Using disaster‐induced closures to evaluate discrete choice models of hospital demand
利用自然灾害导致医院意外关闭的事件，比较实际患者分流比例与标准离散选择模型的预测，发现模型系统性低估了大规模分流，并探讨了改进方法。
Abstract Although diversion ratios are important inputs to merger evaluation, there is little evidence about how accurately discrete choice models predict diversions. Using a series of natural disasters that unexpectedly closed hospitals, we compare observed post‐disaster diversion ratios to those predicted from pre‐disaster data using standard models of hospital demand. We find that all standard models consistently underpredict large diversions. Both unobserved heterogeneity in preferences over travel and post‐disaster changes to physician practice patterns can explain some of the underprediction of large diversions. We find a significant improvement using models with a random coefficient on distance.