Assessing Urgent Medical Accessibility in Monocentric Megacities: A Bayesian-Optimized GP2SFCA Study of Chengdu

Evaluating urban residents'access to emergency medical care by car is crucial for optimizing public health resources in megacities. To address the challenges of subjective parameter settings and the integration of behavioral preferences with real-time traffic data, this study introduces an integrated analytical framework that combines "behavioral selection, real-time traffic, and parameter optimization." Expanding on the traditional Two-Step Floating Catchment Area (2SFCA) method, this framework incorporates multi-period and multi-threshold real-time travel data. It develops a multi-constraint optimization model that includes spatial equity, Bayesian priors, and operational logic to objectively calibrate the distance decay coefficient (λ), treating it as a variable to be solved rather than preset. The model assesses Chengdu residents' access to Grade-A Tier-3 hospitals during emergencies, simulating hospital choice as a probability-based decision informed by predicted travel times. The analysis uncovers a complex spatiotemporal structure that surpasses a simple core-periphery division. At the macro level, a three-tier zonal differentiation is evident, while at the micro level, a patchy heterogeneity is influenced by fine-grained road networks and polarized medical resources, particularly under 5-minute thresholds. The study also highlights a systematic “dual vulnerability”: regular weekday vulnerability caused by tidal traffic flows, and complex weekend vulnerability due to a spatiotemporal displacement effect from leisure travel, which unexpectedly decreases accessibility even in central urban areas. These patterns reveal deeper structural issues of medical resource “functional lock-in” and “configuration rigidity” within a monocentric city model. Parameter calibration reveals consistent patterns across scenarios: λ increases with shorter emergency time thresholds and during peak hours, mathematically supporting the hypothesis that time pressure heightens sensitivity to distance. This finding, along with the model's robustness confirmed by extensive sensitivity analysis (93.1% of scenarios show minimal fluctuation), ensures that the results are reliable and not artifacts of parameter subjectivity. Theoretically, this research extends the 2SFCA method to dynamic urban settings and offers empirical insight into how human spatial behavior interacts with real-time traffic. Practically, the proposed Gaussian Probability Two-Step Floating Catchment Area (GP2SFCA) model and framework support the optimization of Chengdu’s emergency “life channels” and medical resource allocation, providing a reusable toolkit for similar monocentric megacities.