Measuring and maximizing coverage in the World Trade Center Health Registry
Murphy, J., Brackbill, R. M., Thalji, L., Dolan, M., Pulliam, P., & Walker, D. J. (2007). Measuring and maximizing coverage in the World Trade Center Health Registry: Tenth Biennial CDC and ATSDR Symposium on Statistical Methods. Statistics in Medicine, 26(8), 1688-1701. DOI: 10.1002/sim.2806
The World Trade Center Health Registry (WTCHR) is a database for following people who were exposed to the disaster of 11 September 2001. Hundreds of thousands of people were exposed to the immense cloud of dust and debris, the indoor dust, the fumes from persistent fires, and the mental trauma of the terrorist attacks on the WTC on 9/11. The purpose of the WTCHR is to evaluate the potential short- and long-term physical and mental health effects of the disaster. The definitions of the exposed groups are broad and defined based on an understanding of which groups had the highest exposures to the WTC disaster and its aftermath. The four exposure groups include rescue and recovery workers, residents, students and school staff, and building occupants and passersby in Lower Manhattan.
While one goal of the WTCHR was to maximize coverage overall and for each exposure group, another was to ensure equal representation within exposure groups. Because of the multiple sample types pursued, several approaches were required to determine eligibility. Estimates of the number of eligible persons in each of the exposed populations were based on the best available information including Census, entity-specific employment figures, and public and private school enrollment data, among other publicly available sources. To address issues of undercoverage and overcoverage a variety of methods were assessed or applied, including a capture-recapture analyses test of overlapping sample building list sources and automated deduplication of sample records.
Estimates of the true eligible population indicate that over 400000 unique individuals were eligible for the baseline health survey. Interviewer-administered surveys were completed with more than 71000 persons, resulting in an overall enrollment rate of approximately 17 per cent. Coverage was highest among rescue and recovery workers, followed by residents, students and school staff, and building occupants. Both the accuracy of coverage estimates and the raw number and representativeness of enrollees were maximized by our approach to coverage. In designing a registry which relies on multiple pathways and sources of data to build the sample, it is important to develop a comprehensive approach that considers all sources of error and minimizes bias that may be introduced through the methodology.