Evaluation of the HealthImpact Diabetes Risk Model in the Veterans Health Administration

Area of Science:

• Health services research
• Epidemiology
• Biostatistics

Background:

• HealthImpact is a novel algorithm utilizing administrative healthcare data for diabetes risk stratification.
• Assessing its predictive validity and utility in diabetes screening is crucial for public health.

Purpose of the Study:

• To independently validate the predictive accuracy of the HealthImpact algorithm.
• To evaluate the practical utility of HealthImpact for diabetes screening in a national healthcare system.

Main Methods:

• Utilized National Veterans Health Administration data to form two cohorts for analysis.
• Replication cohort: assessed diabetes incidence over 3 years based on HealthImpact scores.
• Utility cohort: examined diabetes screening rates over 2 years, stratified by HealthImpact scores.

Main Results:

• The 3-year diabetes incidence was 9.1% in the replication cohort (n=3,287,240).
• A HealthImpact score > 90 had a positive predictive value of 29.8% for incident diabetes.
• Screening rates were high (>98%) for patients with HealthImpact scores > 90 in the utility cohort.

Conclusions:

• Independent analysis confirms HealthImpact's validity in stratifying diabetes risk.
• The algorithm's practical utility for enhancing diabetes screening is contingent on existing screening practices and their frequency.