Skip Navigation

2009 Scholarship Awards


  • Brandon Brown

Acceptability of HPV vaccine in Female Sex Workers Faculty advisor: Neal Halsey

Introduction/Importance: Persistent human papillomavirus (HPV) infection is found in nearly all 500,000 cases of cervical cancer each year worldwide 1. HPV prevalence in Peru is approximately 17.7%, much higher than the worldwide prevalence of 10.4%, while cervical cancer is the leading cause of cancer death in Peruvian women (17%) 2,3. A low proportion of Peruvian women are screened for cervical infections and cancer precursors, because a Pap test is more than one day’s pay for a family already in poverty 4.Recent studies have shown HPV vaccines to be effective in decreasing the burden of cervical cancer precursor incidence among women in the general population. HPV vaccine is currently approved for use in Peru, but with the current cost of $375 for 3 doses, it is not possible for the majority of women to receive this life saving medicine. In addition, studies have not included women of higher age and experience, mainly due to opinions of prior exposure. The FDA is considering increasing the recommended oldest age of HPV vaccination from 26 to 49 years, because data show that women who are older and HPV exposed, even with one of the four vaccine subtypes may also benefit from this vaccine. HPV types 16, 18 are present in 20% ofHPV infections, and it is unlikely that women would be exposed to both of these cancer causing subtypes protected with vaccine 5,6. In fact, infection with both subtypes in a sexually active female college student population in Lima, Peru was only 0.3%, showing promise for our study (new results). While it may not be financially possible to vaccinate all women in Lima, a focus on high risk populations may be achievable. Female sex workers (FSWs) are at higher risk of HPV infection and cervical cancer due to multiple partners in their occupation and the resulting exposure to multiple HPV types 7,8. Compared to women in the general population, HPV prevalence in FSWs was found to be approximately 14.4%-75% in several studies 8,9. Registered FSWs in Peru undergo sexually transmitted infection (STI) testing at clinics every 3 months, making completion of a modified HPV vaccine schedule at this time more feasible 10.

Methods: The primary objectives of the research are to determine the acceptance and potential for the effective use of HPV vaccine in FSWs of different ages. We will also be obtaining data on HPV prevalence in this population, which is lacking in the academic literature. As a result, this is a public health practice project. To accomplish these objectives, we will employ two different study designs:1. A cross-sectional study of HPV prevalence and cervical infection in female sex workers2. A prospective open label randomized trial of the acceptance of HPV vaccine at different schedules.This study will take place in Lima, Peru at the non-governmental organization Via Libre over the course of11 months. We estimate 250 women will be screened and 200 women will be included in the protocol. At baseline, HPV prevalence and cytology will be measured among participating females attending Via Libre with a physician administered Pap smear and collection of serum antibody. Correlation of HPV infection with history of sexual partners will be examined. Participants will be randomized to receive HPV vaccine according to the standard schedule (0, 2, 6 months) or the modified schedule (0, 3, 6 months) which matches their required STI testing visits at the clinic. The vaccine will be provided at no charge to study participants. Antibody to HPV types 6, 11, 16 and 18 will be measured among FSWs at baseline and one month after the third dose (month seven). A behavioral survey will be administered at baseline and follow-up to collect information on awareness of HPV, and other relevant information.Work in this project will be overseen by Professor Neal Halsey, with whom I have received partial funding through Merck pharmaceuticals. The IRB application is complete and we are preparing to begin the project in late April. We are currently working on the application to the Peruvian NIH. In preparation for the trial, we recently completed a baseline survey of 319 women on acceptability of HPV vaccine and barriers to acceptance. We will soon submit results of this study to Sexually Transmitted Infections.

Significance: This trial includes several novel innovations. This will be the first trial which studies the immune response of a licensed HPV vaccine in female sex workers. It will also explore the acceptability of the three dose regimen in this population, while examining the benefit of employing a schedule more suitable for FSWs in Peru. Participants in our study will be offered a Pap smear at baseline. For many women this will be their first and only exam in their lifetime. Systematic screening for abnormal cervical cells has been shown to be cost effective compared to treating cancer cases, and efficient in reducing death rates from cervical cancer by 70% or more 11. Women with any abnormal cytology will be referred to an OB/GYN on site, and peer health educators will be available for necessary counseling. Women will have blood analyzed for high cervical cancer risk HPV infection on two occasions.As the subtypes protected by HPV vaccine are found among approximately 60% of women with cervical cancer cases in Latin America, and FSWs are a high risk population, the burden of cervical cancer may be substantially decreased through early and widespread vaccination efforts this study population. Vaccination may not only help prevent cervical cancer in the FSWs, but it would help prevent infection by subtypes 16 and 18 to other sex partners of men who frequent brothels in Lima for sex work. Benefits may include protecting wives, girlfriends and significant others from chronic HPV infection, genital warts, and cervical cancer. FSWs as a group may benefit from the modified (0, 3, 6 months) schedule due to their mandatory presentation to STI clinics every 3 months. Our study will help determine the practicality of HPV vaccination in FSWs as well as evaluate the effectiveness of administering HPV vaccine at the more convenient schedule.In summary, this study will help increase HPV awareness among FSWs, provide already globally recommended cervical screening, as well as provide potential evidence for worldwide early vaccination of female sex workers to prevent cervical cancer in this high risk population. Mathematical modeling has shown that high vaccination coverage over many decades can decrease type specific cervical cancer incidence by as much as 91% 12. We hope to make a contribution to reaching this goal.


  • Kristen Gibson

Tangential Flow Ultrafiltration and Molecular Detection of Surrogates and Pathogens in Large-Volume Environmental Water Samples
Over the past two decades, there as been an increased emphasis on the quality of raw water supplies and finished drinking water. Quality issues, particularly those associated with microorganisms,are of increasing importance as a result of environmental impacts on current water supplies and development of alternative water sources. Accurate and comprehensive assessment of microbial water quality is of paramount importance if both existing and new water sources are to be safely employed. This research project aims to address the inadequacies of the current methods and standards used for determining microbial water quality within drinking water systems and source waters. The objectives of this research are 1) to optimize and implement tangential flow ultrafiltration (TFF) technology for recovery of low levels of bacteria, viruses, and protozoa from large-volume water samples; 2) to develop real time, quantitative molecular techniques, such as polymerase chain reaction (PCR) and loop-mediated isothermal amplification (LAMP), for the rapid detection of select enteric pathogens; and 3) to apply these developed methods to large-volume environmental water samples for the detection and quantification of microbial loads.The combined public health and economic burden of acute gastrointestinal illnesses (AGI) attributable to drinking water is considerable. On average, between 1991 and 2002, 17 waterborne disease and outbreaks (WBDO) associated with drinking water were reported annually in the United States (1). The etiologic agents associated with these outbreaks include Cryptosporidium, Giardia, norovirus, E. coli O157:H7, Shigella, and Campylobacter and caused primarily symptoms of AGI (2). These reported outbreaks, however, may only represent a small portion of AGI cases associated with waterborne transmission that occur each year due primarily to the passive nature of AGI reporting. Some of the most recent estimates of AGI attributable to drinking water in the United States range from 12M cases/year to 19.5M cases/year (3). The economic burden of AGI related to waterborne transmission is even more difficult to establish though it could potentially be equivalent to the burden of foodborne related AGI with an estimated cost of $19.7-34.9B each year (4). In an effort to alleviate these burdens and estimate the true risk of disease transmission via water, research initiatives continually aim to address whether the current standards and methods for water quality monitoring are adequate for the protection of public health. At present, public water systems rely on bacterial indicators (i.e. coliforms) for monitoring water quality, and it has been shown that bacterial indicators are often poorly correlated with the presence of other microorganisms, such as protozoa and viruses, which can be found in various water sources including finished drinking water (5). It is also important to note that in many of the reported outbreaks, the water systems were in compliance with the current water quality standards (2). Therefore, the lack of an adequate method to monitor water quality has lead to the need for developing an effective and efficient method for the simultaneous collection and recovery of low levels of bacteria, viruses, and protozoa that can then be rapidly identified and quantified.In order to detect low levels of microorganisms, one must be able to concentrate large volumes of water, however, current concentration methods are targeted towards specific pathogen groups. This is due primarily to the considerable range in size, composition, and persistence of pathogens, which makes the isolation of more than one microbial category especially challenging. For example, enteric viruses are small (25-100 nm), non-enveloped protein particles that may exhibit polarity and are capable of adsorbing to a wide variety of charged matrices. In addition, viruses are often present at low levels in water requiring volumes in excess of 100 L to be collected in order to have confidence in an assay. The methods for the recovery and detection of protozoa, such as US EPA method 1623, also require relatively large volume samples to be collected and processed. In comparison to the viral and protozoan techniques, the methods for detecting indicator bacteria are relatively simple in which a 100-mL water sample is filtered through a 0.45 m filter and plated on selective media. Monitoring for indicator organisms in this way has been the traditional test for microbiological quality of drinking and recreational water sources, but provide limited information on the presence of protozoa and viruses. In addition, pathogenic bacteria such as Escherichia coli O157:H7, Helicobacter pylori, Salmonella, and Campylobacter, potentially present in source waters, are in many instances present in concentrations too low for detection in 100 mL, even if appropriate growth media and incubation conditions are used. Also, as a class of microorganisms, most vegetative bacteria are much less resistant to environmental degradation and chemical inactivation (including chlorination) than are enteric viruses and protozoa (6-9). Therefore, routine bacterial monitoring is often inadequate for determining the presence and level of viruses and protozoa in source and drinking water. It is evident that a methodology is needed for improving the sensitivity of recognizing risks attributable to viruses and protozoa in source and drinking water.A TFU method has been optimized to concurrently collect and recover each class of microorganism from 100L water samples. The optimized uses a commercially available capsule filter containing hollow fiber membranes of 70,000 dalton pore size to concentrate 100L water samples. Filter type, surfactant addition, and elution steps were evaluated during TFU optimization. For evaluation, microbial surrogates including E. coli CN-13, E. faecalis, C. perfringens spores, MS2 and PRD1 bacteriophages, poliovirus and murine norovirus (MNV-1) were added to either dechlorinated tap water (DTW) or surface water (SW). The final concentrated sample (100-300mL) is then analyzed using culture-based methods and/or molecular-based real time polymerase chain reaction (qPCR) and loop-mediated isothermal amplification (LAMP) for the identification of select, endogenous microorganisms including pathogenic bacteria, viruses, and protozoa of human health concern. Employing the final optimized TFU method, and using standard culture techniques, the average recovery efficiency for bacteria, viruses and spores in DTW (n=25) was 75, 50, 59% and for bacteria and viruses in SW (n=4) was 62 and 49%, respectively. A qRT-PCR method was developed and applied to the detection of MNV-1 in TFU concentrates (n=20). Target MNV-1 RNA was detected in all but one SW sample. In addition, the qRT-PCR assay incorporated an internal standard using Hepatitis G RNA to identify potential sample inhibition. TFU was then applied to 100L surface water (SW) and ground water (GW) samples in Lower Yakima Valley, WA—a region with the densest concentration of dairy operations and milk cows in WA and with known nitrate contamination in GW sources. One hundred liter GW (n=10) and SW (n=11) TFU sample concentrates were evaluated for indicator bacteria including total coliforms, E. coli, and Enterococcus spp. using the IDEXX(r) detection system. Viral and protozoa analysis are ongoing.

The research I have proposed will contribute to the improvement of public health in various ways. First, through the development of a universal method for the recovery of microorganisms, water utilities and regulatory agencies will be better equipped to address problems within public water systems. By moving away from reliance on bacterial indicators for water quality monitoring, we will increase our capabilities for detecting viral and protozoan pathogens, which are far more persistent and resistant to degradation than bacteria. Second, through the utilization of rapid, quantitative molecular detection methods and integrated culture systems, we will be able to provide increased confidence in the sensitivity and specificity critical for estimating levels of risk. Along with this, we will have a more comprehensive understanding of the microbial contamination of various water sources allowing for exposure risk assessments to be generated for individual microorganisms. Overall, this research will assist in the formulation of effective control measures for the reduction of water-related transmission of pathogenic microorganisms.

  • Raydel Valdes Salgado

Changes in intensity of daily tobacco consumption in the US. An application of multilevel and longitudinal analysis to the Tobacco Use Supplement of the Current Population Survey 1992-2007.

INTRODUCTION Tobacco consumption is the single largest preventable cause of death worldwide and a steady reduction in the amount smoked is a trait of the path to successfully quitting1. The main goal of this proposal is to estimate the temporal changes of the number of cigarettes smoked per day (CPD) in the US. I am hypothesizing that race-specific estimates have been hidden behind overall estimates which are the standard for smoking research. In this analysis I am interested in investigating the effects of state-specific tobacco control environments on the CPD and in characterizing whether these effects vary by racial/ethnic groups. The relevance of the topic and the complexity of the analysis motivated me to apply for this award.

METHODS We propose an analysis of the Tobacco Use Supplement of the Current Population Survey (1992-2007) to describe racial/ethnic specific changes in CPD over time and to determine what individual-level and state-level factors are associated with reduction in the number of CPD. The characteristics of this survey will allow us to do a repeated cross-sectional analysis, a multilevel analysis, and a longitudinal analysis to address the issue of racial/ethnicspecific change in average CPD. The specific aims are described below. Specific aim 1: To estimate racial/ethnic-specific changes over time in the average number of CPD among smokers in the US between 1992 and 2007. Hypothesis 1: In general, there is a downward trend in the amount of CPD smoked in the US; however the rate of such decline is unequal among racial/ethnic groups and is also different by gender. The association of individual-level characteristics such as age, income, and education with the temporal change in the average number of CPD is specific for each gender and racial/ethnic group. Specific aim 2: To examine variability across states in the race-specific temporal change of CPD. Hypothesis 2: Race-specific patterns of amounts smoked (CPD) exist, but vary across states. Gender specific patterns within racial/ethnic groups are different across states. Specific aim 3: By using a longitudinal data set for the period 2002 to 2003, to determine whether increasing cigarette tax affect CPD differently for different racial/ethnic populations after adjusting for other state-level tobacco control measures and individual-level characteristics. Hypothesis 3: Increasing cigarette tax impacts differently among racial/ethnic groups. Minorities are more responsive to an increase in the price of cigarettes. The impact of such a measure varies across states. Project’ feasibility: The survey includes a probability sample which is allocated among the states to produce state and national estimates. Here, individual-level variables are soiciodemographic characteristics and person’s smoking behavior, while cigarette price, availability of state-wide secondhand smoke policy at worksites and public places, as well as cessation resources are state-level variables. I gathered information from several sources2-5 to set the databases needed for this project. Because of the huge sample size (more than 1.3 million of observations) and the complexity of the analysis, to have a powerful computer is a core necessity for this project. Currently, I have not such a machine.

Scientific and scholarly merit: The outcome variable CPD conveys many challenges to the analysis, hence the necessity of effective statistical reasoning and methods to model it. The number of CPD are nonnegative, integer-valued responses taking on values (0,1, 2 …) which tend to have a mode at zero and a distribution with a long, heavy right tail as shown in the figure. There are two causes of ‘real overdispersion’ when dealing with count data6: one is a violation of the distributional assumptions upon which the model is based; a second cause relates to misspecified variance. With regard to the first cause, we know that the distributional mean of Poisson and NB specifies an expected number of 0 counts in the response, but the variable CPD has an excess of zeros (happily, most people are non-smokers!7). As for the second cause, we have to be aware of potential extra correlations observed in our models because observations in the surveys are clustered in states. Therefore, we have to use zero-inflated models with state-specific random intercepts for clustered count data to avoid incorrect parameter estimates as well as biased standard errors8. In my 30-page proposal I meticulously described the two-part models related to each specific aim and how to interpret their parameters, but here I will only present the random-effects zero-inflated model, with normally-distributed random effects j ~ N(0, 1) and vj ~ N(0, 2) under its general form: Pr(yij) = p (yij) = ij I (yij) + (1 ij) f (yij), with Part One: logit( ij) = ’wi j + j and Part Two: log( ij) = ’xi j + vj = ’wi j + 1 1j = ’xi j + 2 2j In this approach the probability distribution of yij is modeled as a function of the covariates wij and xij, and the random effects j and vj are specific to the jth cluster. In other words, the first part involves estimating the probability of a zero outcome (we can say ‘smoking participation’ because it is a binary process, where CPD=0 or CPD>0) and the second part involves estimating the probability of a non-zero count (CPD> 0). The interpretation varies by specific aim, but in general we can say that measures the change in the conditional logit of reporting 0 CPD ( ij) for individuals in each cluster j, adjusting by wij (the individual-level and state-level covariates). Likewise, measures the change in the conditional log of the number of CPD ( ij ) for individuals in each cluster j, adjusting by xij (the individual-level and statelevel covariates). The random intercepts represent the combined effect of all omitted state-specific covariates that cause individuals from some state to be more prone to smoke than others ( j in the Bernoulli part) and to smoke more CPD than others (vj in the count response model). The parameters 1 and 2 represent the cluster variance terms for the logistic and Poisson components, respectively. Similarly, 1j and 2j represent the within cluster variance8. There are additional challenges; first, we have to test that the error terms in Part One are not correlated with error terms of the second part of the models. Additionally, in the longitudinal analysis, we have to consider a serial dependence correlation structure9. Because of difference in cluster sample size –particularly for minority groups in some states– the precision of cluster specific estimates may vary greatly. Hence, we will apply Bayesian 2-stage hierarchical models to estimate more efficiently the state-specific and national averages.

Why would be relevant for the School to support this project To the best of our knowledge, there is no precedent of using a national sample for these complex analyses. We are using public data bases, so our study we will be a rich source of examples that can be used in class here at JHSPH. For teaching purposes, this project might illustrate well the challenge due to the hierarchical data collection procedure as zero-inflation and lack of independence may be present simultaneously as a consequence of the inherent correlation structure and underlying heterogeneity. Therefore, our models also account for the unequal racial composition across states and the lack of homogeneity in the strengths of tobacco control programs at state level. These last two features are relevant to any epidemiologic interpretation of our national and state estimates. Very seldom do we have repeated measurements of smoking variables, so a longitudinal analysis is also possible. This kind of statistical analysis almost represents the closest we could be in Epidemiology to a counterfactual scenario to estimate racial/ethnic specific impact of the strongest control measure, increasing cigarette tax. Significance for Public Health: This research is in a context where tobacco-related health disparities among different racial/ethnic groups have been documented10 and the question of whether the benefits of tobacco control interventions are reaching all groups has arisen. Thus, we can potentially contribute to a better understanding of racial/ethnic differences in smoking behavior and the issue of whether tobacco control interventions have a differential impact among groups. Additionally, tobacco control efforts vary widely across the US11, so here we have the opportunity to model gender and race specific patterns of use and decrement in CPD in recent past in each of the US states, which is a unique opportunity to understand how tobacco control measures are more or less effective in different scenarios. Thus, results from this study may not only inform policies that may be race/ethnic-specific, but also contribute to define priorities and allocation of resources for prevention in a more effective way. With this study we could generate sound evidence of the linkages between regulatory interventions at different levels and changes in the intensity of smoking across different subgroups of the population.


  • Kyla Hayford

Using Oral Fluid Samples to Validate Immunization Coverage of Infants:
A Non-Invasive Approach to Identifying Measles Susceptible Children Missed
by Conventional Surveillance

Introduction The provision of basic childhood vaccinations is a cornerstone of an effective health system. Despite impressive gains in vaccination coverage since the 1980s, many low-income countries have failed to sustain the 80% coverage goal set out by the Universal Childhood Immunization initiative(1). Recent outbreaks of measles, pertussis and polio in low-income countries indicate that conventional surveillance methods are not sufficiently identifying susceptible children. Approximately 1.4 million children died of vaccine-preventable deaths in 2002, which could have been largely prevented via improved surveillance and targeting(2).Measuring immunization coverage in a low-income country is difficult and costly. In the absence of a feasible gold standard of population immunity, several proxy indicators of immunization coverage are used but correlation between indicators is often poor due to differences in the methods used to collect the data and calculate coverage(3). The conventional approaches for estimating immunization coverage are based on administrative records from health centers, vaccination card records or parental recall – all of which suffer from social desirability bias, recall error and other measurement errors (4-7). Studies from Bolivia and Kenya found that parental report data underestimated the number of susceptible individuals, especially in populations with relatively high coverage (8, 9). A second concern is that reported vaccination does not account for incorrect administration of vaccines or breaches in the cold chain, leading to an overestimation of actual population immunity. Finally, because immunization coverage is used widely as an indicator of health systems functioning, progress toward the MDGs and improvements in coverage are often tied to donor funding, there is a significant risk of over-reporting by governments. A recent study by Lim and others found that governments receiving performance bonuses for improving immunization access were overestimating true coverage by 33-100% (3). Given the potential for bias and error, new tools are needed that go beyond reported vaccination data to establish alternative estimates of coverage and to explore reasons for discrepancies across sources (4, 5, 7, 10-12).Oral fluid biomarkers of immunity can be used to validate measles immunization coverage estimates. A common approach to validate immunization coverage estimates is to estimate the prevalence of disease-specific antibody from blood samples(13). Its use has ben limited in low-income countries due to the costs, invasiveness and feasibility of collecting blood, but oral fluid collection has emerged as a realistic, minimally invasive and low-cost alternative (14-20). Despite earlier concerns about the accuracy of oral fluid assays, the new disease-specific IgG enzyme-linked immunosorbent assays (ELISAs) for measles have demonstrated high sensitivities and specificities vis-à-vis blood-based assays (9, 18, 19, 21-24). Because oral fluid assays for measles are superior to other oral fluid assays (e.g. vaccine-preventable disease alternatives, this study proposes to explore the validity of immunization coverage estimates by comparing conventional methods for estimating measles coverage to non-invasive oral fluid sampling methods.

Goal and Specific Aims The goal of the proposed study is to use non-invasive oral fluid biomarkers of measles immunity to assess the validity of conventional measles immunization coverage estimates based on administrative records and parental recall. The specific aims are to:- Estimate prevalence of measles immunity using oral fluid assays among 12-23 month olds and compare it to immunization coverage estimates from administrative data and parental recall in the same population.- Determine the sensitivities and specificities of parental report and ‘card plus report’ of measles immunization compared to oral fluid-confirmed immunity.- Develop a prediction model of measles immunity in the population based on individual-level and community-level predictors for misreport of a child’s immunization status.- Test if there is children receiving vaccinations from a measles campaign are more likely to confer immunity than from routine immunization.

Methods Sampling and Eligibility. A population-based representative survey with oral fluid collection will be conducted with 660 randomly selected eligible parent-child pairs in one district of Dhaka, which has one of the worst immunization coverage rates in the country. Eligible children include all 12-23 month olds living in the randomly selected household with a parent/guardian present at time of interview and who consent to the study protocol and oral fluid sampling. Based on the 30-cluster WHO Immunization Coverage Survey design, each cluster will need 22 children to estimate prevalence within ±5%, based on a two-tailed alpha=5% and power = 80%(25). To account for intra-cluster correlation, sample size calculations include a design effect (deff) of 1.3, which is slightly higher than deffs used by the 2004 Bangladesh DHS for measles vaccination questions. Reliability re-tests on a 10% subsample will be conducted.

Data Collection. A survey on vaccination history and an oral fluid sample will be collected from every enrolled child. Prior to requesting vaccination cards, data collectors will ask parents about the child’s vaccination history and corresponding dates of vaccinations for all routine infant vaccines (BCG, DTP3, OPV3, MCV) using standard questions from DHS and WHO immunization surveys. After parental recall is completed, vaccination cards will be requested and dates of all vaccines will be recorded. Health status, history of measles, and additional demographic and socioeconomic characteristics of the household will also be collected. Oral fluid samples will be obtained with Oracol oral-specimen collection devices, stored in a cold box until processing in a lab in Dhaka according to the manufacturer’s protocol. Oral fluid specimens will be tested for measles-specific antibodies with an IgG capture EIA, which has been validated for oral fluid. Government data will be provided by the Bangladesh EPI team.

Analysis Plan. For the primary analysis, a log binomial regression will be used to estimate ‘coverage’ or prevalence of immunity and predictors of immunity. A score test for binomial proportions will be used to test if the prevalence of immunity based on oral fluid samples is significantly different from coverage from government estimates (one-sample test) or parental recall (two-sample test). Correlation, kappa statistics, sensitivities and specificities will be calculated for pair-wise comparisons of oral fluid vs. card+report , and oral fluid vs. parental report. The second analysis will evaluate the accuracy of parental report compared to oral fluid-confirmed immunity. Using multilevel logistic regression models, a prediction model will be developed to estimate measles immunity using individual and cluster-level factors associated with oral-fluid confirmed immunity. By identifying which factors are associated discordance between reported status and actual immune status, we can correct coverage estimates from conventional surveillance methods to better reflect the true proportion of measles-immune individuals. Error models, which have been used to adjust for error due to an imperfect gold standard, will be explored to ensure oral-fluid based estimates of measles immunity are accurately modeled (9, 15, 24).

Significance The proposed study uses an innovative, non-invasive technique of oral-fluid collection to assess population immunity to measles. Only three studies have used oral fluid assays to assess population immunity in low income settings, but none have directly attempted to validate conventional coverage estimates against this tool (8, 9, 19).This study moves beyond estimation by developing a model to estimate true immunity based on comparisons between oral fluid-confirmed immunity and reported immunization. If scaled up to the national level, governments, UNICEF or donor agencies would be able to ‘correct’ reported immunization coverage estimates by collecting a small subsample of oral fluid to identify characteristics of households that are likely to over- or under-report their true immune status.With such broad use of immunization coverage indicators now, it is crucial to develop tools to validate the accuracy of conventional methods for estimating coverage. Measles immunization lags behind the other basic child vaccines and therefore is used as one of three indicators to measure progress toward the 4th Millennium Development Goal (MDG4) of reducing child mortality (26, 27). As we approach the 2015 deadline for the MDGs, there may be incentives for governments to over-report measles immunization coverage. Thus, exploring the accuracy of measles immunization coverage is not only relevant to measles policy but will also generate evidence on how to interpret and validate different sources of coverage data, which is applicable for surveillance of all childhood immunizations.

  • Shegufta Shefa Sikder

Maternal Mortality in Rural Northwest Bangladesh - Dysfunction, delays and solutions: A critical evaluation of underlying pathways leading to death
A Qualitative Study in Collaboration with the JiVitA Project in Gaibandha, Bangladesh

Introduction My thesis project involves the qualitative analysis of deaths in women of reproductive age to examine the social and cultural barriers to maternal health care that women face in rural Bangladesh. Postpartum hemorrhage, eclampsia, unsafe abortions, obstructed labor, and sepsis account for the major direct causes of maternal death in Bangladesh, the majority of which are preventable with timely medical intervention. Yet even in areas of Bangladesh where maternal health care services are available, care seeking remains low. While much research has focused on the biomedical causes and biologic pathways of maternal death, numerous studies call for locally relevant research into the social context in which these deaths occur to enhance understanding of the barriers to health care for women. , My thesis project focuses on elucidating pathways to mortality among women of reproductive age in rural Bangladesh.Overall Goal: To analyze the social and cultural factors involved in pathways to mortality among women of reproductive age in Gaibandha (northwest rural Bangladesh).

Specific Aims: 1. To test whether the three delays model adequately addresses the barriers to health care experienced by women of reproductive age in northwest rural Bangladesh.2. To localize the three delays framework to account for the social and contextual factors surrounding maternal health care for women in Gaibandha.Secondary Aim: To assess whether current health programs in Gaibandha address the barriers to maternal health care experienced by women of reproductive age in Gaibandha.

Methods The proposed study is in collaboration with the JHU-JiVitA project, a double-masked, cluster-randomized, placebo-controlled trial in rural Bangladesh. Since 2001, this project has tested the efficacy of weekly maternal supplementation with vitamin A in reducing all-cause pregnancy-related mortality, maternal morbidity, infant mortality, and other adverse pregnancy and infant health outcomes in Gaibandha, northwest rural Bangladesh. This proposed study has the full support of the JiVitA project; my advisor as well as the other study team members have been assisting me in planning this study.Over the past eight years, research physicians of the JiVitA project have conducted 1192 in-depth verbal autopsies to ascertain the circumstances leading to death among women of reproductive age (aged 14 to 45 years). These verbal autopsies include open-ended narrative histories in which respondents describe the circumstances surrounding the death of a woman who died while under surveillance. Using these narrative histories, supplemented by interviews that I will conduct in the field, my thesis project will analyze barriers to care experienced by women in this study area. Socioeconomic and health indicators of this cohort suggest that it is representative of rural women across Bangladesh. By documenting the experiences of women who died while under surveillance, these narrative histories provide a powerful source of primary data for understanding the barriers faced by these women. Through analysis of these narrative histories, I aim to identify barriers contributing to their deaths, beyond the biomedical causes of mortality. I will test whether the most prominent framework describing factors contributing to pregnancy-related mortality, known as the three delays model, accurately addresses the experiences of women in the study site of Gaibandha.For the first stage of my research, completed in February 2009, I conducted an extensive literature review of documented barriers to health care experienced by women in Bangladesh. For the next phase, I have been performing textual analysis of approximately 600 narrative histories collected in the JiVitA study in order to build a conceptual model of the barriers to care experienced by women in this population. Using Atlas.ti, a software package that allows for in-depth textual analysis of documents, I have been linking keywords and concepts in the text of these narratives to build a data-based conceptual model of barriers to maternal health care. This phase of research will be completed by the end of April 2009. Working with my advisor, I obtained IRB approval for this analysis, and we are applying for IRB approval for the field component of this project.From June 2009 until December 2009, I plan to implement the field component of my project at the JiVitA study site in Bangladesh. Through complete immersion in this setting, I will perform ethnographic fieldwork to document the social and contextual factors surrounding the barriers to health care for women in Gaibandha. I will conduct 25 key informant interviews with a representative sample of women of reproductive age who are currently pregnant or have been pregnant during follow-up in the JiVitA study. These interviews will provide information on the barriers to obstetric and maternal care experienced by women. I will compare the barriers identified from both these interviews and the narrative history analysis to suggest how the three delays model can be localized to rural Bangladesh. For example, many people in rural Bangladesh have a strong belief that evil spirits contribute to pregnancy-related mortality. Using the key informant interviews and narrative history analysis, I will elucidate factors such as the belief in evil spirits to accurately account for the cultural barriers to maternal health care experienced by women in this setting.To achieve a secondary research aim of identifying whether current health programs in Gaibandha address the identified barriers to maternal health care, I will document the services of existing maternal health programs through interviews with five field workers and five project leaders of BRAC (Bangladesh Rural Advancement Committee), the largest NGO provider of maternal health care services in Gaibandha. I have experience in performing key informant interviews in Bangladesh from previous fieldwork; I am fluent in Bengali and able to ask questions without a translator.

Significance This research project will help to advance understanding of the social and cultural factors contributing to preventable pregnancy-related deaths. Much literature on barriers to obstetric and maternal care emphasizes the need for locally relevant research and a deeper understanding of the local context of health care in underserved locations.iii,iv Through complete immersion in Gaibandha for six months, I will add valuable knowledge of the cultural context of the barriers to care experienced by women in rural Bangladesh. This measurement project will utilize qualitative anthropological field methods to help illuminate social and cultural factors contributing to preventable deaths among women of reproductive age in rural Bangladesh.As a Bangladeshi woman, I have always been committed to helping to combat the numerous health obstacles experienced by Bangladeshi women of reproductive age. My previous research projects have dealt with assessing whether NGO health interventions had contributed to improvements in indicators relating to the maternal health Millennium Development Goal (MDG 5). Performing focus group discussions and key informant interviews for these projects allowed me to speak with Bangladeshi women one-on-one about the health problems that they experience. These powerful experiences fueled my passion to continue working to alleviate the obstacles contributing to preventable deaths among this population. I am excited about the opportunity to implement this research project and contribute new information to the field of maternal health in Bangladesh. My training from Hopkins, my prior experience, and most importantly the insightful guidance I continue to receive from my advisor and JiVitA study team members make me confident that I will be able to perform this work with the scientific rigor expected of JHU public health research.

2006 Scholarship Awards

The five Awards below were chosen from 40 excellent submissions from 10 departments and 4 degree programs

Xiadong Cai, PhD candidate in International Health - Impact of Health Insurance on Health Care Utilization in Vietnam

Judith Easterbook, PhD candidate in Molec. Microbio. & Imm. - Regulatory T Cells Mediating Seoul Virus Persistence in Norway Rats

Gabrielle Hunter, MHS candidate in International Health - Effects of Water Availability and Education on the Hygiene Behavior of Residents in a Peri-Urban Community in Lima, Peru

Sang Kim, PhD candidate in Epidemiology - "Center Effect" in Kidney Transplantation: A New Look at an Old Problem

Rachel Singer, PhD candidate in Health Policy and Management - Mobilizing the Masses: Potential for Universal Coverage in Maryland

Project Summaries of Scholarship Awardees

Xiadong Cai, PhD candidate in International Health - Impact of Health Insurance on Health Care Utilization in Vietnam


Impact of health insurance on health care utilization in Vietnam

Introduction/importance, goal

Economic and social reforms in Vietnam had huge impact on its previously completely subsidized health care system by introducing user fees, privatizing medical practices and liberalizing pharmaceutical policies. The most significant result were higher health expenditure paid out-of-pocket by users and lower consumption of essential services among high need groups.

To deal with these problems, the government introduced a health insurance system in 1992, consisting of mandatory and voluntary schemes for different target population. By mid 2002, the system has enrolled a total of 12.6 million individuals, roughly 16% of the total population.

It was expected that the system would lower private health expenditure and thus improve equity and social welfare. If this was indeed the case, insurance coverage that resulted in a reduction in health care expenditure should have also resulted in a proportionally larger increase in utilization among the lower-income individuals. The current research project is set to test such expectations by answering the following three questions about health insurance affiliation and health care utilization: 1) Were insured patients more likely to seek medical care vs. self-treatment than the uninsured when fell ill 2) Among individuals that experienced illness and used medical services, were insured patients more likely to use services from public providers vs. private providers than uninsured patients 3) Were insured individuals more likely to use preventive service than the uninsured


The proposed project will use data from the Vietnam National Health Survey (VNHS, 2001 ~ 2002), which included 36,000 households (158019 individuals) with household information as well as individual health status, health insurance affiliation and health utilization information.

Health care utilization will be set as a binary depend variable for each research question with health insurance affiliation set as a primary independent variable, adjusting for other individual, household and community variables such as age, gender, household wealth quintile and rural/urban status. Interaction between health insurance affiliation and household wealth quintile will be investigated as it pertains to the research aims.

Two special methodological issues arise when modeling the outcome of health care utilization against health insurance affiliation. The first and most important one is selection bias. There could have been immeasurable factors affecting both individual’s self-selection of health insurance affiliation and his/her choice of health care utilization when ill. Such bias is also known as “endogeneity” in econometrics terms. To correct for the selection bias issue, instrumental variables will be identified and used in the modeling process. For each research question, a bivariate probit estimation will be fitted simultaneously on a primary equation using utilization as an outcome and a secondary equation using health insurance affiliation as an outcome, incorporating some instrumental variables in the secondary equation. If the bivariate probit regression does identify correlation between the error terms of the primary and secondary equations, selection bias is then confirmed and, by virtue of the bivariate probit process, corrected at the same time. The second issue is the clustering of observations. Individuals within the same household were clustered and correlated in some way, instead of independent, as assumed by ordinary regression methods. The same is true for households within the same community. This gives rise to correlation among the error terms of the regression equations, which will result in incorrect estimation of the variances of regression coefficients. To correct for such correlation, the Huber-White “sandwich” variance estimator will be used. The estimator estimates coefficient variances by maximum likelihood and allows for unspecified correlation structure within clusters and sub-clusters and therefore works in the settings of the current study.

All statistical procedures will be carried out using STATA 8.0 package.


The proposed research project should be able to confirm (or disprove) the expected effects of health insurance programs, especially in the perspective of social welfare and equity. Lower-income individuals often had larger unmet needs for health care services, therefore a larger positive impact from enrolling into a health insurance program on this group would improve social welfare and equity in health care services and resource allocation. The study would also generate policy implications such as indications for insurance program modification in terms of public/private provider access and preventive care incentives.

Moreover, there is very limited information on the impact of public health insurance programs on health care utilization in Vietnam and other developing countries in south-east Asia. This study develops and uses methodologies for evaluating program effectiveness using household survey data and therefore serves as a basis for future researches in the field.

Allocation of expenses by category




Data collection

Purchase of raw data


Equipment and environment

STATA 8.0 package upgrade


Computer hardware upgrade


Additional funding

There is no additional funding for this project.

Judith Easterbook, PhD candidate in Molec. Microbio. & Imm. - Regulatory T Cells Mediating Seoul Virus Persistence in Norway Rats

The Role of Regulatory T Cells in Mediating Seoul Virus Persistence in Norway Rats


Hantaviruses are noncytopathic and cause two diseases in humans, hantavirus pulmonary syndrome (HPS) and hemorrhagic fever with renal syndrome (HFRS). Human pathology is hypothesized to be caused by sustained, elevated levels of proinflammatory cytokines (5). Hantaviruses are transmitted by rodents and each hantavirus has coevolved with a specific rodent host. In contrast to humans, hantaviruses infect their rodent hosts and cause persistent infection in the absence of pathology. The mechanisms mediating hantavirus persistence in rodents are unknown, but may involve suppression of host proinflammatory immune responses. My data illustrate that production of IL-1 in the spleen is suppressed during persistent Seoul virus infection in Norway rats (Rattus norvegicus), the natural rodent host for this species of hantavirus (3). Mechanisms mediating suppression of IL-1 remain unknown.
Regulatory T cells can suppress proinflammatory responses to infection as well as contribute to persistence of a number of pathogens (7). Regulatory T cells are currently characterized by the surface markers CD4 and CD25 and the intracellular marker forkhead box p3 (foxp3). These cells constitute approximately 5-10% of peripheral CD4+ T cells (9) and activity is primarily through local production of anti-inflammatory mediators, including IL-10 and TGF- (2). During Leishmania infection in mice, regulatory T cells are recruited to suppress proinflammatory T cell responses, which prevents the host from eliminating the parasite. The host, however, displays minimal pathology and remains immune to reinfection due to the regulatory T cells remaining at the site of infection (1). Following administration of anti-CD25 antibodies, Leishmania parasites are cleared, but the host suffers immune-mediated pathology and is susceptible to subsequent reinfection (1). The balance between effector and regulatory T cells seems to reflect an evolutionary compromise between survival of the host and survival of the pathogen. The ultimate goal of my proposed experiments is to assess the role of regulatory T cells in mediating Seoul virus persistence and preventing pathology in Norway rats.

Methods: Experiment 1

Mouse anti-rat CD25 IgG monoclonal antibody (mAb) will be produced by the cell line NDS-63 (Dr. Kathryn Wood, Oxford, UK) in supplemented medium using the hollow fiber system at the JHU Core Cell Center. The antibody will be purified using a Protein G column and adjusted to a 4 mg/ml solution. An appropriate dose of anti-CD25 antibody and time course for administration will be determined in Experiment 1. Recent data has revealed that administration of anti-CD25 mAb results in shedding of CD25 from the cell surface, not depletion of the regulatory T cell population, which was the current dogma (6). Because regulatory T cell activity is dependent on CD25, administration of anti-CD25 mAb causes functional suppression (11). Efficiency of regulatory T cell depletion of CD25 from CD4+foxp3+ regulatory T cells following administration of anti-CD25 mAb has not been determined in rats. One study used what is considered to be a less effective antibody against rat CD25 as compared with NDS-63 (i.e. OX-39) (10) and achieved a 33% depletion of CD25 on the surface of CD4+ T cells measured one week after a single intraperitoneal (i.p.) injection of 1 mg/rat anti-CD25 mAb; measurements, however, were not made at earlier time points (4). Studies in mice report almost complete CD25 depletion from CD4+ T cells (>80%) up to 10 days following administration of a single i.p. injection of 400 g/mouse of anti-CD25 mAb (8). The doses of anti-rat CD25 mAb that I propose to test are based on available data from rats and mice, as well as personal communication with Dr. Yasmine Belkaid, an expert in regulatory T cells. In Experiment 1, rats will not be inoculated with Seoul virus, but the days for sample collection will correspond with the days for sample collection during an infection experiment (i.e. 0 days post inoculation (p.i.) as uninfected, 7 days p.i. during the acute phase of infection, and 30 days p.i. during the persistent phase of infection). Adult male Long Evans rats (Rattus norvegicus) rats will be housed individually in polypropylene cages covered with polyester filter bonnets in a pathogen-free animal facility with food and water available ad libitum. Animals will be maintained on a constant 14:10 light:dark cycle with lights on at 0600 hours Eastern Standard Time, constant temperature (21±2ºC), and constant humidity (50±5%). All procedures have been approved by the JH ACUC (protocol #RA04H338).

Rats will be anaesthetized with isofluorane vapors and injected i.p. with 1 or 2 mg/rat of anti-CD25 mAb (N = 10/dose) or vehicle alone (N = 10) at Experimental Day -1 and at Experimental Day 1. On Experimental Days 10 and 20, the Experimental Day 30 rats will be injected i.p. with the same dose of anti-rat CD25 mAb to maintain CD25+ T cell depletion. On Experimental Days 7 and 30, rats administered the anti-CD25 antibody (N = 5/dose/time point) and rats administered vehicle alone (N = 5/time point) will be euthanized by using CO2 (Figure 1).

Figure 1. Time line for the administration of anti-CD25 mAb and sample collection

Because regulatory T cells act locally at the site of infection, I will not only collect the spleen, which is a secondary lymphoid organ with low Seoul virus replication, but also the lung, which is a primary site of Seoul virus replication. Following lymphocyte separation using a Lympholyte density gradient, CD4+ cells will be enriched by MACS. Regulatory T cells will be fluorescently labeled using antibodies against CD4, CD25, and intracellular foxp3 and fluorescent staining will be measured by FACS and analyzed using FlowJo software. Because administration of anti-CD25 mAb has been reported to induce shedding of CD25 from the cell surface, not only will conventional markers (CD4+CD25+foxp3+) be used to characterize the functional regulatory T cell populations, but also CD4+CD25-foxp3+ staining will be included to characterize the functionally inactive regulatory T cell population that has shed CD25 in response to anti-CD25 mAb administration.

After an appropriate dose and time course for administration have been determined to effectively deplete functional regulatory T cells, I will begin Experiment 2. In Experiment 2, I will monitor pathology and immune responses in rats that are uninfected or infected with Seoul virus and subsequently administered anti-CD25 mAb or vehicle alone. If regulatory T cells suppress antiviral effector responses to mediate Seoul virus persistence, I hypothesize that the CD4+CD25+ regulatory T cell population will be greater in persistently infected rats (i.e. at Day 30 p.i.) in tissues that support high virus replication (i.e. lungs) as compared with uninfected rats (i.e. at Day 0 p.i.) or during the acute phase of infection (i.e. Day 7 p.i.). Additionally, production of anti-inflammatory cytokines, including IL-10 and TGF- , should be elevated and production of pro-inflammatory cytokines, including IL-1 and IFN- , should be reduced or at baseline in persistently infected as compared with uninfected or acutely infected rats. To further test the hypothesis that regulatory T cells mediate Seoul virus persistence, if the CD4+CD25+ population is depleted, rats should display proinflammatory cytokine-mediated pathology in target tissues following Seoul virus inoculation.


1. Belkaid, Y., C. A. Piccirillo, S. Mendez, E. M. Shevach, and D. L. Sacks. 2002. CD4(+)CD25(+) regulatory T cells control Leishmania major persistence and immunity. Nature 420:502-507.
2. Belkaid, Y., and B. T. Rouse. 2005. Natural regulatory T cells in infectious disease. Nature Immunology 6:353-360.
3. Easterbrook, J. D., and S. L. Klein. submitted. Proinflammatory and anti-inflammatory responses are suppressed during elevated Seoul virus replication in male Norway rats. Journal of Virology.
4. Ghiringhelli, F., N. Larmonier, E. Schmitt, A. Parcellier, D. Cathelin, C. Garrido, B. Chauffert, E. Solary, B. Bonnotte, and F. Martin. 2004. CD4(+)CD25(+) regulatory T cells suppress tumor immunity but are sensitive to cyclophosphamide which allows immunotherapy of established tumors to be curative. European Journal of Immunology 34:336-344.
5. Khaiboullina, S. F., and S. C. St Jeor. 2002. Hantavirus immunology. Viral Immunology 15:609-625.
6. Kohm, A. P., McMahon JS, Podojil JR, Begolka WS, DeGutes M, Kasprowicz, Ziegler SF, Miller SD. 2006. Cutting edge: Anti-CD25 monoclonal antibody injection results in the functional inactivation, not depletion, of CD4+CD25+ T regulatory cells. Journal of Immunology:3301-3305.
7. Mills, K. H. G. 2004. Regulatory T cells: Friend or foe in immunity to infection Nature Reviews Immunology 4:841-855.
8. Morgan, M. E., R. P. M. Sutmuller, H. J. Witteveen, L. M. van Duivenvoorde, E. Zanelli, C. J. M. Melief, A. Snijders, R. Offringa, R. R. P. de Vries, and R. E. M. Toes. 2003. CD25+cell depletion hastens the onset of severe disease in collagen-induced arthritis. Arthritis and Rheumatism 48:1452-1460.
9. Sakaguchi, S. 2003. Regulatory T cells: mediating compromises between host and parasite. Nature Immunology 4:10-11.
10. Tellides, G., M. J. Dallman, and P. J. Morris. 1989. Mechanism of action of Interleukin-2 receptor monoclonal antibody therapy: target cell depletion or inhibition of function Transplantation Proceedings 21:997-998.
11. Thornton, A. M., E. E. Donovan, C. A. Piccirillo, and E. M. Shevach. 2004. Cutting edge: IL-2 is critically required for the in vitro activation of CD4(+)CD25(+) T cell suppressor function. Journal of Immunology 172:6519-6523.

Gabrielle Hunter, MHS candidate in International Health - Effects of Water Availability and Education on the Hygiene Behavior of Residents in a Peri-Urban Community in Lima, Peru

The Effects of Water Availability and Education on the Hygiene Behavior of Residents in a Peri-Urban Community in Lima, Peru

Introduction: Importance & Goals

The scarcity of water in conjunction with a lack of proper hygiene practices continues to result in infection and the spread of diarrheal diseases, a major cause of infant mortality in developing countries. As a result, developing countries face the challenge of creating feasible, multi-faceted solutions to reduce the prevalence of these diseases.

A recently published study conducted in Las Pampas de San Juan de Miraflores, a shantytown approximately 15 kilometers south of Central Lima, revealed that inadequate water and sanitation conditions increased the risk of diarrhea. Lima is located in a desert, and as such water is a particularly scarce resource in the shantytowns, known as pueblo jovenes, surrounding the city. The results of an earlier study in these communities revealed that providing information about hygiene without increasing the availability of water will not result in better sanitation practices, and that, in water-scarce communities, hygiene practices could be improved with the provision of convenient, cheap, in-house water. Hygiene promotion can also improve behaviors, and accordingly hygiene promotion is now recognized as an essential component of water and sanitation programs if maximum health benefits are to be achieved.

In 2004, an open, longitudinal research project was begun in Las Pampas de San Juan de Miraflores to determine if and how the quantity of water available, before and after the installation of at-home water connections, affects water usage and personal hygiene behaviors. Such an understanding of the elements that control water handling practices could increase the benefits of infrastructural installation, as well as the implementation of educational programs that aim to better existing conditions and perceptions of water usage. The results of the study at this point have revealed for example that, prior to the installation of at-home water connections, only 11% of defecation events of mothers were followed by handwashing with soap and that more than half of their defecation events were not followed by any form of handwashing. Further information on hygiene behavior, subsequent to the installation of water connections, is currently being collected.

Proposed here is the design, implementation and evaluation of the final stage of the project: a participatory hygiene education intervention. The proposed research project is an attempt to determine if personal hygiene behaviors can be improved through a hygiene promotion intervention, once an at-home water connection is made available. These results will help to interpret and evaluate future interventions relating to water and its usage.

The specific goals of the proposed project are to

  1. Determine if a participatory educational intervention following an increase in affordable, convenient water supply can improve personal hygiene behaviors in a peri-urban community in Lima, Peru.
  2. Develop, through community participation, a hygiene education and promotion program
  3. Improve personal hygiene behaviors of residents in the community.
  4. Foster community empowerment, cohesion and collective action in this disenfranchised setting.


The primary population for this research project are the residents of Manuel Scorza, a community located in the outskirts of Las Pampas de San Juan de Miraflores. This community has just recently received water and sanitation systems, and residents participated in the installation of these systems.

In earlier stages, baseline information was collected through three days of 12-hour continuous monitoring of hygiene behaviors and water usage in residents’ homes before the installation of at-home water and sewage connections. Observations were repeated again the following year, several months after the connections had been activated. This project will begin by employing formative qualitative and quantitative research techniques to understand and interpret findings from these earlier stages of the project; these results will then be shared with community residents in order to facilitate a collaborative effort to develop the form and message of the hygiene-behavior educational intervention. The impact of this hygiene intervention will then be evaluated through continuous monitoring structured observations in order to gather information on the actual change in personal behaviors due to the hygiene promotion intervention.

As such, the proposed project will be comprised of four stages:

Stage 1 – Description of Existing Community Hygiene Behaviors

In order to provide an in-depth and complete understanding of existing hygiene practices, the project will begin with formative qualitative and quantitative research techniques in order to interpret and augment the findings from the earlier stages of the project.

A Knowledge, Attitudes, and Practices (KAP) questionnaire will be used to interview the female heads of approximately fifty households randomly selected from similar but geographically separate communities of Las Pampas de San Juan de Miraflores to determine existing community-wide knowledge of and beliefs regarding hygiene behaviors. The results from the questionnaire will be evaluated in comparison with the results from earlier observations to examine how knowledge and beliefs correspond to actual behaviors.

Focus Group Discussions (FGD) with eight community volunteers from Manuel Scorza in separate groups of men, women, children, and adolescents will provide the first opportunities to share the findings from the earlier observational studies and the KAP surveys with community residents and receive their feedback, which will help with the identification of risk behaviors and which behaviors would be suitable for an intervention.

Purposive sampling, based on the findings of the earlier observations, will be used to select respondents for in-depth interviews. Interviews will be held with five community members from households shown to possess better than average hygiene behaviors and five community members from households that displayed worse hygiene behaviors. These interviews will employ a consultative research approach; first, gathering information on what factors: personal, social, and environmental, motivate or hinder handwashing, and then identifying ways that hygiene practices can be improved and this change sustained.

Stage 2 – Development of Hygiene Intervention

The development of the hygiene intervention will begin with the formation of a community-based hygiene working group, composed of volunteers recruited from the FGD. This group of volunteers will already be comprised of residents who view hygiene behavior as a significant problem in the community and who agree that the situation can and should be improved. Therefore, the principal dialogue will involve developing the means with which hygiene practices can be improved. Group activities, such as “planning posters” and “problem boxes,” will be employed to further discussions regarding the form of a locally-acceptable intervention message and the most effective channels for its distribution within the community.

Stage 3 – Implementation of Hygiene Intervention

As described above, the exact form of the hygiene intervention will be determined during the course of the project by the community-based hygiene working group. As such, the intervention will begin with a participatory approach and will later assume parts from other models, such as health education and social marketing.

Stage 4 – Evaluation of Hygiene Intervention

Continuous monitoring structured observations of the personal hygiene behaviors of residents will be used to determine the impact of the hygiene intervention. Post-intervention results will be compared with those of the baseline observations to determine the changes in behavior due to the intervention. Additionally, the KAP questionnaire will also be repeated to measure the changes in hygiene knowledge in a similar manner.

Evaluation will be based on various measures of 3 outcomes: risk awareness, improved hygiene behaviors, and improved community cohesion. It is expected that the community-based hygiene working group that will be formed to develop and implement the education program will continue to function during and following the conclusion of the project, representing a minimum level of increased community cohesion.


Water and sanitation projects and hygiene education aimed at behavior change are two important and widely-implemented health interventions. This project is aimed to further knowledge about how to best combine these two interventions to achieve maximum health-related behavior change and outcomes.

Methodologically, continuous monitoring is rarely used for intervention evaluations or measuring behavior change due to the complexity and expense involved, but this form of structured observation has the benefit of providing information on people’s actual behaviors as opposed to reported behaviors. As such, this proposal presents a unique opportunity to examine exactly the extent to which behaviors can be improved with a collaborative health education and promotion intervention. Additionally, community-based participatory research is currently receiving unprecedented attention, but much remains unknown about this methodology of attaining sustainability, so this project will add to this body of knowledge.

The study staff for the research project will also include a local nurse, who will be trained in participatory methods and qualitative research techniques in order to carry-out in-depth interviews, facilitate focus group discussions, and community meetings. Four other nurses will be recruited to help with the structured observations. In this way, local capacity building is built into the project.

A large part of the significance of this project lies in the potential benefits to the community. A collaborative approach will be employed where the findings from the earlier stages will be shared with community residents. This is intended to raise awareness about hygiene behaviors in the community. A hygiene working group will be formed, and together with the investigators, this group will use existing data and any additional data they wish to collect to develop the education and promotion intervention. Designed in this participatory and collaborative manner, the educational messages are more likely to be sustainable and locally acceptable. Furthermore, community participation in the development of the intervention may impart upon participants some critical skills, such as leadership, problem-solving, conflict resolution, and potentially the independence and initiative to formulate subsequent interventions and community-development projects. The community will benefit not only from improved hygiene and subsequent health outcomes but from the strengthening of community ties. Due to the nature of social development in Peru, a community’s organization and ability to advocate with local government is critical for the achievement of local development, in the form of infrastructure and service provision. If successful, this approach may be used in other disenfranchised peri-urban communities in Lima to promote community capacity building for sustainable positive change.

The proposed dissemination of findings is also key to this project’s significance. Final results and education messages will be shared with a national handwashing and hygiene behavior campaign in Peru, to help inform their tailoring of messages for peri-urban shantytown communities of Lima. Additionally, infrastructure installations are on the rise in these areas of Lima, and improved understanding of water usage and hygiene practices in such a peri-urban community may be useful for infrastructure projects in other shanty-towns, into which a similar hygiene education intervention could be included. As such, findings related to changes before and after the hygiene intervention and before and after the installation of at-home will also be presented to Lima’s drinking water administration.

On a final note, there is much personal significance to this project. Community-based participatory research and action is my primary interest in public health, and hygiene behaviors are a close second. I have been searching for an opportunity to gain practical experience in both.

"Center Effect" in Kidney Transplantation: A New Look at an Old Problem"

Title of Proposal: The “center effect” in kidney transplantation: A new look at an old problem

End-stage renal disease (ESRD) is a growing public health problem and is associated with a mortality rate of ~20% per year, despite advances in dialysis therapy (1;2). Kidney transplantation is the preferred treatment for ESRD patients since it enhances quality of life, improves long-term survival, and is more cost-effective than dialysis (3-5). Although various clinical and immunologic predictors of transplant success have been elucidated over the last several years, there is still uncertainty about the impact of provider (including the transplant center or hospital) characteristics on kidney transplant outcomes (6-20).

Study Goals:
The study hypothesis is that there are identifiable(and potentially modifiable) provider characteristics, at the level of clinicians and transplant centers, that may predict the outcome of kidney transplant recipients and explain, at least in part, the variation in patient and kidney allograft survival across kidney transplant centers in the United States.

To test the above hypothesis, this study will explore the relation between provider factors and shortand long-term patient and allograft survival (at 1-, 3-, and 5-years post-transplant) in U.S. adult (age 18 years) kidney transplant recipients via the following two aims: (i) longitudinally assess the impact of transplant center (e.g., average annual procedural volume, geographic locale, hospital size) and physician/surgeon characteristics (e.g., clinical volume, years of practice, board certification status) on the patient and allograft survival of 106,362 KTR (transplanted from January 1, 1994 to December 31, 2002 followed until December 31, 2003 by the United States Renal Data System) using a multi-level modeling approach; the effect of provider factors on outcome will also be studied by year of transplantation and in high-risk patient subgroups (e.g., second transplants); (ii) conduct a national survey of medical and surgical directors of all 282 adult kidney transplant centers throughout the United States; information on clinical protocols (e.g., use of induction therapy), follow-up care, and other practice patterns will be ascertained; items that measure the degree and efficacy of teamwork between medical and surgical staff
will also be included.

Study objective 1 will use information from the United States Renal Data System (USRDS) on all kidney transplants performed in the United States from January 1, 1994 to December 31, 2002, along with the American Hospital Association (AHA) and American Medical Association (AMA) databases, to evaluate the role of transplant center and clinician (i.e., physician/surgeon) characteristics in influencing patient and allograft survival in a cohort of KTR. Center characteristics of interest will include average annual procedural volume, the number of years in existence as a kidney transplant program, total number of hospital beds, profit status, number of transplant programs within the same city, presence of an accredited transplant (nephrology/surgery) fellowship program, multi-organ vs. kidney-only transplant program, and geographic locale. Transplant clinician factors will include physician and surgeon volumes (number of KTR cared for per year), years of physician and surgeon experience (years of practice), board certification status, and physician/surgeon specialty. Transplant physicians (i.e., clinicians trained in general internal medicine and/or nephrology) are distinguished from transplant surgeons (i.e., clinicians trained in general surgery, transplant surgery or urology) since the role of each group tends to differ in the short- and long-term management of KTR at most U.S. centers. A secondary aim will be to determine the effect of provider characteristics on the patient and allograft survival of high-risk subgroups (such as second transplants and recipients of expanded criteria donor kidneys) and the role of transplant era (defined by time points at which novel immunosuppressive agents were introduced into clinical practice) in potentially altering the relation between transplant center/clinician-level factors and outcome. Study objective 2 will involve a national survey of medical and surgical directors from all 282 adult kidney transplant centers for the purpose of acquiring more detailed information on transplant center and clinician characteristics (e.g., clinical practice patterns, measures of the degree and efficacy of teamwork among medical and surgical transplant clinicians). An existing survey instrument will be revised and extended in order to carry out this task (21). Measures of reliability and validity will be assessed during the piloting phase of survey development. Three periodic survey mailings, along with telephone, e-mail, and/or fax reminders, will be used to maximize response rates. Endorsement of the survey by the American Society of Transplantation and the American Society of Transplant Surgeons will be sought. The results of the survey will provide a cross-sectional look at current practice patterns and the organizational characteristics of U.S. adult kidney transplant centers. Moreover, it will provide a measure of the teamwork that exists among medical and surgical transplant clinicians and how this teamwork is manifested in the management of KTR. Future research will prospectively link these data to short- and long-term outcomes of KTR to evaluate the relation between center-level practice patterns, organizational characteristics, and measures of teamwork with patient and allograft survival.

Statistical Analyses:
For study objective 1, exploratory data analyses with descriptive statistics will be undertaken to examine the distribution of exposure, confounder, and outcome variables. Potential confounders will be examined within quartiles of total and average annual transplant center and clinician volume. Univariable statistical relations will be tested using appropriate parametric and non-parametric tests for continuous and categorical data. The Kaplan-Meier product limit method will be used to graphically examine the relation between center/provider volume quartiles and both allograft and patient survival. The Cox proportional hazards model will be used to examine the independent effect of covariates on time-to-event outcomes (i.e., allograft failure and death). Two statistical approaches for the analysis of clustered data will be pursued and the results compared: (i) marginal models using generalized estimating equations (GEE) to account for the correlation of patient and allograft outcomes within transplant centers and physicians/surgeons; (ii) random effects or multilevel models. Generally, marginal models will provide hazard ratios with appropriate standard errors but, unlike random effects models, they can only address questions regarding population average responses (i.e., no statements about center-specific effects can be made) and do not permit the examination of variance components within and between clusters (22).

The random effects or multi-level Cox model provides a method to account for the nested data structure, examine the variance components at each level of the data (i.e., center, clinician, and patient), and help to stabilize estimates from centers with smaller volumes (via a procedure that uses the posterior distribution of the overall effect estimate and the standard errors for each center) (23;24). With an anticipated sample size of approximately 106,000 patients distributed over ~ 280 centers and a type 1 error rate of 0.05, this study will have 90% power to detect a hazard ratio of 1.15 for the outcome of graft survival when comparing the lowest center volume quartile to the highest quartile (25).

For study objective 2, descriptive analyses will be performed using histograms, box plots, and lowess smoothing techniques along with standard parametric and non-parametric methods for comparing continuous and categorical data. Bivariate analyses will compare characteristics of kidney transplant programs across center volume categories (e.g., tertiles or quartiles) and the presence of a linear relationship in the responses of surgical and medical directors will be measured using Pearson’s correlation coefficient. Assuming that a response rate of 54% is achieved (26), 282 x 0.54 = 152 centers will return completed questionnaires. Given that differences between responders and non-responders with respect to key factors such as the geographic location of transplant programs are not significant, a sample size that constitutes > 50% of all centers should provide sufficient data to adequately describe the variation in the characteristics of interest across U.S. kidney transplant centers.

The anticipated results of this study will further the scientific debate regarding the factors that contribute to the center effect in kidney transplantation. The study will identify potentially modifiable characteristics of transplant centers and/or clinicians that may help to mitigate existing variations in center-specific outcomes. This information may also be used develop and implement interventions to improve the performance of transplant centers and enhance the quality of care provided to kidney transplant recipients. Finally, the results of this study will inform the efforts of policy makers in weighing the pros and cons of regionalizing complex and expensive health services such as kidney transplantation.

(1) USRDS 2005 Annual Data Report: Atlas of End-Stage Renal Disease in the United States. 2005. Bethesda, MD, National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases. Ref Type: Report

(2) Gilbertson DT, Liu J, Xue JL, Louis TA, Solid CA, Ebben JP et al. Projecting the number of patients with end-stage renal disease in the United States to the year 2015. J Am Soc Nephrol 2005; 16(12):3736-3741.

(3) Evans RW, Manninen DL, Garrison LP, Jr., Hart LG, Blagg CR, Gutman RA et al. The quality of life of patients with end-stage renal disease. N Engl J Med 1985; 312(9):553-559.

(4) Laupacis A, Keown P, Pus N, Krueger H, Ferguson B, Wong C et al. A study of the quality of life and cost-utility of renal transplantation. Kidney Int 1996; 50(1):235-242.

(5) Wolfe RA, Ashby VB, Milford EL, Ojo AO, Ettenger RE, Agodoa LY et al. Comparison of mortality in all patients on dialysis, patients on dialysis awaiting transplantation, and recipients of a first cadaveric transplant. N Engl J Med 1999; 341(23):1725-1730.

(6) Mickey MR. Center effect. Clin Transpl 1986;165-173.

(7) Mickey MR. Center variability. Clin Transpl 1989;435-446.

(8) Gjertson DW, Terasaki PI. The large center variation in half-lives of kidney transplants. Transplantation 1992; 53(2):357-362.

(9) Gjertson DW. Center and other factor effects in recipients of living-donor kidney transplants. Clin Transpl 2001;209-221.

(10) Opelz G, Mickey MR, Terasaki PI. Comparison of kidney transplant survival among transplant centers. Transplantation 1975; 19(3):226-229.

(11) Gilks WR, Selwood N, Bradley BA. The variation among transplant center results in the United Kingdom and Ireland from 1977 to 1981. Transplantation 1984; 38(3):235-239.

(12) Briganti EM, Wolfe R, Russ GR, Eris JM, Walker RG, McNeil JJ. Graft loss following renal transplantation in Australia: is there a centre effect Nephrol Dial Transplant 2002; 17(6):1099-1104.

(13) Gjertson DW. Update: center effects. Clin Transpl 1990;375-383.

(14) Hunsicker LG, Edwards EB, Breen TJ, Daily OP. Effect of center size and patient-mix covariates on transplant center-specific patient and graft survival in the United States. Transplant Proc 1993; 25(1 Pt 2):1318-1320.

(15) Terasaki PI, Cecka JM. The center effect: is bigger better Clin Transpl 1999;317-324.

(16) Axelrod DA, Guidinger MK, McCullough KP, Leichtman AB, Punch JD, Merion RM. Association of center volume with outcome after liver and kidney transplantation. Am J Transplant 2004; 4(6):920-927.

(17) Evans RW, Manninen DL, Dong F. The center effect in kidney transplantation. Transplant Proc 1991; 23(1 Pt 2):1315-1317.

(18) Kim SJ, Schaubel DE, Jeffery JR, Fenton SS. Centre-specific variation in renal transplant outcomes in Canada. Nephrol Dial Transplant 2004; 19(7):1856-1861.

(19) Ogura K, Cecka JM. Center effects in renal transplantation. Clin Transpl 1991;245-256.

(20) Benlahrache C, Cecka M, Mickey MR, Cicciarelli J. The center effect. Clin Transpl 1987;325-337.

(21) Loberiza FR, Jr., Zhang MJ, Lee SJ, Klein JP, LeMaistre CF, Serna DS et al. Association of transplant center and physician factors on mortality after hematopoietic stem cell transplantation in the United States. Blood 2005; 105(7):2979-2987.

(22) Goldstein H. Multilevel Statistical Models. 2 ed. London: Edward Arnold Publishing Ltd, 199.

(23) Louis TA, Shen W. Innovations in bayes and empirical bayes methods: estimating parameters, populations and ranks. Stat Med 1999; 18(17-18):2493-2505.

(24) Yau KK. Multilevel models for survival analysis with random effects. Biometrics 2001; 57(1):96-102.

(25) Hsieh FY, Lavori PW, Cohen HJ, Feussner JR. An overview of variance inflation factors for sample-size calculation. Eval Health Prof 2003; 26(3):239-257.

(26) Asch DA, Jedrziewski MK, Christakis NA. Response rates to mail surveys published in medical journals. J Clin Epidemiol 1997; 50(10):1129-1136.

Rachel Singer, PhD candidate in Health Policy and Management - Mobilizing the Masses: Potential for Universal Coverage in Maryland

Mobilizing the Masses: The Potential for Universal Coverage in Maryland

According to a March 2006 Gallup Poll, 68 percent of Americans worry “a great deal” about “the availability and affordability of healthcare.” At the beginning of the twenty-first century, the United States is the only industrialized nation that does not provide health care for all its citizens. While proposals for universal health insurance have routinely come before Congress since the First World War, they failed to pass time and time again. Over the last century, efforts towards universal coverage resulted in incremental, rather than monumental, reforms, providing coverage only for individual segments of the population. In 2004, 46 million Americans did not have health insurance (U.S. Census Bureau 2005). According to a recent report by the Institute of Medicine (2004), uninsured children and adults do not receive timely care, and die earlier than those with insurance. Insurance matters not just to the health of the individual, but to the very survival of community and society. The lack of a unified grassroots effort to combat this injustice has enabled pharmaceutical and insurance companies, and their supporters to get away with murder.
 Scholars frequently cite the lack of a unified, grassroots movement to counter the well-organized and well-financed stakeholders in opposition to universal coverage as the reason why past health reform efforts have failed. Analyses of past failures have concluded that an organized grassroots movement will be necessary for the success of future reform (Derickson 2005; Gordon 2003; Mayes 2003; Quadagno 2005). As employment-based insurance continues to de-stabilize in tandem with changes in the economy and increasing healthcare costs, the number of individuals without insurance has increased dramatically and the need to address the problem has become more critical. While the public has expressed the desire for changes in the system, change is not possible without grassroots efforts on the part of citizens, locally and nationally. Since the late 1980s, various state movements emerged to advocate for universal healthcare. In Maryland, the “Maryland Healthcare For All!” initiative was established in 2000 with the goal of “create(ing) a comprehensive, economically sound health care” plan for every citizen. In order for this movement to be successful, its direct connection with as many citizens in the state of Maryland is integral.

Beatrix Hoffman (2003) critiques certain statewide health reform movements for “emphasizing coalitions of professional advocacy groups as the centerpiece of their organizing strategies.” Hoffman explains that reform leaders often fail to accomplish their goals because of reform leaders’ “lack of knowledge, lack of interest in, or outright exclusion of popular reform constituencies and grassroots organizing strategies.”

Efforts to achieve national health insurance throughout the twentieth century were generally products of elite policymakers, social scientists, public health professionals, and labor leaders. My goal is to determine whether the Maryland Citizens’ Health Initiative is a true “grassroots” movement, or if it is an organization designed by well-intentioned non-grassroots advocates. If the latter is indeed the case, it is critical that the initiative make it a priority to connect directly with citizens, not merely other interest groups whose individual goals will stand in the way of universal healthcare for all. In its communication with interest groups, the initiative must also make it a priority to speak to those groups in the language of their own movements.

Using the case study method, I plan to analyze the movement from its inception, to trace the lines of communication between the movement’s leadership and the citizens, and to determine the movement’s potential success or potential failure in context of other grassroots movements and their results throughout twentieth century American history. The case study method is best suited to answering “how” and “why” questions, and for understanding phenomenon and their context (Yin 2003). I selected the Maryland “Health Care for All” coalition as my single case study. Sources of evidence for this study, as described by Yin, include documents (e.g. organizational materials, news articles, newsletters), in-depth interviews, and participant observation.

Some social movement researchers argue that social movements have become so professionalized since the early 1970s to render them useless. Has corporate capitalism so overwhelmed the policy-making process that ordinary citizens are no longer vital contributors to decision-making Or, does the potential exist for an unprecedented grassroots movement to combat runaway abuse of American democratic principles by all-powerful drug and insurance companies