A Rational Agent Model for the Spatial Accessibility of Primary Health Care

Abstract

Accurate modeling of the spatial accessibility of health care is critical to measuring and responding to physician shortages. We develop a new model in which patients choose the primary care location that minimizes their combined accessibility and availability costs. This model offers several advantages with respect to existing access frameworks. It incorporates feedback between patient decisions and endogenizes the trade-off between travel times and congestion at the point of care. It allows for patients to seek care from their home or workplace and can account for multiple travel modes. Our open-sourced implementation scales efficiently to large areas and fine spatial granularity. Using distributed computing, we calculate travel times for this model at the census tract level for the entire United States, and we also make this resource available. We compare the results to those from existing primary care access models. Key Words: floating catchment areas, primary health care, rational agent models, spatial accessibility.

健康照护空间可及性的准确模式化，是评估并因应医师短缺的关键。我们发展一个崭新的模型，其中病患选择可最小化可及性与可取得花费的初级医疗照护地点。此一模型相对于既有的可及性架构而言提供若干优势。它纳入病患决策间的反馈，并内化旅行时间和照护地点壅塞之间的权衡。该方法考量病患从家中或工作场所寻求照护，并且能够说明多旅次模式。我们的开源执行，有效地比例化至大面积与细緻的空间颗粒度。我们运用分布式计算，估计在全美国统计区块层级此一模型的旅行时间，并且提供该资源。我们将研究结果与既有的初级医疗取得模式进行比较。关键词：移动搜索面积，初级医疗照护，理性行动者模型，空间可及性。

Modelar con rigurosidad la accesibilidad espacial de la asistencia médica es crucial para medir y enfrentar la escasez de médicos. Nosotros desarrollamos un nuevo modelo en el que los pacientes escogen la localización de la atención primaria, lo cual minimiza los costos combinados de accesibilidad y disponibilidad que les correspondan. Este modelo tiene varias ventajas con relación a los marcos de acceso existentes. Incorpora la retroalimentación entre las decisiones del paciente y promueve compensación endógena entre los tiempos del viaje y la congestión en el punto de la atención. Permite a los pacientes buscar asistencia respecto de su hogar o del lugar de trabajo, y puede ajustarse a los modos de viaje múltiple. Nuestra implementación de fuente abierta acomoda la escala para áreas grandes y granularidad espacial fina. Usando computación distribuida, calculamos los tiempos de viaje para este modelo a nivel de tractos censales para el conjunto total de los Estados Unidos, y también facilitamos la disponibilidad de este recurso. Comparamos resultados con los que se obtienen mediante los modelos existentes de acceso primario a la asistencia médica. Palabras clave: accesibilidad espacial, áreas de captación flotante, asistencia médica primaria, modelos de agente racional.

Supplemental Material

Supplemental data for this article can be accessed on the publisher’s website.

Notes

Notes

1 We base this statement on the total number of individual origin locations considered and the complexity of the road network. We consider every census tract in the United States—upwards of 70,000 cells. Others have preceded us with finer granularity or national scale. For example, Li, Serban, and Swann (2015) performed a county-level calculation for the continental United States. McGrail and Humphreys (2015) calculated access to primary care in Australia, using cells with just a few hundred people in rural areas (U.S. census tracts have a few thousand). The Australian road network is far less complex, though, and the total number of origins is less than a third of ours.

2 Strictly speaking, this is true only if all doctors see at least one patient.

3 Congestion might indicate higher quality or more efficient service or amenities like foreign language comprehensions, but these go beyond the scope of this article. In itself the congestion is a purely “bad.”

4 Alternative initialization strategies yield consistent results.

Additional information

Notes on contributors

James Saxon

JAMES SAXON is a Postdoctoral Fellow in the Center for Spatial Data Science and the Harris School of Public Policy at the University of Chicago, Chicago IL 60637. E-mail: [email protected]. He uses large data sets and graph theory to measure and characterize the accessibility and use of public and human resources.

Daniel Snow

DANIEL SNOW is a Research Assistant in the Center for Spatial Data Science and the Harris School of Public Policy at the University of Chicago, Chicago IL 60637. E-mail: [email protected]. His research interests include the economics of health care, social network theory, and the spatial accessibility of transportation and primary care.