Developing a Model to Predict High Health Care Utilization Among Patients in a New York City Safety Net System.

Area of Science:

• Health Services Research
• Predictive Analytics
• Public Health

Background:

• Traditional predictive models for health care utilization often overlook social determinants of health and may not apply to safety net populations.
• Existing models frequently rely on commercial insurance claims and focus narrowly on readmission risk.
• There is a need for payer-agnostic risk models generalizable to diverse patient populations, particularly those served by safety net hospitals.

Purpose of the Study:

• To develop and validate a payer-agnostic predictive model for high future health care utilization.
• To identify patients at risk who could benefit from targeted interventions within a large US safety net hospital system.
• To incorporate social determinants of health into risk prediction for a vulnerable patient population.

Main Methods:

• Utilized electronic health record and administrative data from 833,969 adult patients treated between July 2016 and July 2017.
• Transformed data into demographic, utilization, diagnosis, medication, and social determinant variables (e.g., homelessness, incarceration history).
• Developed and validated multiple models to predict various measures of acute health care utilization, prioritizing positive predictive value for the top 1% of utilizers.

Main Results:

• The final model, predicting the continuous number of acute days, included 17 variables.
• Achieved a positive predictive value of 47.6% and a sensitivity of 17.3% for the top 1% of high acute care utilizers.
• Previous health care utilization and psychosocial factors emerged as the most significant predictors of future high utilization.

Conclusions:

• Demonstrated a feasible method for predicting high acute care utilization in safety net hospitals.
• Successfully incorporated social risk factors into predictive modeling using electronic health record data.
• The developed model offers a potential tool for improving resource allocation and patient care interventions in safety net settings.