

Our study of STH reinfection (Chapter 4) relied primarily upon descriptive methods to examine spatial patterns of STH reinfection which were incorporated into models of reinfection. It should be noted that these methodologies have limitations. Specifically, the use of cluster analysis should be considered an exploratory method for identifying spatial patterns that can be used to generate further hypotheses around the causes of these clusters [17,30]. By including residence in a high risk cluster and household density in our negative binomial models we considered two spatial variables that may increase exposure to infection but we do not account for the ways in which spatial location may relate to variation in risk factors for reinfection [31]. As such, our spatially implicit models may lead to spurious associations between risk factors and infection outcomes. Spatially explicit methods in epidemiology provide an alternative approach that avoids the risk of spurious associations by accounting for the 244

spatial variation in infection risk factors (ie vector or reservoir distribution) [17,32]. Models of STH distribution and infection that employed spatially explicit methods have shown the association of socio-economic and ecological (ie vegetation, rainfall) variables while accounting for the natural spatial variation that occurs in these risk factors [7,33-35]. Thus, our findings are in line with the more sophisticated spatial models; however, definitive conclusions on the spatial patterns of reinfection in our region would require additional analyses.

7.5 Public Health Implications Taken together our results have several important implications for public health. First and foremost, we have demonstrated the multi-dimensional nature of preschool child health in the comarca Ngäbe Buglé. Thus, comprehensive health interventions that address diet as well as sanitation and hygiene will have the biggest impact on preschool child health. There are several discipline specific messages that have emerged from this thesis and could help inform a more holistic intervention. To improve child linear growth and weight gain, dietary interventions should promote the reduction in the proportion of preschool child meals that come from carbohydrates and also promote meat and fish intake. Furthermore, especially in the CT region, the intake of sweets and chips should be replaced with healthier alternatives to ensure children receive the nutrients necessary for growth. Importantly, the resulting improvements in child anthropometry will also reduce susceptibility to STH infections. However, to further reduce STH infections in the region, regular preventive chemotherapy should be provided to preschool children in addition to school age children as they are significantly contributing to the transmission of infection in the region. It is also of note that monitoring the efficacy of ABZ in the region will be particularly important for Trichuris infection which is already showing signs of reduced efficacy [36]. In addition to dietary and chemotherapeutic interventions, improvements to regional and household sanitation infrastructure

245

and hygiene behaviours will have a significant effect on protozoan and diarrheal infection as well as STH transmission. This is especially important given the consequences of infection on child stunting. Finally, two programmatic considerations have also become evident through this thesis work. First, the regional differences we demonstrated in diet and infection outcomes suggest the need to consider context in informing interventions. Specifically, the types of foods promoted should consider the current diet in the area as well as what foods are available and infection intervention messages should take into consideration the primary infections in the region (ie water borne vs soil transmitted infections). Finally, we have demonstrated the importance of local perspectives for identifying health priorities and locally relevant methods for addressing these problems.

246

7.6 References

1. Koski KG, Scott ME (2001) Gastrointestinal nematodes, nutrition and immunity: Breaking the negative spiral. Annu Rev Nutr 21: 297-321.

2. Hall A, Hewitt G, Tuffrey V, De Silva N (2008) A review and meta-analysis of the impact of intestinal worms on child growth and nutrition. Matern Child Nutr 4: 118-236. 3. Dewey KG, Mayers DR (2011) Early child growth: How do nutrition and infection interact? Matern Child Nutr 7: 129-142.

4. Nundy S, Gilman RH, Xiao L, Cabrera L, Cama R, et al. (2011) Wealth and its associations with enteric parasitic infections in a low-income community in Peru: use of principal component analysis. Am J Trop Med Hyg 84: 38- 42.

5. Van De Poel E, Hosseinpoor AR, Speybroeck N, Van Ourti T, Vega J (2008) Socioeconomic inequality in malnutrition in developing countries. Bull World Health Organ 86: 282-291.

6. Hong R, Mishra V (2006) Effect of wealth inequality on chronic under-nutrition in Cambodian children. J Health Popul Nutr 24: 89-99.

7. Raso G, Vounatsou P, Gosoniu L, Tanner M, Goran E, et al. (2006) Risk factors and spatial patterns of hookworm infection among schoolchildren in a rural area of western Cote d'Ivoire. Int J Parasitol 36: 201-210.

8. Gwatkin D, Wagstaff A, Yazbeck A (2005) Reaching the poor with health nutrition and population services: What works, what doesn't and why. Washington, DC: World Bank. 354 p.

9. Gaudart J, Poudiougou B, Dicko A, Ranque S, Toure O, et al. (2006) Space-time clustering of childhood malaria at the household level: A dynamic cohort in a Mali village. BMC Public Health 6.

10. Odoi A, Martin SW, Michel P, Middleton D, Holt J, et al. (2004) Investigation of clusters of giardiasis using GIS and a spatial scan statistic. International Journal of Health Geographics 3.

11. Duncan C, Jones K, Moon G (1998) Context, composition and heterogeneity: Using multilevel models in health research. Soc Sci Med 46: 97-117.

247

12. Diez Roux AV (2001) Investigating neighborhood and area effects on health. Am J Public Health 91: 1783-1789.

13. Kessel F, Rosenfield PL (2008) Toward Transdisciplinary Research. Historical and Contemporary Perspectives. Am J Prev Med 35: S225-S234.

14. Annerstedt M (2010) Transdisciplinarity as an inference technique to achieve a better understanding in the health and environmental sciences. International Journal of Environmental Research and Public Health 7: 2692-2707. 15. Rosenfield PL (1992) The potential of transdisciplinary research for sustaining and extending linkages between the health and social sciences. Soc Sci Med 35: 1343-1357.

16. Leung Z, Middleton D, Morrison K (2012) One Health and EcoHealth in Ontario: A qualitative study exploring how holistic and integrative approaches are shaping public health practice in Ontario. BMC public health 12.

17. Waltner-Toews D (2004) Ecosystem sustainability and health: A practical approach. Cambridge: Cambridge University Press. 138 p.

18. Dankwa-Mullan I, Rhee KB, Stoff DM, Pohlhaus JR, Sy FS, et al. (2010) Moving toward paradigm-shifting research in health disparities through translational, transformational, and transdisciplinary approaches. Am J Public Health 100: S19-S24.

19. Spiegel J, Bonet M, Garcia M, Ibarra AM, Tate RB, et al. (2004) Building capacity in central Havana to sustainably manage environmental health risk in an urban ecosystem. Ecohealth 1: 120-130.

20. Craig G, Mayo M, editors (2004) Community empowerment: A reader in participation and development. New Jersey: Zed Books Ltd.

21. Williams-Blangero S, VandeBerg JL, Subedi J, Jha B, Corrêa-Oliveira R, et al. (2008) Localization of multiple quantitative trait loci influencing susceptibility to infection with Ascaris lumbricoides. J Infect Dis 197: 66- 71.

22. Creswell JW, Klassen AC, Plano Clark VL, Clegg Smith K (2011) Best practices for mixed methods research in the health sciences. National Institutes of

248

Health. Retrieved Nov. 8, 2012: http://obssr.od.nih.gov/mixed_methods_research. .

23. Creswell JW (2008) Research Design: Qualitative, quantitative and mixed methods approaches. Nerbraska: Sage Publications.

24. Clark VLP (2010) The adoption and practice of mixed methods: U.S. trends in federally funded health-related research. Qualitative Inquiry 16: 428-440.

25. O'Cathain A, Murphy E, Nicholl J (2007) Why, and how, mixed methods research is undertaken in health services research in England: A mixed methods study. BMC Health Services Research 7. 26. Parkes M, Panelli R (2001) Integrating catchment ecosystems and community health: The value of participatory action research. Ecosystem Health 7: 85-106.

27. Westhues A, Ochocka J, Jacobson N, Simich L, Maiter S, et al. (2008) Developing theory from complexity: Reflections on a collaborative mixed method participatory action research study. Qual Health Res 18: 701-717.

28. Wilcox B, Kueffer C (2008) Transdisciplinarity in EcoHealth: Status and future prospects. EcoHealth 5: 1-3.

29. Halpenny CM, Koski KG, Valdés VE, Scott ME (2012) Prediction of child health by household density and asset-based indices in impoverished Indigenous villages in rural Panamá. Am J Trop Med Hyg 86: 280-291.

30. Elliott P, Wartenberg D (2004) Spatial epidemiology: Current approaches and future challenges. Environ Health Perspect 112: 998-1006.

31. Olsen SF, Martuzzi M, Elliott P (1996) Cluster analysis and disease mapping - Why, when, and how? A step by step guide. Br Med J 313: 863-866.

32. Ostfeld RS, Glass GE, Keesing F (2005) Spatial epidemiology: An emerging (or re-emerging) discipline. Trends in Ecology and Evolution 20: 328-336.

33. Pullan RL, Bethony JM, Geiger SM, Cundill B, Correa-Oliveira R, et al. (2008) Human helminth co-infection: Analysis of spatial patterns and risk factors in a Brazilian community. PLoS Neglect Trop D 2: e352.

34. Brooker S, Kabatereine NB, Tukahebwa EM, Kazibwe F (2004) Spatial analysis of the distribution of intestinal nematode infections in Uganda. Epidemiol Infect 132: 1065-1071. 249

35. Brooker S, Alexander N, Geiger S, Moyeed RA, Stander J, et al. (2006) Contrasting patterns in the small-scale heterogeneity of human helminth infections in urban and rural environments in Brazil. Int J Parasitol 36: 1143-1151.

36. Diawara A, Drake LJ, Suswillo RR, Kihara J, Bundy DAP, et al. (2009) Assays to detect β-tubulin codon 200 polymorphism in Trichuris trichiura and Ascaris lumbricoides. PLoS Neglect Trop D 3: e397.

250

Appendix

Research Questionnaires

The following questionnaires have been translated from Spanish for the purpose of the thesis.

251

A Multidisciplinary Study of the Interaction between Malnutrition and GI Parasitic Infections in Rural Panamá

Water Questionnaire

Date (dd-mm-yy)

Primary Caregiver Code M Interviewer

1. a) In the last months, where does the water in your home come from? (since Februrary)

Source

b) Is this the same source that you use in the rainy season??

□ – Yes □ – No

2 a) Can you show me the container where you store your water? How many do you use a day?

Type of Number □ –Yes Volume Covered container x day □ –No b) How many people share this water? ______

3 Do you drink the water as is? If not, what do you do?

What? □ – □ – No

4. a) What methods do you know to purify your water? (Ask the exact process) ______

252

A Multidisciplinary Study of the Interaction between Malnutrition and GI Parasitic Infections in Rural Panamá

Water Questionnaire b) If you purify your water daily, do you think your children will have diarrhea more often, the same amount or less often?

– More – Equal – Less □ □ □

Comments/Observations

Verified by ______Date: ______

253

A Multidisciplinary Study of the Interaction between Malnutrition and GI Parasitic Infections in Rural Panamá

Health Questionnaire

Date (dd-mm-yy)

Primary Caregiver Code M Interviewer

1. In the last month can you tell me how many times your child has had: C1 C2 C3 Skin infection Respiratory problems Coughed more than normal Vomiting Fever Diarrhea Ear infection Eye infection

2.a) Does this house have a latrine?

□ – Yes □ – No b) If yes, do you use it? If not, what do you go to the bathroom?

Don’t Use Alternative use Adults □ □ Child 1 □ □

254

A Multidisciplinary Study of the Interaction between Malnutrition and GI Parasitic Infections in Rural Panamá

Health Questionnaire

3. a) Do your children have shoes, boots or sandals? How often do they wear them?

Always Sometimes Very little Don’t use Child 1 □ – Yes □ – No

4 a) Do your children wash their hands before eating? With soap? Wash Hands With Soap Child 1 □ – Yes □ – No □ – Yes □ – No b) Can you show me the soap you use? □ – Has soap □ – Doesn’t have soap

Verified by: ______Date:______

255

A Multidisciplinary Study of the Interaction between Malnutrition and GI Parasitic Infections in Rural Panamá

Socio-economic Questionnaire

Date (dd-mm-yy)

Primary Caregiver Code M Interviewer

1. Does this house have a tap?

□ Si □ No

2. What fuel do you use most frequently to cook? □ – Gas □ - Charcoal □ – Wood □ – Kerosene □ – Other ______

3. Which of the following ítems do you have in your house? How long have you had them?

Articulo Tiempo Poseído (años)

□ – Radio

□ – Cell Phone

□ - Bicycle

□ – Horse

□ - Shovel □ - Axe □ - Hatchet □ - Chainsaw □ - Sewing Machine □ - Stove

4. a) Does this house receive food vouchers or cash transfers?

Food Vouchers Cash Transfer

256

A Multidisciplinary Study of the Interaction between Malnutrition and GI Parasitic Infections in Rural Panamá

Socio-economic Questionnaire b) Who receives them? (First and last name)

______

5. a) What is the house made of?

Walls □ – Blocks □ – Zinc □ – Cane □ – Branches □ – Wood □ – Other ______

Floor □ – Soil □ – Cement □ – Wood □ – Other ______

Roof □ – Zinc □ – Wood □ – Palm Leaf □ – Other ______

How many separate rooms are in the house (include any outbuildings) ______

Verified by: ______Date:______

257