Validating and Improving Adjusted Clinical Group's Future Hospitalization and High-Cost Prediction Models for Dutch...

Area of Science:

• Health Services Research
• Medical Informatics
• Public Health

Background:

• Rising healthcare costs necessitate effective Population Health Management strategies.
• Risk stratification, assigning scores based on medical records, is crucial for identifying patients at risk of future high costs and utilization.
• Existing predictive models, often US-based, require validation and adaptation for diverse healthcare systems like Dutch primary care.

Purpose of the Study:

• To validate the accuracy of Adjusted Clinical Group (ACG) risk stratification models for predicting hospitalization and high healthcare costs in Dutch primary care.
• To adapt and improve US-based predictive models for the specific context of Dutch primary care data.
• To assess the statistical validity and discriminatory ability of these models.

Main Methods:

• Utilized Dutch primary healthcare data registries, training models on 95,262 patients and validating on 48,780.
• Incorporated patient data including age, sex, general practitioner visits, diagnoses (ICPC), and prescribed medications (ATC).
• Employed C-statistics for discriminatory ability assessment and calibration plots for calibration assessment.

Main Results:

• Adjusting the hospitalization model improved C-statistics from 0.69 to 0.75.
• Adjusting the high-cost model improved C-statistics from 0.78 to 0.85.
• Both adjusted models demonstrated good discrimination and calibration, indicating enhanced predictive performance.

Conclusions:

• Locally adjusted ACG prediction models show significant potential for improving risk stratification in Dutch primary care.
• Validated and adapted models can enhance the efficiency and effectiveness of Population Health Management.
• These findings support the use of tailored risk prediction tools for optimizing healthcare resource allocation.