Validation of a Rule-Based ICD-10-CM Algorithm to Detect Fall Injuries in Medicare Data

Area of Science:

• Gerontology
• Health Services Research
• Epidemiology

Background:

• Diagnosis-code-based algorithms are essential for identifying fall injuries in Medicare data for research.
• Previous algorithms lacked validation against external standards, particularly in ICD-10-CM and Medicare Advantage (MA) data.

Purpose of the Study:

• To validate a diagnosis-code-based algorithm for identifying fall injuries against a reference standard.
• To assess algorithm performance in both Medicare fee-for-service (FFS) and MA data.

Main Methods:

• Linked self-reported fall injuries (reference standard) from the STRIDE trial to Medicare FFS and MA data (2015-19).
• Calculated the area under the receiver operating characteristic curve (AUC) to measure sensitivity and specificity of the algorithm.
• Varied date window sizes and stratified results by data source, trial arm, and healthcare system.

Main Results:

• The algorithm demonstrated 45% sensitivity and 99% specificity for identifying fall injuries within the same calendar month.
• The overall AUC was 0.79, with similar performance across FFS and MA data sources and trial arms.
• Performance varied by participating healthcare system, with AUCs ranging from 0.71 to 0.84.

Conclusions:

• An ICD-10-CM algorithm shows acceptable performance for identifying fall injuries in both Medicare FFS and MA data.
• This validated algorithm can reliably ascertain fall injury outcomes in observational and interventional studies using Medicare claims data.