Understanding the Clustering Patterns in Physician Distribution Through Affinity Propagation.

Area of Science:

• Public Health
• Geospatial Analysis
• Health Services Research

Background:

• Physician spatial distribution is crucial for public health research.
• Distinguishing between home and practice addresses is key to understanding maldistribution, accessibility, and disparity.
• Current data often lacks clarity on physician address types, impacting spatial accessibility research.

Purpose of the Study:

• To develop and apply methods for differentiating physician home and practice addresses.
• To explore clustering patterns in physician distribution using Affinity Propagation (AP).
• To derive potential practice location extents for physicians providing home addresses.

Main Methods:

• Utilized the 2014 American Medical Association (AMA) Physician Masterfile for physician data.
• Applied the Affinity Propagation (AP) clustering algorithm to physician location data in Fulton and DeKalb counties, Atlanta, Georgia.
• Incorporated both Euclidean and driving distances, and gravity models within the AP calculations.

Main Results:

• Successfully applied AP clustering to identify spatial patterns in physician distribution.
• Demonstrated the ability to infer potential practice locations using home address data.
• Hierarchical clustering patterns were derived by adjusting AP parameters.

Conclusions:

• The study provides a novel approach to address physician distribution uncertainty in spatial accessibility research.
• Affinity Propagation is a viable method for understanding physician clustering and inferring practice locations.
• This pilot study has broader implications for public health research and policy.