Highlights

Since 2008, SSRS has conducted the Massachusetts Health Insurance Survey (MHIS), and currently partners with the Center for Health Information and Analytics (CHIA), an agency of the Commonwealth of Massachusetts, to collect 5,000 or more interviews biannually. The MHIS is a household-based survey that collects data about health and health care across Massachusetts. Additionally, SSRS partners with CHIA to conduct relevant follow-up surveys of populations of interest from the main biennial MHIS.

Challenge

Providing representation of smaller subpopulations in Massachusetts, including specific racial and ethnic groups and the uninsured, allowing for important research about these key demographics.

Approach

SSRS uses sophisticated stratification and statistical modeling to reach groups who are less likely to respond to surveys within the address-based sample (ABS). Models predict which households are likely to contain the desired populations, and these households are oversampled as part of the overall ABS design. In addition to the ABS, SSRS dials prepaid cell phone numbers to take advantage of this sample’s higher incidence of potential respondents with lower income, lower education, lower English fluency, and higher likelihood of being Hispanic or uninsured. Additionally, in order to control the costs associated with the prepaid sample, SSRS developed and implemented an “adaptive design” approach that models the likely outcome of subsequent dial attempts based on the outcomes of the initial attempts.

Results

Overall, the statistical modeling for the ABS and prepaid cell sample dialing produced the desired outcome of a representative sample of Massachusetts residents including traditionally underrepresented groups in survey research. The ABS modeling was particularly effective in targeting low-income respondents. The inclusion of a prepaid cell sample frame also increased the number of completed interviews with lower-income and younger respondents, non-White respondents, and those who are uninsured. The use of adaptive design allowed unproductive prepaid sample to be retired and new sample to be dialed in its place, as needed.