How well do principal diagnosis classifications predict disability 12 months postinjury?

Area of Science:

• Public Health
• Epidemiology
• Health Services Research

Background:

• Disability weights are crucial for calculating disability-adjusted life years (DALYs).
• Existing injury classifications lack empirical data for homogeneity in disability outcomes.
• This limits informed recommendations for measuring injury burden.

Purpose of the Study:

• To validate and improve injury burden estimates using empirical data.
• To assess the predictive capacity of injury groupings and ICD-10 diagnoses for disability outcomes.
• To inform best practices in measuring injury burden.

Main Methods:

• Pooled data from over 30,000 injured participants across six cohort studies (Injury-VIBES Study).
• Mapped International Classification of Disease 10th Revision (ICD-10) codes to existing injury burden groupings.
• Developed prediction models to assess the capacity of groupings and ICD-10 codes to predict 12-month disability outcomes, adjusted for age, gender, and data source. Model performance evaluated using AUC and Hosmer-Lemeshow statistics.

Main Results:

• Model performance varied by outcome; detailed ICD-10 codes showed highest discrimination (AUC 0.627-0.847).
• Prediction was best for functional outcomes like independent living, mobility, and self-care.
• Classifications were poorer predictors for anxiety/depression and pain/discomfort.

Conclusions:

• Detailed ICD-10 codes offer better prediction of functional disability outcomes compared to broader groupings.
• No single classification system demonstrated clear superiority across all disability measures.
• Findings highlight the need for empirically-derived, homogeneous injury groups for accurate burden estimation.