Deep Network Capacitated Covering Location Model: Spatial Location-allocation Optimization of Community Healthcare Facilities in Consideration of Public Health Emergencies

Equitable and efficient healthcare service is a critical issue for policymakers and planners in public health emergency events. However, current optimization methods of spatial location-allocation for healthcare facilities at community level often overlook the potential requirements stemming from outbreaks of infectious disease, which lead to biases in facility layout. This study proposes a deep network capacitated covering location model to optimize the spatial location-allocation of community healthcare facilities by considering the site suitability of facilities with an emphasis on emergency preparedness and residents’ access on foot. In the experimental analysis in Panyu district, Guangzhou, we establish the site suitability and access evaluation criteria to build a feature cube comprising 18 criteria maps. By labeling positive and negative samples for the deep site evaluation network model training, we input the feature cube to the model to identify suitable candidate sites. Then we evaluate the healthcare facility distribution against residents’ demands to optimize the spatial layout of healthcare facilities considering capacity constraints. The results indicate significant spatial disparities in community healthcare facility access, with the current distribution failing to meet demands. According to our approach, 637 suitable site locations as candidate sites are classified. Moreover, we find that only 51 additional facilities are needed to extend coverage to 91.2% of the population within a 30-minute walk. The proposed model outperforms the analytic hierarchy process by more accurately addressing residents’ actual healthcare needs. From a policy perspective, optimizing healthcare facility location-allocation using the proposed method improves equity and efficiency for residents at a walking scale, while maintaining emergency preparedness.