Case Control Studies after RDD: New Sample Designs and Analysis Strategies

• Activity Number: 382
• Type: Contributed
• Date/Time: Wednesday, August 11, 2004 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Ralph DiGaetano*+ and Barry I. Graubard and Sowmya R. Rao and Jacqueline Severynse and Sholom Wacholder
• Companies: Westat and National Cancer Institute and National Cancer Institute and Westat and National Institutes of Health
• Keywords: case-control ; sample frame ; oversampling ; design effects

Abstract

Sample selection of population-based controls for case-control studies is often done using an RDD screener to obtain stratification variables such as age, race, and sex, and then subsampling among study eligible persons within strata. The sample distribution of controls is commonly matched to the expected sample case distribution on these variables. With RDD response rates dropping over the past decade, alternative sample designs are being explored. Department of Motor Vehicle lists could serve as sample frames for controls, but coverage issues and the general absence of race data have posed concerns. Using geocoding to link addresses to Census areas permits the oversampling of high-density black areas, helping to increase black sample sizes efficiently. This paper examines available results from a study using such an approach, increasing targeted sample sizes to compensate for design effects so that precision and power are not adversely affected. Coverage of the general population seems fine, but issues related to coverage of subgroups by DMV lists as well as logistical concerns remain. Results on the sample distribution by race and "area density" strata are provided.