Evaluating regional healthcare disparities using a gravity–radiation framework: A multi-resolution accessibility analysis

Access to healthcare facilities is essential and should be considered a fundamental right in modern societies. However, due to geographic characteristics, access to primary healthcare services varies significantly across regions. Evaluating healthcare accessibility is challenging in the absence of detailed mobility data and requires integrating geographic structure, population distribution, and travel behavior. This study implements a gravity–radiation framework to assess primary healthcare accessibility across the state of Florida. The model integrates OpenStreetMap-derived drivable travel distances with probabilistic destination selection to estimate patient travel patterns without relying on arbitrary distance thresholds. Analyses are conducted at multiple spatial resolutions to enhance spatial precision; however, finer resolutions substantially increase computational demands due to the growth of origin–destination matrices. To mitigate this challenge, a gravity-based filtering mechanism is applied to identify the most probable healthcare destinations, reducing OD matrix dimensionality while preserving analytical fidelity. A comprehensive sensitivity analysis evaluates alternative distance decay functions and spatial resolutions. Validation against observed patient visit data demonstrates strong predictive performance (correlation ≥0.97; R 2 ≥ 0.94) at finer scales. Benchmarking against spatial-buffer filtering and the standalone radiation model indicates that the gravity–radiation framework achieves statistically comparable accuracy while delivering substantial computational savings (approximately 92% CPU-time reduction). In contrast, restrictive fixed-distance buffers, although computationally efficient, risk misrepresenting accessibility patterns. Results further reveal pronounced disparities between urban counties and Florida's Rural Areas of Opportunity, where residents experience substantially longer travel distances. Overall, the study provides a scalable and interpretable framework for accessibility evaluation in data-constrained environments.