Medication Errors

Large language models (LLMs) have emerged as tools to support healthcare delivery, from automating tasks to aiding clinical decision-making. This study evaluated LLMs as alternative to rule-based alert systems, focusing on their ability to identify prescribing errors. This was designed as a prospective, cross-over, open-label study involving 91 error scenarios based on 40 clinical vignettes across 16 medical and surgical specialties. We developed and validated five LLM models using a retrieval-augmented generation framework. The best-performing model evaluated three different implementation strategies: LLM-based clinical decision support system (CDSS) alone, pharmacist plus LLM-based CDSS (co-pilot), and pharmacist alone. The co-pilot arm demonstrated the best performance with an accuracy of 61% (precision 0.57, recall 0.61, and F1 0.59). In detecting errors posing serious harm, the co-pilot mode increased accuracy by 1.5-fold over the pharmacist alone. Effective LLM integration for complex tasks like medication chart reviews can enhance healthcare professional performance, improving patient safety.

ImportanceDeprescribing has the potential to improve patient safety and quality of care by reducing polypharmacy and potentially inappropriate medications (PIMs), which in turn may reduce adverse drug events. Questions remain about the effectiveness of deprescribing interventions in outpatient settings.ObjectiveTo determine the association of deprescribing interventions with reducing medication count and PIMs in community-dwelling older adults.Data sourcesIncluded studies were English-language human studies in PubMed and the Cochrane Library published from January 2019 to July 26, 2024, and results were supplemented with reference-mining and expert consultation.Study selectionStudies were eligible if they were solely or primarily about deprescribing, focused on community-dwelling adults, were multisite, used a randomized trial design, and reported on the primary or secondary outcome.Data extraction and synthesisTwo authors extracted study design, intervention characteristics, population characteristics, and follow-up. Outcomes were extracted by the statistician and checked by a second author. Meta-analyses were conducted using random effects with the Hartung-Knapp-Sidik-Jonkman method. The study was reported following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) criteria.Main outcomes and measuresThe primary outcome was the number of PIMs or total medications, and the secondary outcome was proportion of persons prescribed at least 1 PIM.ResultsTwo authors independently screened 1586 titles from PubMed and Cochrane and 33 from other sources; 321 abstracts and 133 full-text studies were further reviewed, and disagreements were reconciled through discussion, resulting in 17 studies in 18 publications. A total of 8 studies of interventions targeting multiple medications were identified for primary outcome analysis; the random-effects pooled analysis found a mean difference of -0.14 (95% CI, -0.27 to -0.01) medications prescribed. A total of 6 studies of interventions targeting multiple medications were identified for secondary outcome analysis; the random effects pooled analysis found no significant reduction in the proportion of persons prescribed at least 1 PIM (odds ratio, 0.92 [95% CI, 0.74 to 1.14]).Conclusions and relevanceThis systematic review and meta-analysis found moderate-certainty evidence that deprescribing interventions were associated with reduced PIM and medication counts in community-dwelling older adults. While the individual-level association was very small, on an aggregated population level, the outcomes may be large, given the high prevalence of polypharmacy and PIMs in community-dwelling older adults.

PurposePrescribing cascades contribute to potentially inappropriate prescribing, especially among older adults. With the prescribing cascade framework maturation, it is important to distinguish potentially inappropriate from potentially appropriate prescribing cascades. The objective was to create a comprehensive consensus list of Potentially Inappropriate Prescribing Cascades (PIPCs).MethodsA prescribing cascade inventory was compiled using published lists. An international panel of 12 experts in geriatric medicine and pharmacology was selected. Panelists participated in a Delphi consensus process completing two questionnaire rounds and one discussion round.ResultsA total of 107 proposed prescribing cascades were identified. After Questionnaire Round 1, 56 prescribing cascades achieved a rating of agree or strongly agree by ≥ 75% of the panelists and were included in the PIPC list. For 32 of the 107 proposed cascades, 50-74% of panelists provided a rating of agree or strongly agree, and were moved to Round 2. In Questionnaire Round 2, 9 of 32 proposed cascades achieved a rating of agree or strongly agree by ≥ 75% of panelists and were included in the final PIPC list. For 14 prescribing cascades, 50-74% of panelists provided a rating of agree or strongly agree, and were included in the Discussion Round. After discussion, no additional prescribing cascades were included.ConclusionAn explicit list of 65 PIPCs was created using a rigorous Delphi consensus process conducted by international experts on pharmacotherapy for older adults. The PIPC list provides a crucial tool for clinicians and researchers to detect potentially inappropriate prescribing patterns and to foster efforts to improve medication safety.

AimTo examine how medication burden varies across frailty stages in hospitalized older adults, with a focus on potential nonlinear associations: METHODS: We analyzed a multicenter cohort of hospitalized adults aged ≥ 65 years. Frailty was assessed using the preadmission Clinical Frailty Scale (CFS). Associations between CFS and total medication count, potentially inappropriate medication (PIM) count, and anticholinergic burden (Japanese Anticholinergic Risk Scale) were evaluated using restricted cubic spline models adjusted for age, sex, and Charlson Comorbidity Index. Primary analyses were conducted in participants with complete data (n = 1075).ResultsCFS showed significant associations with all medication-related indicators (all overall p ConclusionsMedication burden is strongly and nonlinearly associated with frailty in hospitalized older adults. These stage-dependent patterns suggest that moderate frailty may represent a window for targeted medication review and deprescribing, supporting frailty-stratified approaches to medication management in acute care.