Extended-age Out-of-sample Validation of Risk Stratification Index 3.0 Models Using Commercial All-payer Claims.

Area of Science:

• Health Informatics
• Predictive Analytics
• Clinical Risk Stratification

Background:

• Previous Risk Stratification Index 3.0 (RSI 3.0) models were developed for Medicare patients (≥65 years).
• These models utilized International Classification of Diseases, Tenth Revision (ICD-10) trajectories and admission data.
• The study aimed to assess RSI 3.0 model performance in younger, healthier populations.

Purpose of the Study:

• To evaluate the predictive accuracy of existing RSI 3.0 models in younger and healthier patient cohorts.
• To determine if RSI 3.0 models are generalizable beyond the elderly Medicare population.
• To assess the models' ability to predict utilization and adverse events in diverse hospital admissions.

Main Methods:

• Analysis of All Payer Claims data for medical and surgical hospital admissions in Utah and Oregon (2017).
• Prospective application of pre-existing RSI 3.0 models to younger populations.
• Comparison of model performance against previous validation in Medicare patients.

Main Results:

• Model performance in Utah and Oregon populations was comparable or superior to Medicare validation.
• Average area under the curve (AUC) was 0.83, with R2 values of 0.93 (Utah) and 0.85 (Oregon).
• Mean sensitivity for the highest risk group ranged from 28% (Utah) to 37% (Oregon).

Conclusions:

• Risk Stratification Index 3.0 models demonstrate robust predictive performance across diverse adult hospital admissions.
• Administrative claims-based predictive modeling provides valuable individualized risk profiles at admission.
• The findings support the broad validity and utility of RSI 3.0 models for guiding patient management.