Predicting Health Care Utilization After Behavioral Health Referral Using Natural Language Processing and Machine...

Area of Science:

• Health Services Research
• Clinical Informatics
• Mental Health Services

Background:

• Mental health conditions are significant drivers of increased healthcare utilization.
• Accurate prediction of future healthcare needs is crucial for effective resource allocation and patient management.

Purpose of the Study:

• To develop and validate machine learning models for predicting healthcare utilization following initial behavioral health encounters.
• To identify patients at risk for significantly decreased or ultra-high future healthcare use.

Main Methods:

• Utilized random forest classification algorithms.
• Trained models on data from patients' first behavioral health encounters.
• Evaluated model performance using Area Under the Receiver Operating Characteristic Curve (AUROC).

Main Results:

• Achieved an AUROC of 0.74 for predicting any decrease in utilization.
• Achieved a higher AUROC of 0.88 for predicting ultra-high absolute utilization.
• Demonstrated the predictive power of machine learning in this context.

Conclusions:

• Random forest models show strong potential for predicting healthcare utilization post-behavioral health encounters.
• These predictive models can support clinical decision-making, patient referrals, and healthcare cost management.
• The findings highlight the utility of machine learning for risk stratification in mental health services.