Globally over 249 million children in low- and middle-income countries (LMICs) are at risk of not meeting developmental milestones on the motor, cognitive, language or socio-emotional domains. This compromised development may result in less educational attainment and lower economic earnings later in life. The nurturing care framework, developed by the World Health Organization, summarizes the inputs required to promote healthy child development. These inputs include good health, nutrition, responsive caregiving, access to early learning, and safety and security. In pursuit of improved child development, caregiver support interventions have been developed and implemented in numerous LMICs, and have succeeded at improving child language and cognitive development. Most of these interventions, however, were designed to be implemented on a small scale and focus primarily on access to early learning and responsive caregiving.

This dissertation aims to inform the design and evaluation of large-scale interventions to improve child development over the life-course in low-resource settings. In Chapter 1, I evaluate the effects of a multicomponent intervention, that covers early learning, responsive caregiving nutrition, water, sanitation, hygiene, caregiver mental health and lead exposure prevention on child development and risk factors for poor child development. I use data from a cluster randomized controlled trial of an intervention that was delivered primarily in groups in Kishoreganj, Bangladesh. This research contributes to existing literature on the effects on caregiver support interventions on early child development and expands the scope of these interventions to include additional components that may result in larger and more sustained impacts on child outcomes. In Chapter 2, I assess the concurrent correlation between two measures of child development: (i) the Bangladeshi-adapted Ages and Stages Questionnaire: Inventory and (ii) the Bayley Scales of Infant Development-III. The Ages and Stages Questionnaire: Inventory is primarily a caregiver report measure that is feasible to implement to evaluate intervention effects when financial, human, and time resources are constrained, as may be the case with large-scale evaluations. The Bayley Scales of Infant Development-III is a direct assessment measure for child development that has been adapted for use in Bangladesh, and is considered to be less biased than caregiver report in evaluating the effects of early child development interventions. I use endline data from the same cluster randomized trial used in Chapter 1 to examine correlations overall, by age, and by intervention status. This work illustrates the potential of the Bangladeshi-adapted Ages and Stages Questionnaire: Inventory for use in the evaluation of large-scale child development interventions. Finally, in Chapter 3, I describe how the distribution of risk factors for poor child development shifted in response to experiences during the COVID-19 pandemic in Chatmohar, Bangladesh. I analyze data from the evaluation of a large-scale multicomponent child development intervention in Chatmohar, Bangladesh. This analysis highlights the shift of intervention targets for improving population-level early child development in the context of the COVID-19 pandemic. Taken together, this dissertation contributes to the evidence for the design and measurement of effective and scalable interventions to improve child development in LMICs.

The success of Conditional Cash Transfer (CCTs) program in several health and social domains has led to recent experiments testing CCTs for sexual and reproductive health. However, this approach has yielded mixed results and has been understudied. My dissertation explores whether sexual relationship power - the power women have to decide if and when to have sex, with whom, and with or without a condom - influences the effectiveness of CCTs to reduce STIs, and additionally, if these CCTs influence relationship power. If we can answer these questions, researchers will be better able to design CCT interventions to reduce the spread of STIs and HIV.

My overall hypotheses are: (1) CCTs given to women with high baseline relationship power are more effective at reducing STIs than CCTs given to women with low baseline relationship power (Chapter 1); and (2) that the CCT itself changes the relationship power of those enrolled in the intervention (Chapter 2). A relevant example of women’s power to negotiate safer sex is female sex workers’ interactions with their clients. Therefore, in chapter 3, I draw on theoretical models and previous literature related to the determinants of safer sex among commercial sex workers to analyze qualitatively how Female Sex Workers (FSWs) respond to a novel pilot study using CCTs to incentivize safe sex (Chapter 3).

To test these hypotheses, I conducted three studies: The first explores the effect modification of baseline relationship power on the impact of CCTs to reduce STIs; the second is an intent-to-treat analysis of the effect of the CCT on relationship power; and the third is a qualitative analysis of how female sex workers respond to a pilot CCT intervention to incentivize safe sex. Chapters 1 and 2 draw on data from the RESPECT study in Tanzania and Chapter 3 draws on data from the RESPECT II Pilot study to reduce STIs and HIV among female sex workers in Tanzania.

My analysis reveals that women’s relationship power significantly modifies the effect of the CCT on STIs. In addition, I show that a CCT improved women’s relationship power, but whether or not women were eligible to receive a high, low, or no cash transfer had little relative impact on changes in their relationship power. Finally, I identify the salient domains of power for female sex workers that are necessary to meet the conditions of a CCT intervention.

Taken together, acknowledging and addressing both the degrees of relationship power and domains of relationship power that women experience when enrolled in a CCT will improve the outcomes of these interventions in the context of STIs and HIV.

The age-standardized incidence of cardiovascular disease in high-income countries has fallen precipitously from 844 cases per 100,000 in 1990 to 597 in 2019. However, this reduction in average incidence obscures significant disparities in risk by socioeconomic, genetic, biological, and health behavioral factors. To date, most studies of the risk factors for cardiovascular disease attempt to isolate the effect of a single risk factor. This single-exposure focus belies the dynamic interplay between genetics, socioeconomic status, and traditional health and health behavioral factors that ultimately determine an individual’s cardiovascular disease risk. This dissertation explores the ways in which different domains of risk interact to produce heterogeneous, individualized risk of cardiovascular disease.

The second chapter of this dissertation employs Mendelian randomization, an instrumental variable technique that uses random genetic variation as the instrument, to understand the causal effect of adiposity on cardiovascular disease incidence for individuals at different levels of socioeconomic status. Both high adiposity and low socioeconomic status increase cardiovascular disease risk. However, it is unclear whether the risk caused by an increase in adiposity is itself identical for individuals at varying levels of socioeconomic status or if, instead, individuals at lower levels of socioeconomic status face a greater risk from an increase in adiposity than their peers. If such a disparity did exist, it would imply that the risk implications of weight gain may differ by socioeconomic status. In UK Biobank data, we find that differences in risk in these Mendelian randomization models existed only for individuals with versus without a university degree. The differences in cardiovascular disease risk from higher body mass index by educational attainment or income – if they exist - are small in magnitude, though imprecisely estimated. The results for waist-to-hip ratio adjusted for body mass index, a measure of central adiposity, generally showed even smaller differences between socioeconomic groups.

The third chapter of this dissertation explores whether a health behavior related to lower cardiovascular disease risk, physical activity, might ameliorate some of the risk of coronary artery disease caused by an individual’s genetic predisposition. Specifically, I use wrist-worn accelerometer data combined with the most powerful polygenic risk score in the literature to determine how physical activity volume and intensity impact an individual’s risk of coronary artery disease at different levels of genetic risk. If individuals at higher risk benefit more from greater physical activity than their lower risk peers, this could imply that more personalized physical activity standards should be considered for at-risk patients.Physical activity volume and intensity each had significant independent associations with incident coronary artery disease, with physical activity intensity demonstrating the strongest association. However, no interactive effect of physical activity and genetic risk on coronary artery disease risk was found.

Future research should focus both on the impact of individual risk factors on cardiovascular disease risk and - particularly - how different risk factors and domains might contribute and interact to determine an individual’s overall cardiovascular disease risk. With the proliferation of large-scale cohort datasets such as the UK Biobank, the Million Veteran Program, and All of Us, researchers face an unprecedented opportunity to understand how every aspect of a person’s life from their social and built environment to their medical care access, genetic code, and health behaviors interconnect to determine disease risk.

Objectives Poverty reduction interventions through cash transfers and microcredit have had mixed effects on mental health. In this quasi-experimental study, we evaluate the effect of a living wage intervention on depressive symptoms of apparel factory workers in the Dominican Republic.

Setting Two apparel factories in the Dominican Republic.

Participants The final sample consisted of 204 hourly wage workers from the intervention (99) and comparison (105) factories.

Interventions In 2010, an apparel factory began a living wage intervention including a 350% wage increase and significant workplace improvements. The wage increase was plausibly exogenous because workers were not aware of the living wage when applying for jobs and expected to be paid the usual minimum wage. These individuals were compared with workers at a similar local factory paying minimum wage, 15–16 months postintervention.

Primary outcome measures Workers’ depressive symptoms were assessed using the Center for Epidemiologic Studies-Depression Scale (CES-D). Ordinary least squares and Poisson regressions were used to evaluate treatment effect of the intervention, adjusted for covariates.

Results Intervention factory workers had fewer depressive symptoms than comparison factory workers (unadjusted mean CES-D scores: 10.6±9.3 vs 14.7±11.6, p=0.007). These results were sustained when controlling for covariates (β=−5.4, 95% CI −8.5 to −2.3, p=0.001). In adjusted analyses using the standard CES-D clinical cut-off of 16, workers at the intervention factory had a 47% reduced risk of clinically significant levels of depressive symptoms compared with workers at the comparison factory (23% vs 40%).

Conclusions Policymakers have long grappled with how best to improve mental health among populations in low-income and middle-income countries. We find that providing a living wage and workplace improvements to improve income and well-being in a disadvantaged population is associated with reduced depressive symptoms.