To introduce a statistical inference framework for policy decision making on access to pediatric dental care.
Secondary data were collected for the state of Colorado for year 2019.
The access model was an optimization model, matching the demand (patients) and supply (providers) of dental care. Sampling distributions of model inputs were specified using hierarchical Bayesian models, with hyperparameters informed by prior information derived from multiple data sources. Simultaneous inference was applied to identify areas for access improvement. The model was applied to make inference on the pediatric dental care in Colorado, accounting for financial access, differentiated into public (Medicaid and CHIP), private (commercial and out‐of‐pocket), and without financial access.
Multiple data sources informed the access measurement approach including: 2017 American Community Survey, 2019 Colorado Dental Board, and 2019 National Provider Plan and Enumeration System, 2019 InsureKidsNow.gov among others.
The median access measure (travel distance) was greater than the Colorado access standards in 16.9% and 65.1% of census tracts for children with private financial access and publicly insured, respectively. Accounting for uncertainty (confidence level 99%), these percentages decreased to 14.6% and 25.6%, respectively, with mostly suburban and rural tracts failing to meet the standards. The median disparity for Medicaid and CHIP versus private financial access was greater than 5 miles in 84.5% and 81.6% of census tracts, respectively. Accounting for uncertainty (confidence level 99%), these percentages declined to 19.5% and 10.5%, respectively, with significant disparities around the metropolitan areas.
While many communities failed to meet access standards, when accounting for uncertainty, most urban ones did not fail. Disparities in spatial access between publicly and privately insured were most acute in urban communities. Medicaid insured experienced higher disparities than CHIP insured; those differences were not identified when not accounting for uncertainty.
Data Collection/Extraction Methods