Detecting Uncoded Self-Harm in Veterans' Electronic Health Records Using Positive and Unlabeled Learning:...

Area of Science:

• Public Health
• Data Science
• Mental Health Research

Background:

• Suicide and self-harm are critical public health issues in the U.S.
• Undercoding of mental health conditions in electronic health records (EHRs) leads to missing data.
• Lack of reliable negative examples hinders accurate prevalence estimation and identification of at-risk individuals.

Purpose of the Study:

• To identify U.S. Veterans with self-harm events not captured by diagnostic codes in EHRs.
• To estimate the true prevalence of self-harm among Veterans using a novel Positive and Unlabeled (PU) learning algorithm.
• To address undetected mental health diagnoses within large healthcare datasets.

Main Methods:

• Retrospective observational study of 1.3 million Veterans Health Administration EHRs (1999-2019).
• Application of the PULSNAR (Positive Unlabeled Learning Selected Not At Random) algorithm to estimate uncoded self-harm.
• Independent chart reviews of 97 uncoded Veterans to validate algorithm predictions and refine prevalence estimates.

Main Results:

• Only 1.85% of Veterans had coded self-harm, while PULSNAR estimated an overall prevalence of 10.46%.
• PULSNAR identified an additional 8.77% of self-harm cases among the uncoded population.
• Analysis suggests coded self-harm represents only 23.4% of all documented self-harm events in Veterans.

Conclusions:

• PULSNAR offers a scalable framework for estimating mental health condition prevalence and identifying uncoded cases without requiring negative labels.
• This method addresses diagnostic undercoding in EHRs, improving prevalence estimation and supporting targeted interventions.
• The approach can enhance clinical decision support, resource allocation, and research for better mental health outcomes.