Citations

ABSTRACT Objective To compare outcomes of phenobarbital‐based versus benzodiazepine‐based pathways for alcohol withdrawal syndrome (AWS) in critically ill adults and evaluate whether pathway design and timing of phenobarbital use affect treatment outcomes. Data Sources A librarian‐led search of MEDLINE, Embase, Cochrane Central, CINAHL, Web of Science, Scopus, ClinicalTrials.gov , and the World Health Organization International Clinical Trials Registry Platform was conducted through July 2024, with updates in December 2024, August 2025, and December 2025. Search concepts included phenobarbital, benzodiazepines, and alcohol withdrawal. Study Selection and Data Extraction Comparative studies of adults admitted or transferred to an intensive care unit (ICU) with AWS and treated with phenobarbital‐based versus benzodiazepine‐based pathways were included. Sixteen nonrandomized studies were included in the qualitative synthesis, and 12 treatment‐oriented studies were included in the primary meta‐analysis. Multiple reviewers independently screened studies, extracted data, and assessed risk of bias using Risk Of Bias In Non‐randomized Studies—of Interventions (ROBINS‐I). Data Synthesis Phenobarbital‐based pathways were not associated with significantly different intubation (11 studies; odds ratio (OR), 0.62; 95% confidence interval (CI), 0.20–1.87; I 2 = 73%) or hospital length of stay (five studies; mean differences [MD], −1.75 days; 95% CI, −5.07 to 1.56; I 2 = 63%) compared with benzodiazepine‐based pathways. ICU length of stay was shorter with phenobarbital‐based pathways (eight studies; MD, −0.60 days; 95% CI, −0.79 to −0.41; I 2 = 0%). In exploratory subgroup analyses, front‐loaded, protocolized phenobarbital‐first pathways were associated with lower intubation and shorter hospital length of stay, but certainty of evidence was very low. Relevance to Patient Care and Clinical Practice This review distinguishes phenobarbital‐first, protocolized ICU pathways from adjunctive or mixed phenobarbital strategies, suggesting that implementation design may influence outcomes. Conclusions Although evidence remains very low certainty, these findings suggest that phenobarbital's role in ICU AWS may be pathway dependent. Front‐loaded, protocolized phenobarbital‐first pathways may be associated with favorable outcomes in exploratory analyses, but these findings should be considered hypothesis‐generating and require confirmation in prospective randomized trials. Trial Registration: CRD42024595819

Background/Objectives: Novel vision-language models (VLMs) can integrate patient textual data with image data to support medical diagnosis. Recent studies reported conflicting results regarding the performance of multimodal VLMs compared to other models and physician performance. This systematic review aims to assess the diagnostic performance of multimodal VLMs integrating both patient textual and image data across diverse real-world hospital settings. Methods: We performed comprehensive searches of eight resources, including Embase, MEDLINE, and SCOPUS, on 17 December 2025. Eligible studies reporting diagnostic performance of VLMs integrating both image and patient history textual data from real-world adult patients compared to that of other models and physicians were included. The review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The Prediction model study Risk Of Bias Assessment Tool + AI (PROBAST + AI) was used to assess the quality and risk of bias. The study protocol was registered in the PROSPERO database (CRD420251244054). This review received no external funding. Results: We screened 11,026 records, of which 18 studies met the inclusion criteria. Six studies comparing multimodal and unimodal models demonstrated the consistent superiority of the multimodal models. Four studies evaluating VLM accuracy as standalone agents compared with physician performance reported conflicting evidence. One study assessing VLMs as a clinical copilot demonstrated higher accuracy from the group of physicians using VLM assistance. A meta-analysis could not be performed due to the heterogeneity across study populations and outcomes. The majority of the studies were assessed as having a high risk of bias due to dataset quality. Primary limitations identified across studies include small sample size, a lack of external validation, and the need for prospective clinical deployment studies. No study provided documented considerations regarding model safety or data security. Conclusions: This systematic review suggests that multimodal VLMs consistently outperform unimodal models with access to only image or text. While model performance as standalone agents compared to humans remains inconclusive, a copilot model has demonstrated high diagnostic accuracy. Given substantial methodological concerns across studies, cautious interpretation is required, No firm clinical recommendation can be made regarding the use of standalone VLMs. Further research employing high-quality datasets is needed to ensure the reliability and clinical applicability of future VLMs.

Abstract Introduction Large Language Models (LLMs) have proven effective at analyzing unstructured and structured data from electronic health records (EHR). Their reliance on static EHR data means they miss the dynamic, high-resolution physiological signals that are the first indicators of patient decline, creating a fundamental data lag as clinical documentation of an event may occur hours after its onset. This represents a critical gap in the ability to develop truly effective triage tools. Furthermore, while LLMs can synthesize multi-source data for acuity assessment, they exhibit known limitations, including hallucinations that constrain autonomous use in safety-critical settings. In contrast, a Vision-Language Model (VLM) with strong image-grounded clinical reasoning can interpret vital sign trend images, surpassing the limitations of current EHR-based models. This approach provides immediate triage evaluations that lead to earlier detection of patient deterioration. Methods This mixed-methods study deployed a VLM (Gemini 2.5) integrated within the eICU dashboard. We used near-real-time two-hour trends in heart rate, blood pressure, respiratory rate, and oxygen saturation from consecutive patient cases during alerts in hospital floor units monitored by eICU that admit acute medical and surgical patients. To increase the diversity of the sample, we sought routine cases from patients without alerts and emergent cases from patients admitted to the medical, surgical, trauma, and mixed ICUs. A structured context prompt was used to generate a standardized summary including clinical interpretation, data quality and artifact assessment, triage recommendation, and the model’s chain-of-thought reasoning. Results We included patients from Mayo Clinic Rochester, MN, using an 800-patient cohort for prompt development and a 100-patient cohort for validation against clinically adjudicated results. In the validation cohort the VLM achieved 80% accuracy (80/100). An analysis of the 20 incorrect predictions revealed a clear pattern of conservative over-triage: 80% (16/20) of errors were clinically acceptable adjacent-class discrepancies (routine↔urgent or urgent↔emergent). Only 4/20 discrepancies represented a substantive misclassification. Most errors therefore reflected conservative prioritization rather than incorrect assessment. The model demonstrated a perfect safety profile, underestimating patient acuity in 0% of cases, and thus avoiding the most clinically dangerous error. These results confirm the feasibility of VLM-based approach and provide a strong foundation for the proposed work. Conclusion This VLM performs cross-modal reasoning by fusing deep visual perception with broad clinical context. The result is a fully contextualized and explainable triage prioritization, which provides direct decision support and integrates into the clinical workflow, allowing for faster and more confident acute care assessments. This abstract is funded by: none

This systematic review examined multicenter randomized controlled trials (RCTs) conducted during the Sepsis-3 era that evaluated pharmaceutical and mechanistic interventions in adult patients with sepsis or septic shock, with a focus on identifying interventions that demonstrated mortality benefit. A comprehensive search was performed across Ovid MEDLINE, Embase, Cochrane Central, ClinicalTrials.gov, WHO ICTRP, and Web of Science for trials published since 2016. Studies were included if they enrolled adult patients with sepsis or septic shock using Sepsis-3 criteria and reported mortality as either a primary or secondary outcome. Data were extracted by two independent reviewers and included trial design characteristics, funding source, randomization methods, and reported outcomes. Among the 31 eligible multicenter RCTs enrolling over 7,000 patients across 550 centers, only four trials showed a mortality benefit. These included three studies from China—two evaluating herbal anti-inflammatory compounds and one evaluating aminophylline—and one European study assessing corticosteroid therapy in vasopressor-dependent septic shock. The remaining 27 trials reported no mortality benefit, and substantial heterogeneity was observed in patient populations, sepsis definitions, and outcome reporting. Median mortality in control arms was 31.6% (IQR 24.1–45.5), and 35.4% (IQR 25.7–49.6) in intervention arms. These findings suggest that despite extensive efforts, most interventions have not reduced mortality in multicenter sepsis trials. Future research should focus on optimizing trial design, refining patient selection, and incorporating patient-centered outcomes beyond mortality.

OBJECTIVE: To identify distinct phenotypes of acute respiratory distress syndrome (ARDS) developing after hematopoietic cell transplantation (HCT), using routinely available clinical data at ICU admission. DESIGN: Multicenter retrospective cohort study using latent class analysis. SETTING: ICUs across three Mayo Clinic campuses (Minnesota, Florida, and Arizona). PATIENTS: A total of 166 adult patients who developed ARDS within 120 days following HCT (96 allogeneic, 70 autologous). INTERVENTION: None. MEASUREMENTS AND MAIN RESULTS: Model selection was based on multiple metrics including Bayesian information criteria, entropy, and Vuong-Lo-Mendell-Rubin Likelihood Ratio testing. A two-class model optimally described the cohort. Class 1 (n = 81) was characterized by worse hypoxemia (P/F ratio 157 vs. 210, p = 0.002), higher Pco 2 (41 vs. 36 mm Hg, p < 0.001), and higher bilirubin (1.4 vs. 0.9 mg/dL, p < 0.001) compared with class 2 (n = 85). Both classes included a mix of transplant types, transcending a simple autologous/allogeneic dichotomy, although class 1 had more allogeneic recipients (70.4% vs. 45.9%, p = 0.001). Although time-from-transplant was not a class-defining variable, class 1 occurred later after transplant (30.0 vs. 11.9 d, p < 0.001) with higher frequency of idiopathic pneumonia syndrome (14.8% vs. 2.4%, p = 0.004). Class 2 had more frequent neutropenia (leukocytes 0.4 vs. 5.9 × 109, p < 0.001) and higher frequency of peri-engraftment respiratory distress syndrome (29.4% vs. 9.9%, p = 0.005). Outcomes were significantly worse for class 1 (90-d mortality: 72.8% vs. 48.2%, p = 0.001). An exploratory parsimonious model had good classification accuracy (0.90) using just six variables: leukocyte count, platelet count, bilirubin, Pco 2, body mass index, and temperature. CONCLUSIONS: ARDS after HCT comprises two distinct phenotypes with distinct clinical characteristics and outcomes. These phenotypes align with recognized post-HCT lung injury syndromes and may reflect different underlying biological processes. This framework provides a foundation for investigating targeted therapeutic approaches.

Abstract Rationale: Ongoing healthcare staffing shortages have prompted new care paradigms that incorporate remote care delivery to support timely detection of patient physiological deterioration through mechanisms other than clinician bedside evaluations. In partnership with institutional Information Technology, our team has developed a unique clinical informatics environment, known as the Advanced Medical Practice (AMP) platform, that incorporates data from multiple sources and platforms and supports remote observation of patients from the electronic intensive care unit(eICU). AMP takes advantage of technologies outside of vendor-supported EHRs, and its advanced user-interface design reduces information overload while incorporating real-time data from novel devices and sensors. We partnered with the hospital practice to co-develop a score within the AMP platform(Clinical Events Detection and Response[CEDAR]) that incorporates heart rate, blood pressure, respiratory rate, and SpO2 to stratify patient risk and generate deterioration alerts for the electronic ICU(eICU) team. This formative evaluation employed an implementation science approach to refine CEDAR and ensure it is acceptable to users and can be feasibly implemented in practice. Methods: The study team conducted this pilot work in the eICU at Mayo Clinic (Rochester, MN). We observed the eICU workflow and documented feedback from eICU staff in fieldnotes. eICU staff were asked to log issues in a tracking file during shifts. Issues were reviewed daily and discussed at multidisciplinary team meetings weekly. Observation and tracking data were coded to constructs in the Consolidated Framework for Implementation Research(CFIR 2.0) framework and mapped to Expert Recommendations for Implementation Change(ERIC) strategies that targeted key determinants. Results: Between 9/23/24 and 10/25/24 we completed 25 days of observations. Most identified issues related to technical/design issues like unclear vital trigger alerts (CFIR: Innovation trialability and design); staff knowledge of CEDAR functionality and attitudes about functionality, complexity, or relative advantage compared to current systems (CFIR: Inner setting access to knowledge and information and relative advantage); and difficulties incorporating CEDAR into workflow(CFIR: Inner setting/compatibility). Implementation strategies mapped to CFIR determinants included promoting adaptability of customized interface features; staging implementation scale-up with iterative improvements; providing local technical assistance; assessing for readiness and identifying facilitators and barriers; and conducting educational outreach visits. Conclusions: Formative evaluation during pilot studies are a critical step in the process of intervention development and testing. User engagement and frameworks from implementation science can identify issues of feasibility and inform strategies to ensure acceptability prior to intervention scale-up in clinical practice and readiness for conducting larger implementation studies.

Abstract Objectives This study aimed to evaluate critical care provider perspectives about diagnostic practices for rare and atypical infections and the potential for using artificial intelligence (AI) as a decision support system (DSS). Methods We conducted an anonymous web-based survey among critical care providers at Mayo Clinic Rochester between November 25, 2023, and January 15, 2024, to evaluate their experience with rare and atypical infection diagnostic processes and AI-based DSSs. We also assessed the perceived usefulness of AI-based DSSs, their potential impact on improving diagnostic practices for rare and atypical infections, and the perceived risks and benefits of their use. Results A total of 47/143 providers completed the survey. Thirty-eight out of 47 agreed that there was a delay in diagnosing rare and atypical infections. Among those who agreed, limited assessment of specific patient factors and failure to consider them were the most frequently cited important contributing factors (33/38). Thirty-eight out of 47 reported familiarity with the AI-based DSS applications available to critical care providers. Less than half (18/38) thought AI-based DSSs often provided valuable insights into patient care, but almost three-quarters (34/47) thought AI-based DDSs often provided valuable insight when specifically asked about their ability to improve the diagnosis of rare and atypical infections. All respondents rated reliability as important in enhancing the perceived utility of AI-based DSSs (47/47) and almost all rated interpretability and integration into the workflow as important (45/47). The primary concern about implementing an AI-based DSS in this context was alert fatigue (44/47). Conclusion Most critical care providers perceive that there are delays in diagnosing rare infections, indicating inadequate assessment and consideration of the diagnosis as the major contributors. Reliability, interpretability, workflow integration, and alert fatigue emerged as key factors impacting the usability of AI-based DSS. These findings will inform the development and implementation of an AI-based diagnostic algorithm to aid in identifying rare and atypical infections.