Association between perioperative potentially inappropriate medication exposure levels and postoperative hospital length of stay among chinese older hospitalized patients: a retrospective cohort study

Abstract Background Potentially inappropriate medication (PIM) administration in geriatric surgical patients is increasingly prevalent in the perioperative period. The relation between the variation in PIM exposure and other key postoperative geriatric outcomes, such as postoperative hospital length of stay (POLOS), however, has seldom been reported, and current researches mentioning perioperative PIM concentrate on preoperative long-term home medications or postoperative prescriptions that may not be actually taken. We aimed to investigate whether the level and the change of short-term PIM exposure in the immediate perioperative period was associated with prolonged POLOS (pPOLOS). Methods We performed a retrospective cohort study of patients ≥ 65 years of age who underwent elective inpatient surgery at a tertiary academic hospital from July 2022 and March 2023. PIMs were defined using the Beers Criteria as suggested by the American Geriatrics Society. Stage-varying exposure variables were used to quantify cumulative PIM exposure levels during the preoperative (Pre-PIMs), intraoperative (Intra-PIMs), and even the entire perioperative stage period (Total-PIMs, represented by summing Pre-PIMs and Intra-PIMs) for each participant. A multivariable logistic model and restricted cubic spline model were applied to explore the association and dose-response relationship of PIM exposure with the risk of pPOLOS in the total population and subgroups. Results 196 (44.6%) of 439 participants had a prolonged postoperative length of stay. There was PIM exposure in 378 (86.1%) of the current cohort, analgesics were administered most frequently both preoperatively and intraoperatively albeit via quite different mechanisms of action. Total-PIMs demonstrated superior association with pPOLOS compared to isolated exposure measures, exhibiting greater precision (narrower confidence interval) despite a moderate effect size. This combined metric provided significantly better predictive accuracy (area under the curve=0.763; Delong’s test, P