使用核密度估計分析緊急醫療事件的時空聚集

Abstract

核心密度估計方法(Kernel Density Estimation)現今廣泛的運用於熱點分析進行空間上的分佈研究。但在緊急醫療事件中，除了距離之外，發生時間也是重要的考量。在本研究中，我們提出了叫做ERAKDE-ST (Elevated Relaxed Adaptive Kernel Density Estimation-Spatio-Temporal)的演算法進行空間加上環形時間的研究，針對美國賓州蒙哥馬利縣(Montgomery County, Pennsylvania, United States)2015到2020年，所有求救電話(9-1-1-calls)中的緊急醫療事件進行分析。利用ERAKDE-ST求得的機率密度和K-means找出在時空間中的集群。對於集群中不同的緊急事件，使用費雪正確性檢定(Fisher's Exact test)和卡方檢定(χ2 test)去分析集群之間是否有明顯不同的分布。我們認為這個研究可以影響緊急救護醫療資源的分配，能夠更有效的利用資源去拯救生命。

Kernel Density Estimation (KDE) is widely used in Hotspot Analysis to research spatial patterns. But in Emergency Medical Services (EMS) incidents, the occurrence time is a critical factor, too. In this study, we proposed a method called Elevated Relaxed Adaptive Kernel Density Estimation-Spatio-Temporal (ERAKDE-ST) to analyze the spatial and circular temporal patterns of the EMS events in 9-1-1-calls from Montgomery County, Pennsylvania, United States, from 2015 to 2020. We use the probability density obtain from ERAKDE-ST and K-means to find spatial-temporal clusters. We later apply Fisher's exact test and the χ2 test to examine and analyze the EMS incidents distribution differences between clusters. We think this study might have the chance to change the allocation of EMS resources to save more lives.