State and regional estimates using seven cycles of pooled nationally representative HINTS data

Abstract

BackgroundThe Health Information National Trends Survey (HINTS) is a probability-based, nationally representative survey conducted routinely to gather information about the American public's cancer-related beliefs and behaviors, including the use of cancer-related information. HINTS was created to produce national estimates and has lacked the ability to create accurate and precise state and regional estimates. The motivation for this current work was to create state- and regional-level estimates using a national sample (HINTS) through standard calibration methods. Health estimates at a local level can inform policy decisions that better target the cancer needs within a community. Local-level data allow researchers an opportunity to examine local populations in finer detail without additional costly data collection. MethodsBy combining seven cycles of HINTS data from 2012 to 2018 and then raking the previously created person-level weights, we were able to create tables and maps of HINTS subnational survey estimates for key outcomes that have small variances and little potential bias. Results and conclusionThis paper describes the methods used to harmonize and aggregate data across cycles, create state- and regional-level estimates from the pooled data, and produce survey weights for the pooled datasets. It demonstrates both the opportunities and the challenges of pooled data analysis.