# Behavioral Phenotypes in Electronic Health Record Use by Primary Care Providers: a Cluster Analysis

## TL;DR

This study identifies three distinct patterns of how primary care providers use electronic health records, based on their engagement levels and behaviors.

## Contribution

The paper introduces a data-driven clustering approach to uncover distinct EHR use behaviors among primary care providers.

## Key findings

- Three clusters were identified: high-engagement, low-engagement, and moderate/selective users.
- Each cluster represents different levels of EHR functionality use and time spent.
- The findings suggest potential for tailoring EHR interventions to provider workflow styles.

## Abstract

The use of electronic health record (EHR) systems varies among primary care providers (PCPs). However, little is known about how numerous different EHR use behaviors, such as time spent and collaboration in the EHR, cluster together. Prior efforts to quantify characteristics of PCPs using EHRs have generally focused on single behaviors.

To identify patterns of EHR use among PCPs using a data-driven clustering approach.

Cross-sectional study analyzing EHR data from the 2021 calendar year.

Primary care providers practicing in a large Massachusetts healthcare system.

PCPs were assigned to groups based on patterns of EHR use across 30 monthly variables from EHR data using a k-means clustering approach. We used Elbow, Silhouette, and Gap statistic methods to determine the number of clusters. Cluster characteristics were analyzed descriptively.

In total, 163 PCPs were included; 103 (63%) PCPs were female, and 113 (69%) were White. Three distinct clusters of PCPs were identified, named based on the EHR characteristics that differed most across the clusters: (1) “High-engagement users”: 38% of PCPs; (2) “Low-engagement users”: 42%; and (3) “Moderate and selective users”: 20%.

This study identified three distinct patterns of EHR use among PCPs, characterized by different levels of engagement with EHR functionality and time spent in the EHR. Further studies are needed to explore how EHR-based interventions could be tailored to different provider workflow styles.

The online version contains supplementary material available at 10.1007/s11606-025-09670-9.