A geographic tool to predict community deprivation in health care access: Based on a model of combined personal and ecologic characteristics

Context: Personal characteristics are often used to explain barriers or delays in accessing healthcare for individuals. Ecological measures, like poverty level, are sometimes used as near-proxies for healthcare access problems for communities and populations. We combined these methods to develop indices of healthcare access deprivation or difficulty accessing care for individuals in geographically small areas in the US. In an earlier step we reviewed previous studies, identified factors associated with poor access to care and developed parsimonious models that explain poor access to healthcare at the individual level using a national survey and logistic regression methods.

Design: We used data from the 2002 and 2003 National Health Interview Surveys and our earlier models that identified individual-level predictors of healthcare access deprivation, and constructed individual indices of access deprivation. We geo-coded the data and merged them including the indices with 2000 US Census data at census-tract level. Using multi-level logistic modeling techniques we undertook pair wise substitution of Census for NHIS variables that allowed us to determine the ecologic variables that best predict reported health access problems in the US. We constructed census-tract-level indices of access deprivation, using the best-fit model, as the expected values of the dependent variable. We mapped the resulting indices nationally and showed how they can be overlaid with health care provider availability and other statistics.

Results: 9 percent of the 67 thousand census-tracts matched to the NHIS sample and were used in the modeling. Census-tract level variables that best predict health access problems are percent Black, percent Hispanic, percent 65 years or older, percent disabled, and percent with only 1 adult in household. Our raw census-tract deprivation indices ranged from 6.3% to 40.9%. The results include derived deprivation indices for each of the 66,997 U.S. census tracts, a series of US national maps of all census-tracts indicating tracts or communities with possible health access deprivation, overlaid with provider availability, Health Professional Shortage Areas designation, etc.

Conclusions: Ecologic factors that identify small areas or populations at risk of healthcare access deprivation can be combined with or substituted for individual-level data to model community-level barriers or delays in accessing health care, and enable local community assessments and health planning.