Comparison of four criteria for potentially inappropriate medications in Brazilian community-dwelling older adults

Aim: To compare four potentially inappropriate medication (PIM) criteria from different regions of the world in terms of their characteristics, concordance, sensitivity, specificity and predictive values in a community-dwelling sample. Methods: Across-sectional, observational, epidemiological study was carried out by a door-to-door survey in a Brazilian city. The following PIM criteria were applied: Beers-2015, Screening Tool of Older People's Potentially Inappropriate Prescriptions (STOPP)-2015, The European Union (EU)(7)-PIM list and Taiwan criteria. The associations of criteria with the presence or absence of falls, hospitalizations and cognitive impairment were determined, and their sensitivity, specificity, positive predictive value, negative predictive value and concordance among the criteria were assessed. Results: A total of 368 (92%) older adults were in continuous use of at least one drug. A high prevalence of PIM by Beers (50.0 %), STOPP (46.2%), EU(7)-PIM (59.5%) and Taiwan 31.3%) criteria was found. There was a high concordance among the PIM criteria (66.3-81.8%), and a moderate-to-high intraclass correlation between criteria (0.607-0.851). In general, the Taiwan criterion had lower levels of sensitivity (25.7-34.0%) and higher levels of specificity (67.8-70.3%), The EU(7)-PIM criteria had higher levels of sensitivity (60-75.3%) and lower levels of specificity (41.1-46.9%), whereas the Beers and STOPP had a more balanced sensitivity/specificity ratio (sensitivity: STOPP 50.7-55.3% and Beers 53.0-56.9 %; specificity: STOPP 56-56.6% and Beers 51.6-53.8%). Conclusions: The present study found moderate-to-high levels of concordance among the four PIM criteria assessed, pointing to a consensus in this field. However, each criterion showed particular characteristics: the EU(7)-PIM criterion had higher sensitivity, the Taiwan criterion higher specificity, and the Beers and STOPP a more balanced profile. These results highlight that each criterion has its own characteristics, and should be used according to health providers' objectives.