Researchers develop a machine learning method to reveal self-harm history hidden in veterans' electronic medical records.
Key Details
- 1Study analyzed over 1.3 million Veterans Health Administration patients' electronic records.
- 2Diagnosis codes identified only 1.85% with self-harm history; machine learning estimates true rate at 7.9%.
- 3The PULSNAR algorithm addressed missing or incomplete coding of self-harm in standard record fields.
- 4Problem lists captured self-harm for just 22.6% of those with relevant diagnosis codes.
- 5Findings show a substantial gap in the visibility of clinically important mental health history.
- 6This method is currently a research tool and not yet ready for direct clinical deployment.
Why It Matters
Uncovering hidden self-harm history in health records can significantly improve mental health care planning and research. For radiology-AI and broader medical AI communities, these advances in handling undercoded conditions demonstrate how informatics tools can support more accurate predictive models and population health analytics.
Source
EurekAlert
Related News
•EurekAlert
NYU AI Model Improves 5-Year Breast Cancer Risk Prediction via 3D Mammograms
NYU researchers developed an AI model using longitudinal 3D mammograms that outperforms single-scan and 2D-based tools in predicting 5-year breast cancer risk.
•EurekAlert
Explainable AI Predicts High-Risk Heart Attack Patients for Imaging and Treatment
Researchers developed an explainable AI scoring system to predict intramyocardial hemorrhage risk after severe heart attacks, potentially guiding imaging decisions.
•EurekAlert
AI Framework Accelerates Aortic Aneurysm Risk Prediction from Imaging
Researchers developed BioPINN-LM, combining physics-informed neural networks and multimodal large language models to deliver fast, interpretable risk assessments for ascending thoracic aortic aneurysms.