Improving Case-Control Study Validity By Addressing Sampling Frame Limitations

University Of Wisconsin-Madison, Madison WI

Investigators

Linked publications, trials & patents

Abstract

DESCRIPTION (provided by applicant): In a case-control study, selection bias may result when a sampling frame that serves as the basis for recruitment does not cover the study base completely. Over the past few decades, increasing privacy demands have led to the development of opt-out mechanisms for many population-based sampling frames used in public health research. Using preliminary data from Wisconsin, we know that driver's license lists have been affected by privacy laws, and that approximately 30% of residents have chosen to opt-out. The trend towards opting-out of sampling frames and the effects it may engender are likely not limited to Wisconsin. AIMS: Primary goals of this dissertation are (1) to explore the current variation across the country and trends over time in Wisconsin of driver's license sampling frame undercoverage, (2) to evaluate the determinants of driver's license sampling frame coverage, and (3) to compare methods to improve internal validity when using a sampling frame to select controls with incomplete coverage of the study base. METHODS: After a complete survey of Department of Transportations across the country, we propose to conduct data linkages and statistical analyses using ongoing studies initiated by the Cancer Epidemiology Program at the University of Wisconsin. We will construct predictive models describing the probability of an individual being included in the driver's license sampling frame by describing the study cases, ascertained from a mandated disease registry, that are more (and less) likely to reside on the driver's license sampling frame used to select controls. In addition, we propose to evaluate various methods to minimize selection bias when using sampling frames to select controls with incomplete coverage of the study base. These include: (1) exclusion of cases not found on the sampling frame used to select controls, (2) using selection probability ratios to directly adjust the observed odds ratios, and (3) using sampling frame coverage propensity scores derived using demographic, behavioral, and other health-related variables available for cases and applying those scores to controls. IMPACT: The effects of the selective absence of members of the study base from sampling frames used in population-based epidemiologic studies need to be analyzed and quantified. The proposed project will evaluate validity concerns that are frequently overlooked in case-control studies. This project is strengthened by the rich, diverse datasets at our disposal to explore the proposed aims. In general, biases due to inadequate coverage of the study base are increasing as more individuals and institutions opt-out of inclusion in sampling frames used in public health research. Results from this dissertation project will promote public health research that delivers accurate information to the public. As participation rates decline, understanding the effects of using sampling frames that do not fully enumerate the study base will be critical to establish the validity of study results. PUBLIC HEALTH RELEVANCE: Selection bias due to inadequate coverage of the study base is expected to increase as more individuals and institutions opt-out of inclusion in sampling frames commonly used in public health research. As participation rates in epidemiologic studies decline, understanding the effects of sampling frames that do not fully enumerate the study base will be critical to establish the validity of study results. Results from this dissertation project will promote public health research that delivers accurate and trusted information to the public.