Constructing service areas is an important task for evaluating geographic variations of health care markets. This study uses cancer care as an example to illustrate the methodology, with the nine‐state Northeast Region of the USA as the study area. Two recent algorithms of network community detection are implemented to account for additional constraints such as spatial connectivity and threshold region size. The refined methods are termed “spatially constrained Louvain” (ScLouvain) and “spatially constrained Leiden” (ScLeiden) algorithms, corresponding to their predecessors the Louvain and Leiden algorithms, respectively. Both are network optimization methods that maximize flows within delineated communities while minimizing inter‐community flows. The service areas derived by the methods, termed “cancer service areas”, are more favorable than the commonly used comparable unit, hospital referral regions, for evaluating cancer‐specific variations in care. Between the two, the ScLeiden performs better than the ScLouvain in terms of modularity, localization index and computational efficiency, and thus is recommended as an effective and efficient approach for defining functional regions.

中文翻译：

---

划定医疗保健服务区域的网络优化方法：空间约束 Louvain 和 Leiden 算法

构建服务区是评估医疗保健市场地域差异的一项重要任务。本研究以癌症护理为例来说明该方法，以美国东北部九州地区为研究区域。实施了两种最近的网络社区检测算法，以解决额外的约束，例如空间连通性和阈值区域大小。改进后的方法被称为“空间约束 Louvain”（ScLouvain）和“空间约束 Leiden”（ScLeiden）算法，分别对应于它们的前辈 Louvain 和 Leiden 算法。两者都是网络优化方法，最大限度地提高划定社区内的流量，同时最小化社区间的流量。通过这些方法得出的服务区，称为“癌症服务区”，比常用的可比单位、医院转诊区更适合评估癌症特定的护理差异。在两者之间，ScLeiden 在模块化、定位指数和计算效率方面的表现优于 ScLouvain，因此被推荐为一种有效且高效的定义功能区域的方法。