Diagnostic, pharmacy-based, and self-reported health measures in risk equalization models

Area of Science:

• Health economics
• Risk equalization modeling
• Predictive analytics in healthcare

Background:

• Existing risk equalization models often exclude comprehensive self-reported health measures.
• Current models may use limited medical diagnoses or pharmacy data for risk adjustment.
• Research is often confined to specific high-risk populations, not the general populace.

Purpose of the Study:

• To assess the predictive value of all self-reported health measures for healthcare expenditure modeling.
• To evaluate these measures in a general population of Dutch sickness fund enrollees.
• To determine their contribution beyond existing diagnostic and pharmacy data in risk equalization.

Main Methods:

• Utilized 4 models to evaluate total, inpatient, and outpatient expenditures in 2002.
• Compared Pharmacy-based Cost Groups (PCGs) and Diagnosis-based Cost Groups (DCGs) with self-reported health data.
• Assessed model performance using R2 and mean absolute prediction error, analyzing expenditure discrepancies in subgroups.

Main Results:

• Models including PCGs, DCGs, and self-reported health achieved out-of-sample R2 of 17.2% (total), 2.6% (inpatient), and 32.4% (outpatient).
• Self-reported measures were less predictive than PCGs/DCGs for high-expenditure groups but superior for low-expenditure and specific unhealthy subgroups.
• The SF-36 instrument demonstrated the greatest predictive contribution among self-reported health measures.

Conclusions:

• Self-reported health measures provide an independent contribution to forecasting healthcare expenditures.
• These measures enhance prediction models even when diagnostic and pharmacy data are already incorporated.
• The findings support the inclusion of comprehensive self-reported health data in risk equalization models.