Abstract
AIMS: We estimate trends in severe drug-drug interaction (DDI) prevalence and examine longitudinal associations between DDI exposure and health outcomes (emergency department [ED] visits, quality-of-life [QoL] and functional decline) over approximately 10-years in the older community-dwelling population in Ireland. METHODS: This is a cohort study using linked national-pharmacy-claims data and health outcomes data from waves 1-5 (waves every 2-years; October 2009-January 2019) of The Irish LongituDinal study on Ageing. Community-dwellers aged ≥65 years dispensed ≥2 drugs per claim were eligible for inclusion. DDI prevalence was estimated using a list of 28 225 severe DDIs. Propensity score matching was used to minimize confounding. Generalized linear and Poisson regression was used to model the association between DDI exposure and health outcomes. DDI prevalence, adjusted relative risks (aRR), adjusted β-coefficients and 95% CIs are reported. RESULTS: At baseline, the mean age of n = 1917 eligible participants was 74.9(±6.2) years; 54.6% female, 45.4% polypharmacy (5-9 regular drugs), 12.0% major polypharmacy (≥10 regular drugs). Codispensed DDI prevalence varied from 26.2% (wave 1) to 25.3% (wave 5). After propensity score matching, any severe DDI exposure was associated with an increased risk of ED visit at wave 3 (aRR = 1.49 [95%CI: 1.11, 2.00]) and wave 4 (aRR = 1.40 [1.02, 1.91]); worse QoL at wave 3 (β = -1.22 [-2.24, -0.20]); and increased risk of functional decline at wave 2 (RR = 1.37 [1.08, 1.74]). Individuals with DDIs which increase bleeding risk had greatest risk of ED visit (wave 1: aRR = 1.95 [1.22, 3.10]; wave 3: aRR = 1.82 [1.02, 3.31]) and functional decline (wave 3: aRR = 2.00 [1.05, 3.81]). Those with a contraindicated DDI had worse QoL at wave 1 (β = -3.48 [-6.95, -0.01]) and wave 5 (β = -8.50[-11.00, -6.00]). CONCLUSIONS: We found plausible associations between severe DDI exposure and adverse health outcomes in the older community-dwelling population. DDIs which increase bleeding risk were common and should be targeted for medication optimization.