Wolfson Medical Center

STUDY OBJECTIVES:To evaluate the performance and safety of a large language model in interpreting drug-induced sleep endoscopy (DISE) videos and providing treatment recommendations for obstructive sleep apnea, compared to expert human raters and the contemporaneous clinical report. METHODS:This prospective, blinded study included 16 adults undergoing drug-induced sleep endoscopy at a tertiary academic center. For each case, an anonymized procedural video, clinical vignette, and examination findings were independently reviewed by two sleep surgery experts, a senior otolaryngology resident, and a large language model. All the raters assessed the video quality, airway maneuvers, airway collapse at each anatomical site, and recommended therapy. Concordance with the clinical reference standard was evaluated using Cohen's kappa and intraclass correlation coefficients. Safety and reproducibility were assessed through subgroup and error-type analyses. RESULTS:Human raters demonstrated perfect or near-perfect agreement with the reference standard for the velum, oropharynx, tongue base, and jaw thrust response, and moderate agreement for epiglottic collapse. The large language model matched expert performance for all domains except the epiglottis, where moderate agreement was observed. Model-generated treatment recommendations were safe, consistent with accepted clinical practice, and highly reproducible between independent runs. No unsafe or discordant recommendations were identified. CONCLUSIONS:In this single-center study of 16 patients, a large language model accurately interpreted DISE videos and generated safe recommendations consistent with accepted clinical practice, approximating expert performance in most domains. Larger multicenter cohorts are needed to validate these findings and confirm generalizability. Statement of Significance This study demonstrates that artificial intelligence can interpret complex airway videos in sleep surgery and provide safe, expert-level treatment advice. By comparing the performance of a large language model with that of experienced clinicians, our findings suggest that advanced technology can help standardize decision-making in a highly subjective area of care. This work highlights the promise of artificial intelligence as an adjunct to clinical judgment, especially in settings where expert access is limited. However, important questions remain about the use of artificial intelligence in challenging cases and its integration into real-world practice. Future research should focus on larger, more diverse groups of patients and on ensuring ongoing oversight and safety.

Background: Mechanical ventilation is essential for managing acute respiratory distress syndrome (ARDS) but may contribute to ventilator-induced lung injury (VILI). Mechanical power (MP), an integrative measure combining tidal volume, respiratory rate, and airway pressures, has emerged as a predictor of VILI. However, conventional MP calculations based on airway driving pressure (ΔP) reflect the mechanical load applied to the entire respiratory system, including both lung parenchyma and chest wall. Estimation of transpulmonary pressure using esophageal pressure allows calculation of transpulmonary mechanical power (TPMP), which may more accurately represent lung stress. Methods: We conducted a retrospective observational study of 20 mechanically ventilated ICU patients with available esophageal pressure measurements at Wolfson Medical Center between 2012 and 2015. MP and TPMP were calculated across ventilation modes including continuous mandatory ventilation assist/control (CMV), synchronized intermittent mandatory ventilation (SIMV), and continuous positive airway pressure (CPAP). Paired comparisons were analyzed using the Wilcoxon signed-rank test. Results: At baseline, MP and TPMP were similar but significantly different (20 ± 6 vs. 21 ± 7 J/min; p = 0.0107). During CMV, TPMP exceeded MP (25 ± 7 vs. 23 ± 6 J/min; p = 0.0016). In contrast, TPMP was lower than MP during CPAP (15 ± 8 vs. 22 ± 8 J/min; p = 0.0073) and SIMV (19 ± 7 vs. 23 ± 7 J/min; p < 0.001). Conclusions: Airway-based MP calculations may misrepresent lung parenchymal load. TPMP derived from esophageal pressure provides a more physiologically accurate estimate of lung stress and may improve individualized ventilatory management in ARDS.

Data on the influence of sociodemographic characteristics on the epidemiology of community-acquired bacteremia in vaccinated populations remains limited. We aimed to examine the association between sociodemographic characteristics and the epidemiology of pediatric community acquired bacteremia.Rates of Community Acquired Bacteremia in Low and High Socioeconomic Status Populations This retrospective study was conducted in two hospitals serving populations with significant socioeconomic disparities (low socioeconomic status [SES] rank 2/10 vs. high SES: rank 7/10, based on the Central Bureau of Statistics' scale). We included all cases of community-acquired bacteremia in patients< 18 years between 2016-2023. Bacteremia rates were calculated per 1,000 blood cultures drawn. Rates of true bacteremia and the distribution of isolates were compared between hospitals. S.pneumoniae isolates were further analyzed to determine serotype coverage by the Prevenar 13 and Prevenar 20 vaccines. We identified 298 cases of community-acquired bacteremia. Bacteremia rates were higher in low- vs. high-SES children < 3 years (10.5 vs. 5.2 cases/1,000 cultures; RR -2, P < 0.01 and 10.06 vs. 4.1 per 1,000 cultures; RR-2.4, P < 0.01, in those aged < 3 months and 3–36 months, respectively). S.pneumoniae was the leading isolate among children aged 3–36 months (n=55, 44.3% in low vs. n=36, 22.3% in high-SES populations). In the low-SES population, S.pyogenes (10.4%) and N.meningitidis (7.0%) were the next most common isolates, while K.kingae (16.0%) and M.catarrhalis (11.7%) followed in the high-SES population. Non-Prevnar 13 S.pneumoniae serotypes accounted for 94.4% of cases in high-SES and 89% in low-SES populations (P = 0.4). Prevnar 20 covered isolates accounted for 50.9% in low-SES and 38.8% in high-SES populations (P = 0.2). Low-SES children < 3 years had more than twice the risk of bacteremia compared to high-SES children, with higher rates of S.pyogenes and N.meningitidis. Non–Prevnar 13 S. pneumoniae serotypes accounted for the majority of cases in both populations, with a substantial proportion covered by the recently introduced Prevnar 20 vaccine, highlighting its potential to reduce the burden of invasive pneumococcal disease. All Authors: No reported disclosures

Background Small intestinal neuroendocrine tumors (SiNETs) are rare tumors with propensity to cause bulky fibrotic mesenteric masses by hematogenous and lymphatic invasion. They exhibit variable oncological behavior and reliable prognostic markers remain elusive. The 8 th edition American Joint Commission on Cancer introduced mesenteric mass ≥2 cm to improve prognostication. The clinical utility in prognostication remains unknown. We reviewed outcomes of resected SiNETs to determine if computed tomography (CT) size is associated with oncological outcomes. Methods Retrospective review of resected SiNET patients at two tertiary centers with evaluable preoperative CT scans. Measured the largest diameter of the mesenteric mass and the cutoff of 2 cm used to correlate with oncological outcomes. Results Seventy-three patients operated for SiNET with preoperative CT were included. The largest mesenteric mass was <2 cm in 28 and ≥2 cm in 45 patients. Cohorts were similar in demographic and clinical characteristics. Median follow-up was 47 months. Overall survival (OS) and progression free survival (PFS) were similar. No significant differences in estimated mean OS (136 vs 139 months, P = 0.981) or median PFS (90 vs 71 months, P = 0.126) for the <2 cm and >2 cm cohorts, respectively. Univariable analysis failed to show an association between mesenteric size as a continuous variable, and at the ≥2 cmcutoff. Age, cardiac comorbidities, and palliative intent surgery were the only factors that correlated with survival. Discussion Mesenteric mass size on preoperative CT was not correlated with oncological outcomes and by itself showed little utility in preoperative decision-making.

PURPOSE:Define the etiologic profile of uveitis in China. MATERIALS AND METHODS:Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-guided systematic review of studies from PubMed, Scopus, PubMed Central, Web of Science, and the China National Knowledge Infrastructure (CNKI) (January 2005-January 2025). Eligible cross-sectional and cohort studies reporting etiologies were included. Proportions with 95% confidence intervals (CIs) were pooled, and heterogeneity was assessed with I2. RESULTS:Of 2,273 records, 10 studies met criteria (2005-2025; sample sizes 144-15373). Pooled proportions: idiopathic uveitis 41.75% (95% CI 36.88-46.78%); Vogt-Koyanagi-Harada (VKH) syndrome 14.12% (95% CI 10.56-18.63%); Behçet's disease 7.7% (95% CI 5.2-11.27%); and Ankylosing spondylitis 5.75% (95% CI 3.65-8.96%). Infectious causes were less frequent: herpetic uveitis 2.22% (95% CI 1.47-3.34%), acute retinal necrosis (ARN) 0.94% (95% CI 0.39-2.25%), and syphilis 1.03% (95% CI 0.59-1.79%), although rates varied significantly across centers. CONCLUSION:Uveitis in China is largely idiopathic (~41.75%), with a predominance of autoimmune etiologies, specifically Vogt-Koyanagi-Harada (VKH) syndrome and Behçet's disease. These findings reflect a distinct epidemiological profile.