The Impact of Medications on Fall Risk: A Retrospective Observational Study Based on Medication Data from a Tertiary Hospital (Preprint)

BACKGROUND Patient falls are a critical issue that impacts patient safety and health outcomes while placing a significant burden on nursing staff. The medications administered to patients can strongly influence their risk of falling. Unfortunately, large-scale, multi-departmental studies investigating the relationship between various medications and fall risk are scarce. The PRISCUS list is a catalogue of medications derived by experts that are associated with adverse events, such as falls, in elderly patients. As such, it is a valuable reference for comparing results. OBJECTIVE Conduct a large, multi-departmental pharmacoepidemiologic investigation into the association between various medications and fall risk. The results are then compared with those of PRISCUS 2.0, thus validating existing entries and identifying possible new candidates. METHODS We conducted a retrospective cohort study using routine data from 20 departments at a German tertiary hospital (2016–2022; n=932,103 patients overall; analysis cohort n=8,470; 4,235 fallers and 4,235 random non-fallers). Exposure intervals were derived from medication administrations with a 7-day carryover. For each of 394 medications, exposed vs unexposed patients were matched 1:1 via propensity scores (age, sex, counts of prior diagnoses/procedures); balance was assessed by standardized mean differences (mSMD). We estimated odds ratios (ORs) with 95% CIs and P values (χ² tests) and visualized results in a volcano plot. Pre-specified thresholds (log2(OR)=±0.3; P≤.05) prioritized clinically relevant signals. Expert panels (clinical pharmacists; geriatricians) adjudicated plausibility and potential causality for selected drugs. RESULTS Of the 394 medications examined, 217 exceeded the log₂(OR) threshold. 39 of those showed statistically significant fall-risk-increasing associations (P<.05); 8 were already labeled as fall-risk increasing in PRISCUS 2.0. Clinical pharmacists adjudicated five non-PRISCUS medications (levetiracetam, melatonin, lacosamide, valproic acid, mycophenolic acid) as plausible and potentially causal; geriatricians concurred for three (levetiracetam, melatonin, lacosamide). Three PRISCUS medications without a fall indication (pantoprazole, domperidone, clonidine) were judged by pharmacists as plausible fall-risk-increasing. Geriatricians concurred for clonidine and domperidone. Six medications showed statistically significant fall-risk-decreasing associations. CONCLUSIONS Real-world, multi-department data support and refine the fall-risk entries in PRISCUS 2.0 and suggest additional candidate medications for consideration. Despite warranting consecutive subgroup validation analyses, our findings demonstrate how combining hospital EHR data with targeted clinician adjudication could improve geriatric medication safety. CLINICALTRIAL NCT07181993