(147) Which Medication Counting Method is Useful for Predicting Death and Institutionalization in Older Adults? A Population-Based Cohort Study

Background: Polypharmacy, the use of multiple medications, is associated with adverse health outcomes in older adults. Various medication counting methods exist, complicating polypharmacy analysis. Integrating polypharmacy indicators from these counting methods into prediction models could enhance risk assessment and facilitate the development of risk prediction tools, particularly in a context of chronic diseases and aging.

Objectives: To evaluate performance of 6 polypharmacy indicators based on different medication counting methods in predicting death and institutionalization in older adults, in Québec (Canada).

Methods: Design and setting: Using administrative data, we randomly sampled 110,000 older adults >65 years on March 31st, 2017, covered by the public drug insurance plan. Exclusion: individuals who were transferred to long-term care facilities or hospitalized for more than 30 days.
Main outcome measure: Death and institutionalization risk at 2 years.
Statistical analysis: Medication use in the previous year was assessed with 6 counting methods: 1-cumulative one-year count; 2-average of four quarters’ cumulative count; 3-single day count; 4-count of medications used in first and fourth quarters; 5-count weighted by duration of exposure; 6-count of uninterrupted medication use. Polypharmacy was defined as ≥5 medications. We used Cox multivariable regression to predict risk adjusting for age, sex, social and material deprivation, comorbidity, and use of health services. Predictive performance (R2), discrimination (D statistic), and calibration (integrated calibration index [ICI]) were investigated.

Results: Adjusted hazard ratios for predicting death ranged from 1.01 (95% CI [confidence interval] [0.95;1.07]) to 1.08 (CI [1.00;1.16]) and, for predicting institutionalization, from 1.19 (CI [1.09;1.30]) to 1.29 (CI [1.16;1.43]). The strongest associations were observed with counting methods #2 and #5. Adding a polypharmacy indicator (versus no indicator) maintained or improved predictive performance (no increase in R2 for death and increase of 11% for institutionalization) and calibration (no increase in ICI for death and institutionalization) across all models, but slightly enhanced discrimination (increase in D statistic of 1% and 6% for death and institutionalization respectively).

Conclusions: The medication counting methods has minimal impact on death or institutionalization prediction. Future work could explore qualitative aspects of polypharmacy, such as potentially inappropriate medications or drug drug interactions.