Development and Validation of Algorithms to Predict Activity, Mobility, and Memory Limitations Using Medicare Claims...

Area of Science:

• Gerontology
• Health Services Research
• Data Science in Healthcare

Background:

• Functional impairment is a significant predictor of mortality and healthcare utilization.
• Current methods for assessing functional impairment are not practical for large-scale clinical use or risk adjustment.
• There is a need for efficient, claims-based tools to identify functional limitations in older adults.

Purpose of the Study:

• To develop and validate claims-based algorithms for predicting functional impairment.
• To utilize Medicare Fee-for-Service (FFS) data linked with post-acute care (PAC) assessments.
• To identify predictors of memory and activity/mobility limitations using machine learning.

Main Methods:

• Employed supervised machine learning techniques on Medicare FFS claims data (2014-2017).
• Linked claims data with post-acute care (PAC) assessment data for validation.
• Weighted data to represent the broader Medicare FFS population.

Main Results:

• Developed an algorithm for memory limitation with moderate sensitivity and specificity.
• The algorithm for activity/mobility limitations accurately identified those with five or more limitations, but overall accuracy was low.
• The developed algorithms show potential for use within PAC populations.

Conclusions:

• Claims-based algorithms can predict certain functional impairments, particularly memory limitations, in PAC populations.
• Generalizability of these algorithms to the wider older adult population requires further investigation and validation.
• This approach offers a potential pathway for large-scale risk adjustment and intervention targeting if accuracy can be improved.