Area-Level Sociodemographic Differences Between Indian Health Service Purchased/Referred and Non-Purchased/Referred Care Delivery Areas

Purpose: Purchased/Referred Care Delivery Area (PRCDA) counties are those where resident American Indian and Alaska Native (AIAN) people are eligible for Indian Health Service care. Due to concerns about racial misclassification, cancer statistics for AIAN people are often restricted to PRCDA counties. Differences in sociodemographic characteristics may exist between PRCDA and non-PRCDA counties, but have not been described; therefore, the potential selection bias associated with the restriction to PRCDA counties remains unknown. Methods: We used data from the University of California, San Francisco Health Atlas to explore ecological differences in county-level demographic, socioeconomic, healthcare access, and health outcomes data between PRCDA and non-PRCDA counties (n = 3152 counties). We tested for statistical differences in mean levels of demographics between PRCDA and non-PRCDA counties using Pooled or Welch t-tests. Results: We observed small, but statistically significant differences between PRCDA and non-PRCDA counties in county-level demographic and socioeconomic characteristics (age, poverty, utility services threat, unemployment, educational attainment, computer access, and median income), neighborhood and environment characteristics (overcrowding, severe mortgage/rent burden), healthcare access and utilization (uninsured, annual checkup, annual dental visit, mammography, binge drinking, smoking, physical inactivity, social isolation), and health outcomes (poor mental health, arthritis, poor self-rated health, high blood pressure, diabetes, high cholesterol, and obesity). Conclusions: These results indicate variability in county-level measures between PRCDA and non-PRCDA counties. While these data do not speak specifically to AIAN peoples’ experiences, they provide critical contextual information to understand how exclusion of AIAN people residing in non-PRCDA counties from cancer statistics may bias risk estimates.