Helping CDC Provide Timely H1N1 and Other Public-Health Estimates
Abt Associates statisticians are assisting CDC in providing timely estimates of the prevalence of H1N1 virus symptoms and other health conditions at the national and state level. Data on these topics are collected through the Behavioral Risk Factor Surveillance System (BRFSS), a major CDC surveillance survey conducted since 1984. To maintain the viability of the BRFSS as a state-of-the-art surveillance system, Abt Associates will provide support and expertise to the CDC and to CDC grantees in two main task areas: sampling and weighting, and consultation and technical assistance on analytical issues. The BRFSS is the largest telephone health survey in the U.S. It is an important resource for monitoring state and sub-state risk factor levels and also the prevalence of several health conditions. To maintain the viability of the BRFSS as a state-of-the-art surveillance system, Abt Associates will provide support and expertise to the CDC and to CDC grantees in two main task areas: sampling and weighting, and consultation and technical assistance on analytical issues. Because the BRFSS is conducted by telephone, CDC is able to release the survey’s annual estimates only a few months after the end of the year — which is a considerably faster turnaround than in any other risk factor surveillance system. An important part of the rapid data release process is the timely calculation of the survey’s sample weights. Abt Associates statistical staff are helping CDC maintain a weighting system that is state-of-the-art in terms of both the socio-demographic variables the survey uses and the source of the BRFSS population weighting targets. They are assisting the agency in developing estimates for shorter time periods than the usual annual estimates — quarterly, monthly, and every four weeks based on the use of rolling samples. (A four-week rolling sample would involve, for instance, data collection in weeks 2-4 of September and week 1 of October, weeks 3-4 of September and weeks 1-2 of October, etc.) This rapid data delivery is of particular interest for the survey’s H1N1 question module. Implemented in September 2009, this module monitors reports of influenza-like illness (ILI) symptoms. Some states also implement a child module, and Abt Associates statisticians have developed an estimation procedure to provide four-week rolling sample ILI child estimates to the CDC. The BRFSS is primarily a landline telephone survey, but it currently includes a sample of cell phone-only adults. It may evolve into a full dual-frame survey, with one of its sampling frames being landlines and the other comprising all cell phones. Abt Associates statisticians have developed an estimation procedure that allows these samples to be combined so that estimates for the entire adult population of the U.S. can be provided to the CDC. Other planned innovations that the Abt Associates statisticians may support include:
- Creating a small area estimation system to provide estimates for all counties in the U.S.
- Including cell-only adults who do not live in households (i.e., the group quarters population).