Environmental Biostatistics Training Program
For more information, please click here.
Statistical Methods in Environmental Epidemiology
Evidence from environmental epidemiology research often contributes to the foundation of major policy decisions, driving policy makers to pose challenging questions to researchers. These questions are often best answered by using statistical methods that characterize the risk of a targeted environmental agent while taking other environmental variables into account. The nature and characteristics of environmental data and health outcomes make the risk estimation challenging and require the development of novel statistical methods. The purpose of this research is to develop models for integrated analyses of spatio-temporal data on exposure, health outcomes and covariates, incompletely observed and available at different levels of aggregation. Such models are needed for addressing a broad class of environmental agents that vary over time and across geographical regions. The focus is on the development of new statistical methods for
- estimating temporal associations between health outcomes and current and past environmental exposures, when the underlying function is unknown and exposure is measured with error;
- estimating spatial associations between health outcomes and environmental exposures which take proper account of non-random sampling designs, and
- conducting integrated analyses of spatio-temporal data on health and environmental exposures taking into account sources of bias arising from spatial and temporal aggregations.
- We apply some of the proposed statistical methods to data on air pollution, mortality and temperature
National Medicare Cohort Air Pollution Study
This application addresses a key uncertainty in the evidence on the effects of particulate matter (PM) on public health: that of longer-term exposure to airborne PM, including PM2.5, on morbidity and mortality. By using large, existing cohorts established by federal agencies, we will
- test the hypothesis that long-term and short-term exposure to PM increases mortality and morbidity;
- estimate the joint effects of PM and other pollutants; and
- compare risks of PM exposure in persons with and without underlying heart and lung disease.
The focus of analysis will be the longitudinal data in the Medicare Cohort (MC), comprising a 5% sample of Medicare participants followed for hospitalization and death. In 1998, the MC included 2 million persons; we project 80,000 deaths in 327 counties with substantial data over follow-up from 19989-2001. We will use hospitalization data to construct comorbidity measures. A separate smaller survey, the Medicare Current Beneficiary Survey (MCBS), provides information on a large number of potential confounding factors. We will also use the Veterans Health Study Cohort (VC), which includes 900,000 persons. Data on hospitalization and mortality from these cohorts will be combined with the air monitoring data routinely collected by the Environmental Protection Agency and weather data from the National Oceanic and Atmospheric Administration (NOAA). Over the 3 years of the project, we will obtain data and create the analytic data base, address methodological issues related to imputation of potential confounding factors, and estimate the acute and chronic effects of PM2.5 and other pollutants on morbidity and mortality, including characterizing effects on differing timescales and regions of the country.
We have identified epidemiologic cohorts with reasonably precise estimates of the health risks posed by PM2.5 and other pollutants. They will provide sufficient events to directly address risks of PM2.5 without using surrogate indicators. The data will offer insights into cardiac and respiratory comorbidity as a determinant of susceptibility and into the differing magnitudes of risks found in the time-series and PM cohort studies. The findings will advance understanding of the public health consequences of air pollution and support the development of more certain risk models for air pollution and health.
Internet-based Health and Air Pollution Surveillance System (iHAPSS): A New Model for Health Monitoring
The purpose of this research program is to create an internet-based system for monitoring the effects of air pollution on mortality and morbidity in US cities. The association between shorter-term fluctuations in concentrations of particles and other air pollutants with daily mortality and morbidity is important evidence as the US Environmental Protection Agency and other bodies deliberate future air quality standards. Key constituents to these deliberations, including government, industry, non-government organizations, and the public depend upon reliable, up-to-date information about pollution-associated health effects. The data necessary for estimating these effects, including mortality and morbidity statistics, air pollution, weather, and demographic data, are routinely collected by several government agencies at enormous expense to the public. However, knowledge that is relevant to environmental policy can only be produced through systematic analysis of this information. Such analyses have been performed and reported sporadically- for example, when individual investigators conducted studies or when HEI initiated the National Morbidity and Mortality Air Pollution Study (NMMAPS) which forms the basis for this proposal.
This grant proposes to create a web-based system that regularly accesses, analyzes and disseminates policy-relevant data about the association of air pollution and mortality and morbidity in US cities. In this Phase I of the project, the requisite public information and results of statistical analyses will be posted to an iHAPSS website so that all constituents can stay informed about the most recent knowledge and its component parts. In Phase to be considered under a separate proposal, a web-based methodology for easily conducting new, innovative analyses will be created.
Goto the iHAPSS webpage.
Air Pollution and Health: A European America Approach (APHENA)
Multi-city studies of air pollution and daily mortality and hospital admissions have now been carried out in Europe, the US, and Canada. They the provide strong evidence that exposure to particulate air pollution increases rates of mortality and morbidity from cardiovascular and respiratory disease. They also suggest that these effects vary in magnitude across Europe and North America.
As an appropriate follow-up to these multi-city studies APHENA will bring together the investigators who have carried out the European Air Pollution and Health: A European Approach (APHEA 1 and 2) studies, the US National Morbidity and Mortality Study (NMMAPS) in the United States, and several national studies in Canada.
The overall goal is to produce analyses that characterize effects of air pollution on mortality and morbidity in Europe and North America, using a common analytic framework, in order to describe and explain spatial variation in the health effects of air pollution. The goal will be accomplished by initial methodologic work, followed by a joint and parallel analysis of the air pollution and health data. The methodologic research is intended to: 1) establish the comparability of methods used by the investigative groups; 2)develop and apply analytic methods for characterizing heterogeneity of air pollution effects across locations, and 3) explore the degree of mortality displacement ("harvesting").
These methods will then be applied to already developed databases on mortality and hospitalization from the APHEA1 and 2 studies, from the NMMAPS, and from the Canadian Studies. In total, the databases cover many of the major time-series studies from developed countries that have been reported over the last decade.
Hierarchical Models for Health Service and Research
The data structure and inferential goals for Health Services Research (HSR) and evaluation require use of a hierarchical model (HM) that accounts for the structure and specifies both population values and random effects for units such as clinics, physicians and health service regions. HMs properly account for the sample design and structure scientific and policy-relevant statistical inferences. Applications require valid and efficient estimation of population parameters (such as the average death rate), estimation of between unit variation (variance components) and inferences on unit-specific random effects. These include unit-specific ranks (to be used in profiling/league tables) estimates of the histogram of random unit-specific effects, estimation of how many and which unit-specific effects exceed a threshold, and identification of extremely poor and good performers.
No single set of estimates can effectively address these multiple goals, and we will develop and evaluate a panel of goal-specific summaries and inferences. The panel will combine efficiency (making good use of the available information) with robustness (high efficiency over a broad range of underlying assumptions). Our specific aims are: structuring inferences via the Bayesian formalism; development, implementation, and application of robust priors; evaluation of the robustness, efficiency and operating characteristics of the panel of summaries; development and implementation of computational approaches, with focus on massive data sets; preparation of case studies based on RAND projects; hosting a summer intern at RAND.
The proposed research will produce innovative statistical methodology and computer implementation, provide guidance on addressing multiple HSR goals, communicate interesting and informative case studies, educate and acculturate pre-doctoral students.