Enhancing recruitment methodologies: leveraging the Tailored Design Method to survey populations with varied engagement in healthcare

BACKGROUND: Successfully recruiting respondents to complete surveys is integral for ensuring representative samples and providing adequate power for analyses. However, obtaining survey responses for public health and health services research can be challenging, especially when working with populations with varied engagement with the providers of the institution conducting the study. Recruitment may be particularly challenging when selecting potential participants based on historical medical encounters that they may not remember, such as a brief health screen. Many researchers implement the Tailored Design Method (TDM) to optimize response rates. Adaptations to TDM, however, may be needed for researchers in federal institutions conducting surveys of populations with varied levels of healthcare engagement. This manuscript describes TDM implementation techniques and modifications that comply with federal regulations, and factors associated with response rates and survey response type. METHODS: We surveyed a random, stratified sample of 9,775 U.S. Iraq and Afghanistan war veterans whom the Veterans Health Administration (VHA) had screened between 2007 and 2018 for post-deployment health concerns. We implemented a modified TDM that complied with federal regulations to survey veterans by email and USPS between July 2020 and August 2021. RESULTS: Twenty-one percent (N = 2,025) responded. It took a median of 21 and 69 days to receive responses via electronic survey and paper booklets, respectively. The random sample of veterans had varied levels of engagement with the VHA at the time of health screening and survey. Respondents and non-respondents used equal amounts of neurology, physical medicine and rehabilitation, and mental health VHA services in the 6 months post-screens. At the time of the survey (an average of 9 years post-screen), 21% of respondents reported not having recently used VHA services. Respondents were more likely to be older, White, and female than non-respondents, although differences were small in magnitude. Compared to electronic survey respondents, paper booklet respondents were more likely to be older, rural, and screened earlier, and less likely to live in the same state as the research team. When controlling for first contact mode (email vs. USPS), only age predicted whether veterans responded by paper booklet or electronic survey. CONCLUSIONS: Investigators working in federal organizations should be aware of possible lower response rates when compared to surveys of civilians. Researchers in federal institutions seeking to optimize response rates may find our TDM adaptations useful when surveying historical cohorts or populations without regular interaction with the research institution. Our response rate (~ 21%) may be useful for researchers conducting sample size calculations with similar populations.