Predicting All-Cause Mortality Using Two Claims-Based Measures in Medicare Beneficiaries With Dementia.

Area of Science:

• Gerontology
• Health Services Research
• Biostatistics

Background:

• Predicting mortality in dementia patients is crucial for healthcare planning.
• Existing comorbidity indices like the Chronic Conditions Warehouse (CCW) and Elixhauser Comorbidity Index are used for risk stratification.
• The performance of these indices in predicting mortality among Medicare beneficiaries with dementia requires further comparison.

Purpose of the Study:

• To compare the predictive performance of the CCW and the 38-condition Elixhauser Comorbidity Index for all-cause mortality in community-dwelling Medicare beneficiaries with dementia.
• To evaluate the utility of elastic net logistic regression in developing parsimonious predictive models for mortality using claims data.

Main Methods:

• A national sample of 1,566,359 community-dwelling Medicare beneficiaries (age ≥65) with dementia was identified from 2018 claims data.
• Elastic net logistic regression was used to model mortality at 30 days, 60 days, 180 days, and 1 year, utilizing 30 CCW conditions and 38 Elixhauser comorbidities.
• Model performance was assessed using C-statistics for discrimination and calibration measures.

Main Results:

• Mortality rates at 1 year were 19.0%.
• Both CCW and Elixhauser measures demonstrated good discrimination (C-statistics: 0.696-0.731) and calibration.
• No significant performance differences were observed between the CCW and Elixhauser measures in predicting all-cause mortality.
• Elastic net models yielded parsimonious predictors with performance comparable to traditional logistic regression.

Conclusions:

• The Chronic Conditions Warehouse and Elixhauser Comorbidity Index exhibit similar accuracy in predicting all-cause mortality among Medicare beneficiaries with dementia.
• Elastic net logistic regression provides a robust and efficient approach for developing claims-based mortality prediction models in this population.