Precision Health-Enabled Machine Learning to Identify Need for Wraparound Social Services Using Patient- and...

Area of Science:

• Precision health
• Machine learning applications in healthcare
• Social determinants of health

Background:

• Social factors significantly impact health outcomes but are often unaddressed in medical care.
• Limited expertise exists in using patient- and population-level data with machine learning to predict social service needs.
• Opportunities to improve health, reduce costs, and enhance care coordination are missed due to these gaps.

Purpose of the Study:

• To develop decision models for identifying patients needing wraparound social services.
• To leverage clinical, behavioral, social risk, and social determinants of health data.
• To predict the need for behavioral health, dietitian, social work, and other social services.

Main Methods:

• Utilized comprehensive patient- and population-level datasets to build predictive models.
• Employed machine learning techniques within a safety-net health system.
• Evaluated model performance using metrics like AUROC, sensitivity, precision, F1 score, and specificity.
• Assessed the impact of social determinants of health data on model performance.

Main Results:

• Developed decision models with performance measures ranging from 59.2% to 99.3%.
• Achieved statistically superior results compared to previous models using limited data (38.2% to 88.3%).
• Inclusion of additional population-level social determinants of health data did not improve model performance.

Conclusions:

• Precision health decision models utilizing extensive data and machine learning can effectively predict the need for social services.
• These models demonstrate good performance in identifying vulnerable patients requiring support.
• Further research may explore optimal integration of diverse data sources for enhanced prediction.