Associations of potentially inappropriate medications in older adults with mortality and hospitalizations: methodological challenges in pharmacoepidemiology.

Abstract

Previous studies on the association of potentially inappropriate medication (PIM) use with hospitalization risk and all-cause mortality among older adults were prone to confounding and biases. Using data from 217,111 participants of the population-based United Kingdom Biobank, aged 60-69 years, including 95,187 participants with primary care data linkage, the main analysis was a prospective new user design with 1:1 propensity-score stratified by indication matching of new PIM users and new appropriate medication (AM) users (assessed with the EURO-FORTA list). Results were compared to previous approaches with a prevalent user design and a new user design without propensity score matching. 43,307 (19.9%) participants used at least one PIM at baseline. Among 11,812 propensity score matched individuals with new PIM or new AM prescription within 2 years after baseline, new PIM use was associated with non-significantly 20% increased 1-month hospitalization (hazard ratio (HR) [95% confidence interval (95% CI)]: 1.20 [0.76-1.90]) and 23% increased 1-year mortality (1.23 [0.80-1.89]). Null-results were obtained with the prevalent user design (HRs [95% CIs]: 1-month hospitalization: 1.04 [0.83-1.31]; 1-year mortality: 1.01 [0.82-1.23]) and slightly stronger associations in new user design without propensity score matching stratified by indication (1-month hospitalization (1.24 [0.95-1.61]); 1-year mortality (HR [95% CI] 1.57 [1.24-2.00]). This first study with an appropriate methodology showed that previous pharmacoepidemiologic studies on the risk of PIM for hospitalization and mortality have either under- or overestimated the risk. Effect sizes of about 20% appear biologically plausible and larger studies are needed to detect such weak associations with statistical significance.