Development of a Predictive Algorithm to Identify Adults With Mobility Limitations Using VA Health Care...

Area of Science:

• Health Services Research
• Biostatistics
• Gerontology

Background:

• Millions of Americans experience mobility limitations, necessitating research into their healthcare needs.
• Existing health care administrative data offer potential for studying this population, but lack validated identification methods.
• Accurate cohort identification is crucial for research on individuals with mobility limitations.

Purpose of the Study:

• To develop and validate an algorithm for reliably identifying adults with mobility limitations within U.S. Department of Veterans Affairs (VA) health care data.
• To assess the performance of logistic regression models in classifying mobility limitation status.

Main Methods:

• Linked VA administrative data (diagnosis, encounters, durable medical equipment, demographics) for 964 veterans.
• Validated against self-reported mobility limitation data from the Medicare Current Beneficiary Survey (MCBS).
• Evaluated logistic regression models for binary (yes/no) and multilevel mobility limitation classification.

Main Results:

• The binary logistic regression model demonstrated good performance, with 70% sensitivity and 79% specificity in identifying mobility limitation.
• The algorithm effectively discriminated between veterans with and without mobility limitations.
• Multilevel approaches for predicting mobility limitation did not perform well in this study.

Conclusions:

• The developed algorithm reliably identifies adults with a binary mobility limitation status in VA health care data.
• This method can facilitate research on veterans with mobility limitations.
• Future research should explore multilevel prediction models and include more diverse populations (women, younger adults).