Background/Objective: Hand hygiene is a cornerstone of infection prevention; however, compliance is inconsistent in intensive care units (ICUs), particularly in resource-constrained settings. This study evaluated whether a low-cost, multimodal quality improvement intervention could improve process-level hand hygiene compliance using routine, episode-based audits embedded in the ICU practice. Methods: We conducted a single-cycle Plan-Do-Study-Act quality improvement project in a 12-bed mixed medical–surgical ICU in Pakistan (December 2023–January 2024). Hand hygiene performance was assessed using the unit’s routine weekly episode-based audit protocol, aligned with the WHO Five Moments framework. A targeted multimodal intervention comprising education, point-of-care visual reminders, audit feedback, and leadership engagement was implemented between the pre- and post-intervention phases (four weeks each). Non-applicable moments were scored as “compliant by default” according to the institutional protocol. A sensitivity analysis was performed excluding these moments to calculate pure adherence. Compliance proportions were summarized using exact 95% Clopper–Pearson confidence intervals without inferential testing. Results: A total of 942 audit episodes (471 per phase) generated 4710 moment-level assessments were generated. Composite hand hygiene compliance increased from 63.1% pre-intervention to 82.0% post-intervention [absolute increase: 18.9 percentage points (pp)]. Sensitivity analysis excluding non-applicable moments demonstrated pure adherence improvement from 54.2% to 82.5% (+28.3 pp), confirming a genuine behavioral change rather than a measurement artifact. Compliance improved across all five WHO moments, with the largest gains in awareness-dependent moments targeted by the intervention: before touching the patient (+27.0 pp) and after touching patient surroundings (+40.0 pp). Week-by-week compliance remained stable within both phases, without immediate post-intervention decay. Conclusions: A pragmatic, low-cost multimodal intervention embedded in routine ICU workflows was associated with substantial short-term improvements in hand hygiene compliance over a four-week observation period, particularly for awareness-dependent behaviors. Episode-based audit systems can support directional process monitoring in resource-limited critical care settings without the need for electronic surveillance. However, its long-term sustainability beyond one month and generalizability to other settings remain unknown. Sensitivity analyses are essential when using “compliant by default” scoring to distinguish adherence patterns from measurement artifacts.

Peer-assisted learning enhances conceptual understanding in anatomy education; however, its impact on structured cognitive outcomes remains unclear. We evaluated the effect of the Peer-Assisted Uniform Learning (PAUL) program on musculoskeletal (MSK) anatomy objective structured practical examination (OSPE) performance. We used a cross-sectional, controlled, observational design to analyse the OSPE scores of 304 students. Seventy-nine students who attended at least two PAUL sessions formed the exposure group, while 225 served as the control group. The PAUL participants alternated between tutor and tutee roles during the sessions. The OSPE items were divided into Cluster 1 (identification/recall) and Cluster 2 (application/analysis). A validated survey was used to assess the participants perceived learning outcomes and satisfaction. PAUL participants outperformed non-participants in Cluster 1 (78.4 vs. 73.7, p = 0.01) and Cluster 2 (87.2 vs. 82.8, p = 0.02). The relative gain was greater for application-based items (interaction effect: F = 8.37, p = 0.004), supporting the program's effect on higher-order reasoning. Survey responses (91 % response rate) showed high satisfaction (mean 4.3 ± 0.6) and perceived learning gains (mean 4.1 ± 0.7); 85 % of students reported improved understanding of the subject. The PAUL program, grounded in social constructivist principles and supported by faculty oversight, improved performance in both recall and application of MSK anatomy OSPE items, with the strongest impact on application-based learning.

Medical education faces increasing demand for scalable grading solutions. Large language models have been proposed as automated graders for high-stakes assessments, but evidence for their reliability in multimodal practical examinations remains limited. This retrospective inter-rater reliability study compared three LLMs: ChatGPT-4o, Gemini 2.5 Flash, and Claude 3.5 Haiku with original human reference scores on 19 integrated anatomy, histology, and physiology objective structured practical examination items completed by 309 pre-medical students. Human reference scores were assigned during live summative grading by two content experts who divided items, with each item marked by one expert across all students; no human inter-rater reliability estimate was available. Items required simultaneous image interpretation and short-answer responses. Models were evaluated using zero-shot prompting reflecting deployment-realistic conditions. Agreement was assessed using Spearman's ρ, Cohen's κ, intraclass correlation coefficients, and Bland-Altman analysis. Item-level gap analysis compared student success rates with LLM performance across items. Rank-order correlations were strong across all models (ρ = 0.784-0.921), but categorical agreement diverged substantially. Pass/fail agreement ranged from 52.1% (ChatGPT; κ = 0.067, slight) to 84.1% (Claude; κ = 0.680, substantial). Bland-Altman analysis showed inconsistent systematic bias: Claude over-scored by +2.5 points, Gemini under-scored by -5.0 points, and ChatGPT by -10.0 points. Item-level gap analysis highlighted three recurring divergence patterns: non-standard histological staining, three-dimensional spatial reasoning from two-dimensional images, and semantic inflexibility in short-answer evaluation. The most extreme case was a pelvic three-dimensional model item on which 93.3% of students succeeded by human grading but all three LLMs assigned a mean score of zero. Strong rank-order correlation alone does not support LLM grading for categorical decisions in high-stakes assessment contexts. LLMs may serve as assessment assistants for first-pass ranking and discrepancy flagging, but item-level human review of flagged discordances, error-profile monitoring, and preserved human authority over pass/fail decisions are required before high-stakes deployment.

Background: The co-listing of liver disease and hypertension (HTN) codes on U.S. death certificates is a documentation pattern that may reflect changing clinical recognition, certification practices, or underlying disease burden; however, these explanations cannot be distinguished from death certificate data alone. Understanding co-listing patterns may help identify populations warranting further clinical and epidemiological investigation. This study analyzed 25 years of U.S. mortality data to characterize temporal, demographic, and geographic patterns in deaths due to both conditions. Methods: We conducted a population-based cross-sectional analysis of CDC WONDER multiple cause-of-death data (1999–2023) for adults aged ≥25 years. Deaths listing both liver disease (ICD-10: K70–K76) and HTN (I10–I15) were identified. Age-adjusted mortality rates (AAMRs) per 100 000 people were standardized to the 2000 U.S. population. Temporal trends were assessed using Joinpoint regression with annual percent changes (APCs). Analyses were stratified by sex, race/ethnicity, region, urbanization, and place of death. Sensitivity analyses using four alternative certificate definitions, including underlying-cause-only and co-listed definitions, were conducted to assess robustness to cause-of-death coding choices. Results: Between 1999 and 2023, 208 666 deaths listed both conditions, increasing from 2307 to 18 769 annually. The national AAMR increased from 1.30 to 6.75 per 100 000 (average APC: +6.71%, 95% CI: 5.64–7.80). Joinpoint regression identified four phases: rapid early growth (1999–2001, APC +19.50%), steady increase (2001–2018, APC +4.27%), substantial increase (2018–2021, APC +17.64%), and stabilization (2021–2023, APC +0.28%). Males showed higher rates than females (9.02 vs 4.71 per 100 000 in 2023). Non-Hispanic American Indian/Alaska Native populations showed the highest rates (12.74 per 100 000), followed by Hispanic (8.15), non-Hispanic White (6.95), and non-Hispanic Black (6.84) populations. Geographic variation ranged from 2.87 (Connecticut) to 21.54 (Oklahoma) per 100 000. Sensitivity analyses confirmed that increasing trends were robust to the choice of cause definition, although the post-2021 trajectory was definition-sensitive. Conclusion: The number of deaths listed with both liver disease and hypertension codes increased substantially over 25 years, with notable acceleration during 2018–2021 that coincided temporally with the COVID-19 pandemic; however, causal attribution cannot be made from death certificate data. Because the findings reflect death certificate documentation rather than validated clinical comorbidity, they should be interpreted as surveillance indicators identifying demographic and geographic patterns that warrant further investigation through clinical registries and prospective study designs.

Background: The efficacy of tenecteplase for acute ischemic stroke (AIS) beyond 4.5 h remains uncertain, particularly across care pathways with and without endovascular thrombectomy (EVT). We performed an updated systematic review and meta-analysis using an EVT-stratified framework. Methods: PubMed, Embase, Scopus, and the Cochrane Library were searched through February 2026 for randomized controlled trials comparing tenecteplase with control in imaging-selected patients with AIS presenting 4.5–24 h from last known well. The primary outcome was excellent functional outcome (mRS 0–1) at 90 days. Secondary outcomes were good functional outcome (mRS 0–2), recanalization, early neurological improvement, symptomatic intracranial hemorrhage, and 90-day mortality. Random-effects models with Hartung–Knapp adjustment were used. Subgroup analyses by EVT availability were interpreted as exploratory because of the limited number of trials. Results: Five trials including 1844 patients were analyzed. Tenecteplase improved excellent functional outcome (RR 1.25, 95% CI 1.10–1.42; p = 0.0005) with no heterogeneity (I² = 0%) and no interaction by EVT status (p-interaction = 0.961). Good functional outcome was not significantly different overall (RR 1.10, 95% CI 0.97–1.24; p = 0.135). Significant subgroup interactions were observed for recanalization (p-interaction = 0.004) and early neurological improvement (p-interaction = 0.002), with benefits concentrated in non-EVT settings. However, the larger effect on recanalization did not translate proportionally into functional recovery, supporting separation of vessel-opening outcomes from patient-centered outcomes. Symptomatic intracranial hemorrhage showed a nonsignificant increase in four estimable studies (RR 1.88, 95% CI 0.94–3.78; p = 0.074), whereas 90-day mortality did not differ significantly (RR 1.11, 95% CI 0.85–1.43; p = 0.43). Conclusions: In imaging-selected AIS presenting 4.5–24 h after onset, tenecteplase improved excellent functional outcome irrespective of EVT availability, while benefits for recanalization and early neurological improvement were largely confined to non-EVT settings. Because recanalization is an intermediate endpoint, these findings should not be interpreted as proof of a proportional clinical benefit. Future extended-window trials should specify EVT status.