The impact of pharmaceutical care interventions for medication underuse in older people: a systematic review and meta-analysis

Abstract

Aims: The aim of the present study was to conduct a meta-analysis of controlled trials assessing the impact of pharmaceutical care interventions (e.g. medication reviews) on medication underuse in older patients (≥65 years).

Methods: The databases MEDLINE and EMBASE were searched for controlled studies, and data on interventions, patient characteristics and exposure, and outcome assessment were extracted. Risk of bias was assessed using the Cochrane Collaboration's 'risk of bias' table. Results from reported outcomes were synthesized in multivariate random effects meta-analysis, subgroup meta-analysis and meta-regression.

Results: From 954 identified articles, nine controlled studies, mainly comprising a medication review, were included (2542 patients). These interventions were associated with significant reductions in the mean number of omitted drugs per patient (estimate from six studies with 1469 patients: - 0.44; 95% confidence interval -0.61, -0.26) and the proportion of patients with ≥1 omitted drugs (odds ratio from eight studies with 1833 patients: 0.29; 95% confidence interval 0.13, 0.63). The only significant influential factor for improving success was the utilization of explicit screening instruments when conducting a medication review (P = 0.033).

Conclusion: Pharmaceutical care interventions, including medication reviews, can significantly reduce medication underuse in older people. The use of explicit screening instruments alone or in combination with implicit reasoning is strongly recommendable for clinical practice.

Keywords: elderly; inappropriate prescribing; medication reviews; medication underuse; meta-analysis; systematic literature review.

Figure 1

PRISMA flow chart

Figure 2

Forest plots of intervention effects on the continuous mean of omitted drugs per patient (A) and the proportion of patients with ≥1 omission (B). (* mean changes from baseline or Mantel–Haenszel odds ratio were calculated, accounting for baseline imbalances; † number of potential prescribing omissions (PPOs) was used instead of number of patients unavailable). CI, confidence interval

Figure 3

Subgroup meta-analysis of different outcomes converted into a common effect size (Hedges’ g). Studies are stratified according to the setting of outcome assessment (i.e. inpatient setting, including long-term care and outpatient setting) (A) and the application of explicit screening instruments (B). ACOVE, Assessing Care of Vulnerable Elderly; CI, confidence interval; START, Screening Tool to Alert Doctors to Right Treatments; RE, random-effects

Figure 4

Meta-regression of different outcomes converted into a common effect size (Hedges’ g). Study-level covariates include the mean number of baseline omission per patient (A), patients’ mean age (B), the mean number of drugs per patient (C) and the follow-up duration (D). Regression lines are visualized as solid lines with 95% confidence regions (dotted lines). Studies were weighted according to their sample sizes as indicated by different spot sizes.