Regional health care planning: a methodology to cluster facilities using community utilization patterns

Area of Science:

• Health Services Research
• Spatial Analysis
• Clustering Algorithms

Background:

• Effective community health care planning requires grouping facilities into regions with similar utilization patterns.
• Existing methodologies for creating these regions are not well-documented.
• This study addresses this gap by presenting a novel clustering approach.

Purpose of the Study:

• To develop and detail an objective, replicable, and sustainable methodology for grouping health care facilities.
• To apply this methodology to create Hospital Groups in Michigan for the Bed Need Methodology.
• To consider scientific, practical, and political factors in methodology development.

Main Methods:

• A 2-step K-means and Ward's clustering algorithm was employed to group hospitals.
• A heuristic integrating statistical cluster fit and group characteristics determined the optimal number of clusters.
• The methodology focused on similarities in patient utilization patterns and geographic location.

Main Results:

• The clustering methodology successfully identified 33 distinct Hospital Groups in Michigan.
• This was based on recent hospital utilization data.

Conclusions:

• An objective, replicable, and sustainable methodology for creating Hospital Groups was developed, even within a regulated environment.
• The methodology is transferable and suitable for grouping various facilities or areal units due to its foundation in sound clustering and health care utilization principles.