Search

Objectives We aimed to examine the adherence of large language models (LLMs) to bacterial meningitis guidelines using a hypothetical medical case, highlighting their utility and limitations in healthcare.Methods A simulated clinical scenario of a patient with bacterial meningitis secondary to mastoiditis was presented in three independent sessions to seven publicly accessible LLMs (Bard, Bing, Claude-2, GTP-3.5, GTP-4, Llama, PaLM). Responses were evaluated for adherence to good clinical practice and two international meningitis guidelines.Results A central nervous system infection was identified in 90% of LLM sessions. All recommended imaging, while 81% suggested lumbar puncture. Blood cultures and specific mastoiditis work-up were proposed in only 62% and 38% sessions, respectively. Only 38% of sessions provided the correct empirical antibiotic treatment, while antiviral treatment and dexamethasone were advised in 33% and 24%, respectively. Misleading statements were generated in 52%. No significant correlation was found between LLMs’ text length and performance (r=0.29, p=0.20). Among all LLMs, GTP-4 demonstrated the best performance.Discussion Latest LLMs provide valuable advice on differential diagnosis and diagnostic procedures but significantly vary in treatment-specific information for bacterial meningitis when introduced to a realistic clinical scenario. Misleading statements were common, with performance differences attributed to each LLM’s unique algorithm rather than output length.Conclusions Users must be aware of such limitations and performance variability when considering LLMs as a support tool for medical decision-making. Further research is needed to refine these models' comprehension of complex medical scenarios and their ability to provide reliable information.

Objective To examine sex differences in prevalence, volume and distribution of vascular brain lesions on MRI among patients with atrial fibrillation (AF).Methods In this cross-sectional analysis, we included 1743 patients with AF (27% women) from the multicentre Swiss Atrial Fibrillation study (SWISS-AF) with available baseline brain MRI. We compared presence and total volume of large non-cortical or cortical infarcts (LNCCIs), small non-cortical infarcts, microbleeds (MB) and white matter hyperintensities (WMH, Fazekas score ≥2 for moderate or severe degree) between men and women with multivariable logistic regression. We generated voxel-based probability maps to assess the anatomical distribution of lesions.Results We found no strong evidence for an association of female sex with the prevalence of all ischaemic infarcts (LNCCI and SNCI combined; adjusted OR 0.86, 95% CI 0.67 to 1.09, p=0.22), MB (adjusted OR 0.91, 95% CI 0.68 to 1.21, p=0.52) and moderate or severe WMH (adjusted OR 1.15, 95% CI 0.90 to 1.48, p=0.27). However, total WMH volume was 17% larger among women than men (multivariable adjusted multiplicative effect 1.17, 95% CI 1.01 to 1.35; p=0.04). Lesion probability maps showed a right hemispheric preponderance of ischaemic infarcts in both men and women, while WMH were distributed symmetrically.Conclusion Women had higher white matter disease burden than men, while volume and prevalence of other lesions did not differ. Our findings highlight the importance of controlling risk factors for cerebral small vessel disease in patients with AF, especially among women.

Purpose The Thrombolysis in Ischemic Stroke Patients (TRISP) collaboration was a concerted effort initiated in 2010 with the purpose to address relevant research questions about the effectiveness and safety of intravenous thrombolysis (IVT). The collaboration also aims to prospectively collect data on patients undergoing endovascular treatment (EVT) and hence the name of the collaboration was changed from TRISP to EVA-TRISP. The methodology of the former TRISP registry for patients treated with IVT has already been published. This paper focuses on describing the EVT part of the registry.Participants All centres committed to collecting predefined variables on consecutive patients prospectively. We aim for accuracy and completeness of the data and to adapt local databases to investigate novel research questions. Herein, we introduce the methodology of a recently constructed academic investigator-initiated open collaboration EVT registry built as an extension of an existing IVT registry in patients with acute ischaemic stroke (AIS).Findings to date Currently, the EVA-TRISP network includes 20 stroke centres with considerable expertise in EVT and maintenance of high-quality hospital-based registries. Following several successful randomised controlled trials (RCTs), many important clinical questions remain unanswered in the (EVT) field and some of them will unlikely be investigated in future RCTs. Prospective registries with high-quality data on EVT-treated patients may help answering some of these unanswered issues, especially on safety and efficacy of EVT in specific patient subgroups.Future plans This collaborative effort aims at addressing clinically important questions on safety and efficacy of EVT in conditions not covered by RCTs. The TRISP registry generated substantial novel data supporting stroke physicians in their daily decision making considering IVT candidate patients. While providing observational data on EVT in daily clinical practice, our future findings may likewise be hypothesis generating for future research as well as for quality improvement (on EVT). The collaboration welcomes participation of further centres willing to fulfill the commitment and the outlined requirements.

Background We investigated outcomes in patients with intracerebral haemorrhage (ICH) according to prior anticoagulation treatment with Vitamin K antagonists (VKAs), direct oral anticoagulants (DOACs) or no anticoagulation.Methods This is an individual patient data study combining two prospective national stroke registries from Switzerland and Norway (2013–2019). We included all consecutive patients with ICH from both registries. The main outcomes were favourable functional outcome (modified Rankin Scale 0–2) and mortality at 3 months.Results Among 11 349 patients with ICH (mean age 73.6 years; 47.6% women), 1491 (13.1%) were taking VKAs and 1205 (10.6%) DOACs (95.2% factor Xa inhibitors). The median percentage of patients on prior anticoagulation was 23.7 (IQR 22.6–25.1) with VKAs decreasing (from 18.3% to 7.6%) and DOACs increasing (from 3.0% to 18.0%) over time. Prior VKA therapy (n=209 (22.3%); adjusted ORs (aOR), 0.64; 95% CI, 0.49 to 0.84) and prior DOAC therapy (n=184 (25.7%); aOR, 0.64; 95% CI, 0.47 to 0.87) were independently associated with lower odds of favourable outcome compared with patients without anticoagulation (n=2037 (38.8%)). Prior VKA therapy (n=720 (49.4%); aOR, 1.71; 95% CI, 1.41 to 2.08) and prior DOAC therapy (n=460 (39.7%); aOR, 1.28; 95% CI, 1.02 to 1.60) were independently associated with higher odds of mortality compared with patients without anticoagulation (n=2512 (30.2%)).Conclusions The spectrum of anticoagulation-associated ICH changed over time. Compared with patients without prior anticoagulation, prior VKA treatment and prior DOAC treatment were independently associated with lower odds of favourable outcome and higher odds of mortality at 3 months. Specific reversal agents unavailable during the study period might improve outcomes of DOAC-associated ICH in the future.