Senior Researcher Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, United Kingdom
Background: Polypharmacy has been defined using cross-sectional thresholds, like for example the use of 10 or more ingredients in a given year. Longitudinal data allow for a more accurate alternative definition, incorporating the element of-and trajectories over- time.
Objectives: To identify trajectories of longitudinal polypharmacy over five years in older people, and to study their association with mortality.
Methods: 300,000 random participants from the UK Clinical Practice Research Datalink GOLD database aged ≥65 on December 31st 2014 were included. Polypharmacy, defined as the cumulative number of prescribed ingredients was calculated at baseline, and at the end of each subsequent follow-up year (2015-2019). Patients exited the study at the earliest of leaving the practice, study end date (31st December 2019), or death. Joint latent class model (JLCM) includes longitudinal and time-to-event sub-models. It divides the population into subgroups (clusters) with different trajectories and associated hazard ratios (HR). To determine the optimum number of clusters, Bayesian Inference Criterion was used. Models were adjusted for baseline age and gender. Intercepts and slopes, mortality hazard ratios (HR), baseline characteristics, and the most common drug classes (WHO anatomical therapeutic chemical classification) of the identified clusters were reported.
Results: Participants were allocated to three distinct clusters based on their progression in polypharmacy: low-steady (n=261985 (87.3%), intercept=5.88, slope=0.14), medium-increasing (14031 (4.7%), intercept=10.68, slope=3.10), and high-decreasing (23984 (8.0%), intercept=22.37, slope=-1.39). Medium-increasing was the oldest; mean age 77.1 (SD=7.8) vs. 75.1 (7.9) and 76.6 (7.9), for low-steady and high-decreasing, respectively. High-decreasing had the highest comorbidity with mean Charlson comorbidity index 2.7 (2.1) vs. 1.3 (1.6) and 2.4 (2.0) for low-steady and medium-increasing, respectively. The most common drug classes were drugs prescribed for alimentary tract and metabolism, and cardiovascular system. Taking low-steady as reference, medium-increasing had the highest HR for death (8.55 (95% CI 8.19 to 8.92)), while high-decreasing had a HR 4.15 (4.02 to 4.30).
Conclusions: We demonstrate the successful use of JLCM to define clusters of older people with distinct longitudinal trajectories of polypharmacy. While having a high number of prescribed ingredients increases the risk of mortality in the next five years, a rapid progression in polypharmacy is associated with an even higher mortality risk than baseline polypharmacy. This can result in time-varying confounding in geriatric pharmacoepidemiology. Further research into the generalisability of these findings is needed.
