Skip Navigation

Delta Omega

2012 Delta Omega Scholarship Winners

Four winners were chosen.


Anna Ettinger, PhD Student
Parenting trajectories and child obesity in low-income ethnic minorities


Childhood overweight and obesity is a leading public health problem in the US, as estimated 17% of children are overweight or obese, with approximately 10% of infants/toddlers to 18% of adolescents having a high body mass index. Being overweight or obese during childhood is related to increased risk of a variety of chronic health conditions in adulthood, including metabolic syndrome, type 2 diabetes, and cardiovascular disease. Child overweight/obesity is also associated with psychosocial and behavioral problems, such as low self-esteem, increased depression, and reduced quality of life. Socioeconomic and racial/ethnic disparities in child obesity have emerged, as low-income, Black, and other ethnic minority children and adolescents have higher rates of overweight/obesity than higher income or White children. While childhood overweight/obesity is a complex and multifactorial problem, it has been linked to caregiving/parenting styles and practices. Experiencing early maternal sensitivity has been associated with decreased risk of obesity in adolescence, possibly through its effect on modulating children’s stress response system. Studies of preschool children have found that those experiencing authoritarian (characterized by strict discipline) or permissive and neglectful parenting are more likely than children of authoritative (characterized by warmth and ageappropriate limits) mothers to be overweight. Responsive and non-responsive parent feeding practices have also been linked to both under- and over-nutrition in children. Variation in parenting styles and feeding practices among different racial/ethnic groups may help explain some of the differences in obesity rates, although research has not been conclusive. Often characterized as a stable trait, parenting quality can change over time in response to changing environmental contexts, parent characteristics, and child developmental needs. However, few studies have examined changes in parenting over time, particularly among ethnic minorities, or its relation to child obesity. Recent advances in latent growth curve analysis, including growth mixture modeling (GMM), provides a tool for modeling different trajectories of parenting over time. Understanding how parenting trajectories impact child health is important for developing effective family policies and interventions, promoting child resilience, and reducing disparities in child obesity. This study will meet this need by characterizing parenting trajectories over time, identifying the factors influencing these trajectories, and analyzing how these patterns influence child overweight/obesity in a diverse, high-risk population.

The aims of this project are to:

1.) Analyze the stability and variation in parenting quality during early and middle childhood among low-income Black and Latino families,
2.) Identify the family, maternal, and child factors associated with these parenting trajectories, and
3.) Examine the relation between parenting trajectories and child growth and overweight/obesity.

This project involves secondary data analysis of the three waves of data from the Welfare, Children, and Families: A Three-City Study. The study follows a random sample of approximately 2,400 low-income (below 200% Federal poverty level) households with children who were either 0-4 or 10-14 years old in Boston, Chicago, and San Antonio from 1999-2005. This project involves secondary data analysis of the three waves of data from the Welfare, Children, and Families: A Three-City Study. The study follows a random sample of approximately 2,400 low-income (below 200% Federal poverty level) households with children who were either 0-4 or 10-14 years old in Boston, Chicago, and San Antonio from 1999-2005.


The Three City-Study used a multistage, stratified-random sampling procedure, sampling at varying rates based on income, race/ethnicity, family structure and receipt of public assistance. This analysis will focus on approximately 1200 children who were 0-4 years old at Wave 1.

Aims 1 and 2: The goals of these aims are to examine variability in parenting over time and identify different trajectories of parenting and the factors influencing those trajectories. Trends in self-reported and observed measures of parent-child interactions and parenting dimensions, including warmth, cognitive stimulation, and discipline, will first be examined across the three waves of data using repeated measures analysis of variance (ANOVAs) to identify any significant population-level changes in parenting. Pearson correlations will be conducted to test the relative stability (ranking of parents) over time on parenting dimensions. Conventional growth curve modeling will be employed as a first step to identify the appropriate number of distinct classes of parenting trajectories. GMM will then be used to characterize the heterogeneity or stability in parenting dimensions over time (e.g., increasing, decreasing, consistently high or low). While conventional growth curve models assume that individuals belong to a single population and covariates influence trajectories equally, GMM allows for different classes of parenting trajectories and allows variables to differentially influence each trajectory. GMM will also be utilized to identify significant predictors of particular trajectories, such as family stability and income, maternal mental health, and child characteristics through multinomial logistic regression using Mplus software Version 6.1 (Muthén & Muthén, 2011).

Aim 3: To examine the relationship between parenting trajectory and later child overweight/obesity, unadjusted and adjusted multivariate logistic regression analyses will be conducted. Cross-tabulations and unadjusted analysis using chi-square tests will first be used to test the associations between child overweight/obesity and parenting trajectories and covariates. Simple logistic regression will be used to estimate the unadjusted odds ratio and 95% confidence interval of overweight and obesity by parenting trajectory. Covariates identified as significant in bivariate analyses or in the literature will be included in the multivariate regression models.


This study will make a significant contribution to the field early child development due to its innovative conceptualization of parenting, unique methodology, and focus on a significant public health issue in a high-risk population. One of the first studies of parenting quality to use GMM, it will provide a more complex understanding of parent trajectories and changes over time. Data on parenting trajectories could help identify optimal time points for intervention and particularly vulnerable periods for parents. In addition, if practitioners can predict from baseline factors the likely trajectory for parents, resources can be targeted to high risk groups to alter and improve their parenting trajectory. Currently, obesity is one of the leading causes of preventable death in the US and a tremendous public health and economic burden, costing an estimated $147 billion in medical spending in 2008. Since parenting patterns and child growth “track over time,” early childhood may the ideal time to intervene with parents and children on potentially problematic parenting or developmental trajectories. Establishing healthy parent-child relationships early in childhood may promote physiological regulation in children and help prevent or mitigate future growth problems and their negative health sequelae. While most child obesity interventions are aimed at nutrition and health behaviors, few target parent-child interactions or the family context. Results could help inform policy and programs to provide a more comprehensive approach to child obesity prevention and reduction. By focusing on low-income Black and Latino families who are at greater risk for child overweight/obesity, this study may help to identify important risk factors to reduce child health disparities among these populations.

Project Budget

Due to the complex statistical requirements of modeling parenting trajectories using growth mixture modeling, additional statistical training is required beyond what is offered during the school year and covered by my dissertation funding.

Budget Item 1

Statistical training Longitudinal Analysis with Latent Variables (3 credit Summer Institute course) and MPlus Workshop

Estimated Cost: $3,200

Budget Item 2

Other funding ($500 PFRH award + personal expenditures)

Estimated cost: -$2,200

Total Projected Expenses: $1,000


Erin Hunter, MSPH Student
But it Tastes Bad-- Overcoming barriers to uptake of liquid chlorine for water treatment to prevent diarrhoeal disease in urban low income communities in Bangladesh

I. Introduction and Goals

Bangladesh’s unique physiography makes it prone to a variety of natural disasters, including floods, cyclones, earthquakes, and droughts. When this is coupled with  an extremely dense population, poor infrastructure, and widespread poverty, it is not difficult to see how Bangladesh is left vulnerable to a host of public health challenges. The top two contributors to DALYs lost due to communicable, maternal, perinatal, or nutritional causes in Bangladesh are lower respiratory diseases and diarrhoeal diseases—both of which are precipitated by poor water, sanitation, and hygiene. Arnold and Clason have provided evidence that water treatment, specifically the use of diluted chlorine, can reduce the incidence of self-reported diarrhoea. Therefore, the Water, Sanitation, and Hygiene Research Group (WASH-RG) at the International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B) is conducting two large randomized trials of water, sanitation, and hygiene intervention packages, including home water chlorination components, to test optimal acceptability and messaging to promote their uptake in urban communities.cost, the main barrier to widespread water chlorination is consumer acceptance of the practice. Water is not a neutral subject, and interventionists must understand that community members do not look at the subject in the same way as chemists. Many of the obstacles to implementation currently being faced by the WASH-RG in Bangladesh were originally highlighted in the classic article by Nichter, based on his work in Sri Lanka. In this context, water may be experienced as a living thing—something that can be alive, weakened, or dead. Water is not only integral to life and health, but has the potential to offer refreshment and enjoyment. Water chlorination requires water be treated and stored before use. This may not only take away the pleasure of drinking cool water directly from a deep well, but storing the water for a period of time may result in dead, lifeless water which is considered neither satisfying nor beneficial for health.

Perhaps the most substantial recurring barrier to uptake of water chlorination is the concern over the strong smell and taste of chlorine. Adults and children alike are averse to the unfamiliar, medicinal smell and taste, and even the community health promoters (CHP) trained to encourage water chlorination do not practice it in their own homes.

The overarching objective of my research is to contribute to the development and monitoring of initiatives to increase the use of liquid chlorine for water treatment in urban low income communities. We need to understand the community’s concerns regarding water treatment and the best ways to communicate about these issues so we can best address them. Drawing upon my training in anthropology along with my experience in qualitative research and program development, implementation, and evaluation, my role in the WASH projects at ICDDR,B is to take a holistic perspective on the challenges faced in home water treatment, pulling together data collected in the field in order to construct a cohesive picture of the communities’ varied experiences with water chlorination. I will use these data and conclusions to present a cohesive understanding of the challenges faced in the current program implementation and to make recommendations for the development of new intervention strategies to further increase the uptake of home water treatment through chlorination to prevent diarrhoeal diseases.

II. Methods

Study participants are randomized at the household level, resulting in families that receive the intervention being scattered across Bangladesh. Because of this, and the fact that the local research staff can become narrowly focused on individual components, it is often a challenge to pull all the moving parts together into a cohesive story of the current experiences in the field. One of my roles in the WASHRG will be to serve in an advisory capacity, working with the local data collectors to pull the pieces together in order to see the bigger picture. This will include debriefing interviewers, talking about their findings, bringing the study objectives back to the forefront and guiding them in discovering how the data fits into the larger picture of what the study is trying to accomplish. I will also assist the research team in coding interview transcripts and identifying emergent themes (including the use of qualitative data analysis software, Atlas.ti) and helping them to relate these to various behavior change theories and models. It will also be important for me to support and to build the confidence of the data collectors, showing interest in their findings, letting them know when they have struck upon interesting data and encouraging them to follow-up on certain themes during subsequent collections in the field. The following is a list of additional methods I will utilize to achieve the study objectives:

• Qualitative interviews with household members and with community health promoters to determine issues surrounding the use of chlorine concentration and adopting home water chlorination

• Work with the implementing NGO, Dushtha Shasthya Kendra (DSK) to develop and implement solutions

• Qualitative interviews with household members and with community health promoters to assess success and limitation of new initiatives with chlorine concentration to increase home water chlorination

• Examine monthly quantitative data before and after introduction of new initiatives to increase home water chlorination

III. Significance

With roughly 2.2 million children under the age of five dying each year from diarrhoeal diseases, lack of safe water, adequate sanitation, and proper hygiene creates a great burden upon many of the world’s developing nations. Research has resulted in numerous efficacious interventions aimed at improving water, sanitation, and hygiene to prevent disease and mortality, but there is still a lack of formative research or understanding of how these interventions are accepted or rejected in communities across the globe. As previously mentioned, water chlorination is an extremely cost effective and efficacious intervention to prevent diarrhoeal diseases, yet the scale-up of such an intervention has proved almost worthless in urban Bangladesh. More attention needs to be given to the reasons why urban communities in Bangladesh are averse to current chlorination campaigns, and how improvements could be made in order to increase uptake. The results of this study could have far reaching implications, not only for improving the current interventions in Bangladesh, but also through informing the development of WASH programs in other developing regions--by highlighting the importance of firm formative research on the local understandings of water and how these affect the acceptability of particular water treatment methods.

IV. Allocation of Expenses

Housing during research (August-December): $450 x 5 months= $2125.00

Have applied for small awards through the Dept of International Health, still awaiting award announcements. Otherwise, the balance will be covered by additional student loans.

*Note: Because my research will be nested within a larger existing study at ICDDR,B, my role as a qualitative researcher will be added to the existing IRB protocol which has already been approved in Bangladesh.


Ricardo Castillo, PhD Student
Role of Guinea Pigs (Cavia porcellus) in the Epidemiology of Chagas Disease


In Latin America Chagas disease affects 8 million people and about 100 million people at are risk of being infected. Once a rural disease, it is becoming an increasing urban problem. Chagas disease is a complex vector-borne disease that includes the human host, several mammal reservoirs, insect vectors, and the parasite. Several blood-sucking triatomine bugs have been described as vectors in rural areas, but Triatoma infestans is the most important vector in Peru; it is domiciliary and nests in domestic animal corrals, thatched roofs, and irregularities in interior walls of homes. The protozoan parasite, Trypanosoma cruzi, is the causative agent of the disease and only mammals can be infected with T. cruzi. Some mammals have been cited as reservoirs of T. cruzi. Several researchers have reported finding T. infestans in close proximity to guinea pigs corrals, and some have suggested the guinea pig has an important role in the epidemiology of the disease. In the Andean area of South America, such a role would have important implications. The guinea pig has been always ubiquitous in rural areas, and due to the ongoing process of urbanization, the number of guinea pigs in cities is increasing. Laboratory data show that guinea pigs can have an infectivity of T. cruzi of over 95% and survive while infected for several years. These facts could halt ongoing efforts to eliminate the disease, as the interventions are focused on treating infected humans and household insecticide application to control the vector, leaving untouched the animal reservoirs of the parasite.


The project will consist of 2 studies: a cross sectional study in 4 rural communities (n~500 households), and a longitudinal study in an urban ecological
transect (n~750 households). We hypothesize that guinea pigs play two roles, (i) in the persistence of T. cruzi in rural areas, and (ii) in the urbanization of
the disease. The cross sectional study in the rural area will allow us to accomplish Aim 1: to determine if the presence of domestic guinea pigs in houses is
associated with the presence of Triatoma infestans, and with the presence of Trypanosoma cruzi in those bugs. In these rural communities, the MoH vector control campaign conducts residual insecticide application in households. At the time of insecticide application, we will conduct an entomological survey in each house to assess vector infestation (T. infestans). Biologists who accompany the insecticide sprayers in houses, will collect T. infestans as they emerge in response to the insecticide. The specific location of bug capture within the house will be recorded, in addition to the presence of domestic animals in that location. The bugs will be sent to a lab to be tested microscopically for the presence of the parasite (T. cruzi). Information about the sex and age of the bugs will be recorded in order to assess population structures of the bug colonies. The data will be analyzed using mixed models adjusted for clustering by household to assess which factors determine the presence of the vector and the parasite in the vector. The longitudinal study in the urban area will allow us to accomplish Aims 2 and 3 which are to determine the association between vector migration into households and guinea pig population dynamics, and to characterize the vector infestation process in new urbanized areas and its relationship to the presence of domestic guinea pig. For this part of the project, field biologists will visit households within a pre-specified transect in the city every 6 months. They will record information about changes in guinea pig and other domestic animal populations, and they will conduct entomological surveys to determine the presence of T. infestans in households by systematic active searches consisting of 1.5 person-hours per house. As above, parasite presence in captured vectors will be assessed by microscopy. At the household level, these data will be analyzed to evaluate the changes in insect vector feeding patterns as domestic guinea pig populations change over time following seasonal trends. I will apply Bayesian analysis to the data for these two aims. In both the urban and rural studies, all households will be mapped using GPS devices and spatial data will be used to create a geographic data system in ArcGIS to visualize data. We will conduct spatial analysis in R.


Chagas disease affects many countries, from the United States of America to Argentina. The parasite and its insect vectors seem to evolve and adapt well to the different conditions that each country presents. Potential reservoirs vary by country and important findings have been made about the role of wild animals and the persistence of the parasite in the Amazon region, leading to targeted interventions in Amazonian urban areas where elimination was possible. More recently, the role of dogs in Chagas disease transmission in Argentina and Mexico has been described and informed control interventions in these countries. In the Andean region, the role of domestic guinea pigs could explain both the persistence and the urbanization of the disease and support efforts to finally eliminate Chagas disease from Peru and other Andean countries with similar ecologies. This study will collect longitudinal data and will have fine spatial resolution that will allow us to detect patterns of vector migration within households. Moreover, I will work closely with the MoH personnel of the Chagas Disease Control Program, such that I can become familiar with their needs and provide them with real time data to adjust
and improve their control efforts.


- 1 round trip ticket Baltimore - Lima, Peru - Baltimore: $650
- 1 round trip ticket Lima - Arequipa (field site) - Lima: $200
- Local transportation and food (2 months): $150

I will be at the field site supervising and conducting data quality control as it is collected. Also, I will be coordinating our study activities with the work of the MoH.

*Housing will be provided by the Peruvian University Cayetano Heredia.
*Other costs related to this study such as data collection, lab testing, etc. are already covered by the project budget.

Sherri Rose, Post-doctoral Fellow
When to initiate combined antiretroviral therapy in HIV-infected persons in the United States


The question of when to initiate combined antiretroviral therapy in HIV-infected individuals is complicated. There is a delicate balance between managing CD4 count, while not starting treatment too early, such that the patient will quickly develop resistance to the drugs. Statistical methodology for handling this intricate longitudinal process to answer the “when to start” question has been limited. We will develop a double robust estimator that accounts for the time-dependent confounding of the changing CD4 count and viral load over time.


When to initiate combined antiretroviral therapy in therapy-naive individuals is a complicated clinical question. As CD4 count drops, individuals become more susceptible to infection, however, early initiation of therapy can lead to drug resistance. When treatment begins, the viral load of drug-susceptible quasispecies will fall, but drug-resistant quasispecies will not be eliminated, leading to incomplete suppression. These drug-resistant quasispecies will then reproduce, and viral load will rise. The same antiretrovial drugs will not be effective at lowering viral load again, and the treatment regime must be changed to avoid disease progression. Most worldwide guidelines for the initiation of combined antiretroviral therapy in HIV-infected individuals recommend falling below a CD4 threshold of 0:350 109 cells/L before beginning treatment. A recent study indicated that starting treatment after falling below 0:500 109 increased AIDS-free survival. Previously, an analysis of early versus deferred therapy demonstrated a 94% increased risk of death associated with deferring treatment for those with CD4 count of 0:500109 and below, and a 69% increased risk of death associated with deferring treatment for those with CD4 count of 0:350109 and below. The recommended threshold in the U.S. is currently 0:500109, although considerable debate remains. Randomized trials have not been effective in providing useful data since enrolled subjects are not therapy-naive, among other issues. Thus, the use of observational longitudinal studies offers the best opportunity to answer this complicated “when-to-start” question. We will develop a novel double robust estimator that accounts for the time-dependent confounding of the changing CD4 count and viral load over time. The clinical impact of correctly identifying the optimal CD4 cutoff is substantial, as over 1 million people are currently infected with HIV in the U.S., and over 30 million in the world.


The North American AIDS Cohort Collaboration on Research and Design (NA-ACCORD) collects clinical HIV and AIDS data across the United States and Canada in multiple study sites as part of the International epidemiological Databases to Evaluate AIDS (IeDEA). The data provided is a rich source of clinical information on treatment and mortality outcomes from 1996 to 2010. Data for each subject include, at each time point t (visit): current CD4 count, current viral load, summary of past CD4 count, summary of past viral load, indicator of death at or before time t, antiretroviral therapy status, and an indicator of censoring due to individual loss to follow-up, among others. Additional baseline variables include age, sex, cohort, hepatitis C status, and intravenous drug user.


We will present and develop a new semiparametric efficient double robust substitution estimator for longitudinal observational data to identify the optimal CD4 count to initiate treatment in therapynaive individuals using the NA-ACCORD cohort. This innovative targeted maximum likelihood estimator (TMLE) has desirable properties compared to two existing methods for controlling the time-dependent confounding found in observational data, inverse probability weighting (IPW) and g-computation. IPW methods are unstable and can provide large confidence intervals since they are not substitution estimators. G-computation estimators are not double robust and also tend to suffer bias since they do not account for the process by which the treatment is generated over time. The TMLE does not suffer from these issues and thus is poised to provide the most accurate estimates of survival for different levels of CD4 initiation.


Given the burden of HIV prevalence, decreasing mortality among HIV-infected persons will have a significant public health impact. The new TMLE methodology can contribute substantially in identifying the optimal CD4 level to initiate treatment, by analyzing the NA-ACCORD data, as well as other longitudinal observational cohorts in the future. Methodology for observational longitudinal data has be slow to develop due to the complicated issues surrounding time-dependent confounding, and the TMLE will also contribute overall to the methodological literature.


The fellow is requesting $1000 in funding for this project from the Delta Omega Society Scholarship. Since the proposal is a statistical methodology project, the costs associated with its completion are low. The fellow is soliciting funding to attend the main statistics conference, the Joint Statistical Meetings. She has already had an invited talk accepted for presentation.