Identifying Hospitals in Nepal for Acute Burn Care and Stabilization Capacity Development: Location-Allocation Modeling for Strategic Service Delivery

Abstract

In Nepal, preventable death and disability from burn injuries are common due to poor population-level spatial access to organized burn care. Most severe burns are referred to a single facility nationwide, often after suboptimal burn stabilization and/or significant care delay. Therefore, we aimed to identify existing first-level hospitals within Nepal that would optimize population-level access as "burn stabilization points" if their acute burn care capabilities are strengthened. A location-allocation model was created using designated first-level candidate hospitals, a population density grid for Nepal, and road network/travel speed data. Six models (A-F) were developed using cost-distance and network analyses in ArcGIS to identify the three vs five candidate hospitals at ≤2, 6, and 12 travel-hour thresholds that would optimize population-level spatial access. The baseline model demonstrated that currently 20.3% of the national population has access to organized burn care within 2 hours of travel, 37.2% within 6 travel-hours, and 72.6% within 12 travel-hours. If acute burn stabilization capabilities were strengthened, models A to C of three chosen hospitals would increase population-level burn care access to 45.2, 89.4, and 99.8% of the national population at ≤2, 6, and 12 travel-hours, respectively. In models D to F, five chosen hospitals would bring access to 53.4, 95.0, and 99.9% of the national population at ≤2, 6, and 12 travel-hours, respectively. These models demonstrate developing capabilities in three to five hospitals can provide population-level spatial access to acute burn care for most of Nepal's population. Organized efforts to increase burn stabilization points are feasible and imperative to reduce the rates of preventable burn-related death and disability country-wide.

Figure 1.

Models of baseline and theoretical chosen hospitals (models A–F) from location-allocation modeling.