Identifying predictors of medication-related harm in older populations: a latent class analysis approach

Abstract

Background The aim of this study was to apply latent class analysis to identify underlying groupings of predictors, including drug classes and clinical predictors, co-occurring in older people at higher risk of medication-related harm. Method The Adverse Drug reactions in an Ageing PopulaTion cohort was used (N = 798 patients aged ≥65 years admitted acutely to hospital). Seven drug classes; antithrombotic agents, diuretics, renin-angiotensin-aldosterone system, calcium channel blockers, beta-blocking agents, psychoanaleptics, non-steroidal anti-inflammatory drugs, and comorbidity, frailty and significant polypharmacy (10+ different drug classes) were included as potential predictors of medication-related harm. Medication-related harm outcomes included adverse drug reactions (ADR)-related hospital admissions, health-related quality of life, functional impairment and emergency department visits. Determination of the best number of latent classes was based on standard comparison of fit statistics. Univariate and multivariable logistic, linear and Poisson regression models were used to examine the associations between the latent groups and the medication-related harm outcomes. Results A five class model was determined to fit best; (i) high-risk prescribing and polypharmacy group (N = 245); (ii) low-risk group (n = 138); (iii) high-risk prescribing only group (N = 332); (iv) antihypertensive group (N = 18); and (v) psychoanaleptics and polypharmacy group (N = 65). Patients in both the high-risk prescribing and polypharmacy group (a.OR = 2.59, 95%CI = 1.51–4.44) and the high-risk prescribing only group (a.OR = 2.85, 95%CI = 1.57–5.20) were more likely to have an ADR-related hospital admission, with the high-risk prescribing and polypharmacy group also having statistically significant higher functional impairment (β = 1.21, 95% CI = 0.09, 2.33) compared to those in the low-risk group. Conclusion Identifying distinct subgroups of older people based on their medications may lead to more targeted and tailored interventions to reduce potential medication-related harm.