Potentially Inappropriate Medication Use in Older Adults in Primary Care According to Four Explicit Criteria: Associations with Patient Factors.

Abstract

<i>Background and Objectives</i>: Potentially inappropriate medication (PIM) use is associated with adverse clinical outcomes in older adults. Several explicit screening tools are available for identifying PIMs, but Romanian data regarding the application of updated criteria remain limited. This study aimed to assess PIM identification according to the STOPP v3, Beers, EU(7)-PIM criteria and PRISCUS v2, to compare PIM counts and PIM presence across criteria, and to explore associations between PIM counts and selected patient characteristics. <i>Materials and Methods</i>: A retrospective observational study with a cross-sectional design was conducted in a single family medicine practice in Iasi, Romania, including 100 patients aged &#x2265; 65 years with at least one chronic condition and receiving at least two medications. Medications documented between January and June 2024 were assessed using the four explicit PIM criteria. One-way ANOVA with Bonferroni correction, repeated-measures ANOVA, agreement analyses, and Poisson regression models were used to examine differences between criteria and associations with patient characteristics, multimorbidity, and medication burden. <i>Results</i>: Mean PIM counts ranged from 0.75 for PRISCUS to 1.31 for STOPP. At least one PIM was identified in 73% of participants according to STOPP, 62% according to Beers, 60% according to EU(7)-PIM, and 53% according to PRISCUS. STOPP identified significantly more PIMs than each of the other three criteria (<i>p</i> &lt; 0.001), while pairwise kappa coefficients ranged from 0.506 to 0.777. Among the 16 one-way ANOVA comparisons, only the association between obesity and EU(7)-PIM counts remained significant after Bonferroni correction (<i>p</i> &lt; 0.001). In adjusted Poisson models, medication burden was associated with PIM counts according to STOPP (<i>p</i> &lt; 0.001), Beers (<i>p</i> = 0.016), and EU(7)-PIM (<i>p</i> = 0.018). Obesity remained associated with EU(7)-PIM counts (<i>p</i> = 0.001), multimorbidity with STOPP PIM counts (<i>p</i> = 0.049), and sex with PRISCUS PIM counts (<i>p</i> = 0.030). Age, type 2 diabetes mellitus, and dyslipidemia were not independently associated with PIM counts. <i>Conclusions</i>: PIM identification varied according to the explicit criteria applied, with STOPP identifying the highest mean PIM count. Medication burden was the factor most consistently associated with PIM counts in adjusted analyses, while other patient characteristics showed criterion-specific associations. Given the small sample, single-practice setting, and non-probability sampling approach, these findings should be considered exploratory and require confirmation in larger, multicenter primary care studies.