Reclassifying Primary Care: A Decision Tree Approach to Improving Workforce Estimates and Research Using Claims Data

Area of Science:

• Health Services Research
• Medical Informatics
• Workforce Studies

Background:

• Identifying primary care (PC) clinicians from medical claims data is crucial for workforce estimation but current methods lack precision and reproducibility.
• Existing specialty or activity-based metrics for PC clinician identification are insufficient for accurate workforce analysis.

Purpose of the Study:

• To develop a simple, accurate decision tree classifier using claims data to identify PC clinicians, trained on family physician (FP) survey responses.
• To apply this classifier to estimate the PC workforce composition in Virginia.

Main Methods:

• Linked Virginia All-Payer Claims Database (APCD) data (2016-2023) with American Board of Family Medicine survey responses to identify PC FPs.
• Trained a decision tree classifier using claims-derived features (e.g., diagnosis codes, place of service, patient visit frequency) to classify PC providers.
• Developed an enhanced classifier with exclusion criteria and applied both to the APCD to estimate the PC workforce, including nurse practitioners (NPs) and other clinician types.

Main Results:

• The base decision tree achieved 93% accuracy in classifying PC clinicians using three key features.
• The enhanced decision tree demonstrated at least 89% accuracy across all clinician types.
• Between 2016 and 2023, the proportion of NPs in Virginia's PC workforce grew from 17% to 32%, surpassing FPs as the most common PC provider type.

Conclusions:

• The developed decision tree classification method offers a scalable, interpretable, and accurate tool for identifying PC clinicians from claims data.
• This approach addresses limitations of previous methods and is valuable for workforce planning and health services research.
• The findings highlight a significant shift in the PC workforce composition, with NPs becoming the predominant provider type in Virginia.