Related Experiment Video
Updated: Jun 9, 2026
08:20
Superior Auto-Identification of Trypanosome Parasites by Using a Hybrid Deep-Learning Model
Published on: October 27, 2023
Identification of Patients for a Community Health Worker Program Using an Artificial Intelligence Algorithm
Samuel T Savitz1, Brendan Broderick1, Margaret M Paul1
1Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery, Mayo Clinic Rochester Minnesota USA.
Learning Health Systems
|June 8, 2026
Summary
An electronic health record (EHR) algorithm effectively identifies patients needing community health worker (CHW) support, including those missed by traditional screening. This tool improves access to vital resources for underserved populations.
Area of Science:
- Health Informatics
- Public Health
- Clinical Decision Support
Background:
- Community health workers (CHWs) are crucial for connecting patients with community resources, improving health outcomes and reducing healthcare use.
- Identifying eligible patients for CHW programs presents a significant challenge, often relying on incomplete screening methods.
Purpose of the Study:
- To develop and evaluate an electronic health record (EHR)-based algorithm for predicting referrals to a CHW program.
- To determine if the algorithm identifies a distinct patient subgroup compared to traditional health-related social needs (HRSN) screening questionnaires.
Main Methods:
- A gradient boosted time-to-event model was developed using EHR data from primary care patients.
- The model incorporated demographics, diagnoses, HRSN questionnaire responses, and HRSNs extracted from clinical notes via natural language processing.
- Algorithm-identified patients' characteristics were compared to those identified by questionnaires.
Main Results:
- The algorithm achieved high performance (AUC 0.87 training, 0.92 hold-out).
- Algorithm-identified patients were more likely to use interpreter services, prefer non-English languages, and have documented health literacy issues.
- A significant portion (53%) of algorithm-identified patients had not completed an HRSN questionnaire within two years; health literacy and financial strain were key predictors.
Conclusions:
- The EHR-based algorithm serves as an effective chart-review prioritization tool for CHW program referrals.
- It addresses challenges in identifying patients with HRSNs, particularly when questionnaire data is incomplete.
- This approach enhances the identification of patients who may benefit from CHW services.
