Background: The GPT-4 is a large language model (LLM) trained and fine-tuned on an extensive dataset. After the public release of its predecessor in November 2022, the use of LLMs has seen a significant spike in interest, and a multitude of potential use cases have been proposed. In parallel, however, important limitations have been outlined. Particularly, current LLMs encounter limitations, especially in symbolic representation and accessing contemporary data. The recent version of GPT-4, alongside newly released plugin features, has been introduced to mitigate some of these limitations. Objective: Before this background, this work aims to investigate the performance of GPT-3.5, GPT-4, GPT-4 with plugins, and GPT-4 with plugins using pretranslated English text on the German medical board examination. Recognizing the critical importance of quantifying uncertainty for LLM applications in medicine, we furthermore assess this ability and develop a new metric termed “confidence accuracy” to evaluate it. Methods: We used GPT-3.5, GPT-4, GPT-4 with plugins, and GPT-4 with plugins and translation to answer questions from the German medical board examination. Additionally, we conducted an analysis to assess how the models justify their answers, the accuracy of their responses, and the error structure of their answers. Bootstrapping and CIs were used to evaluate the statistical significance of our findings. Results: This study demonstrated that available GPT models, as LLM examples, exceeded the minimum competency threshold established by the German medical board for medical students to obtain board certification to practice medicine. Moreover, the models could assess the uncertainty in their responses, albeit exhibiting overconfidence. Additionally, this work unraveled certain justification and reasoning structures that emerge when GPT generates answers. Conclusions: The high performance of GPTs in answering medical questions positions it well for applications in academia and, potentially, clinical practice. Its capability to quantify uncertainty in answers suggests it could be a valuable artificial intelligence agent within the clinical decision-making loop. Nevertheless, significant challenges must be addressed before artificial intelligence agents can be robustly and safely implemented in the medical domain.

Turning during walking adds complexity to gait and has been little investigated until now. Research question What are the differences in spatiotemporal parameters between young and elderly healthy adults performing quarter-turns (90°), half-turns (180°) and full-turns (360°)? The spatiotemporal parameters of 10 young and 10 elderly adults were recorded in a laboratory while turning at 90°, 180° and 360°. Two-way mixed ANOVA were performed to determine the effect of age and turning amplitude. Elderly were slower and needed more steps and time to perform turns of larger amplitude than young adults. Cadence did not differ across age or across turning amplitude. Generally, in the elderly, the spatial parameters were smaller and the temporal parameters enhancing stability (i.e., double-support phase and stance/cycle ratio) were larger, especially for turns of larger amplitudes. In elderly adults, the variability of some spatial parameters was decreased, whereas the variability of some temporal parameters was increased. Stride width of the external leg showed the most substantial difference between groups. Most parameters differed between turning at 90° and turning at larger amplitudes (180°, 360°). Significance This study extends the characterization of turning biomechanics with respect to ageing. It also suggested paying particular attention to the turning amplitude. Finally, the age-related differences may pave the way for new selective rehabilitation protocols in the elderly.

This study aimed to analyze the effect of COVID-19 vaccination on the occurrence of ARDS in hospitalized COVID-19 patients. The study population of this retrospective, single-center cohort study consisted of hospitalized COVID-19 patients with known vaccination status and chest computed tomography imaging between July 2021 and February 2022. The impact of vaccination on ARDS in COVID-19 patients was assessed through logistic regression adjusting for demographic differences and confounding factors with statistical differences determined using confidence intervals and effect sizes. A total of 167 patients (69% male, average age 58 years, 95% CI [55; 60], 42% fully vaccinated) were included in the data analysis. Vaccinated COVID-19 patients had a reduced relative risk (RR) of developing ARDS (RR: 0.40, 95% CI [0.21; 0.62]). Consequently, non-vaccinated hospitalized patients had a 2.5-fold higher probability of developing ARDS. This risk reduction persisted after adjusting for several confounding variables (RR: 0.64, 95% CI [0.29; 0.94]) in multivariate analysis. The protective effect of COVID-19 vaccination increased with ARDS severity (RR: 0.61, 95% CI [0.37; 0.92]). Particularly, patients under 60 years old were at risk for ARDS onset and seemed to benefit from COVID-19 vaccination (RR: 0.51, 95% CI [0.20; 0.90]). COVID-19 vaccination showed to reduce the risk of ARDS occurrence in hospitalized COVID-19 patients, with a particularly strong effect in patients under 60 years old and those with more severe ARDS.

COVID-19 vaccination has been shown to prevent and reduce the severity of COVID-19 disease. The aim of this study was to explore the cardioprotective effect of COVID-19 vaccination in hospitalized COVID-19 patients. In this retrospective, single-center cohort study, we included hospitalized COVID-19 patients with confirmed vaccination status from July 2021 to February 2022. We assessed outcomes such as acute cardiac events and cardiac biomarker levels through clinical and laboratory data. Our analysis covered 167 patients (69% male, mean age 58 years, 42% being fully vaccinated). After adjustment for confounders, vaccinated hospitalized COVID-19 patients displayed a reduced relative risk for acute cardiac events (RR: 0.33, 95% CI [0.07; 0.75]) and showed diminished troponin T levels (Cohen’s d: − 0.52, 95% CI [− 1.01; − 0.14]), compared to their non-vaccinated peers. Type 2 diabetes (OR: 2.99, 95% CI [1.22; 7.35]) and existing cardiac diseases (OR: 4.31, 95% CI [1.83; 10.74]) were identified as significant risk factors for the emergence of acute cardiac events. Our findings suggest that COVID-19 vaccination may confer both direct and indirect cardioprotective effects in hospitalized COVID-19 patients.

Background: Safe reuse of multimodal hospital data for AI development is limited by the absence of reliable, context-aware deidentification across multimodal data and longitudinal patient data. Existing approaches are largely modality-specific and can indiscriminately remove clinically important information. Methods: We developed the Multimodal Anonymizer, a modular, locally deployable multi-agent framework integrating multimodal large language models, task-specific neural networks and rule-based transformations. We evaluated 16 orchestrator model configurations on a benchmark built from publicly available data and hospital data from our institution. The benchmark dataset included data from different origins: 250 MIMIC-IV patients with synthetically injected personally identifiable information (PII) supplemented with head CT, face images, handwriting, audio, German clinical-text datasets and local data. Primary outcomes were deidentification sensitivity and preservation of clinically important content; secondary analyses examined model characteristics, reproducibility, and performance against leading market and open-source solutions. Results: The best local configuration—the orchestrator being Qwen3-VL-235B-A22B-Thinking—achieved near-complete deidentification across all datasets, with per-patient sensitivity of 98.80% (95%-CI 97.20; 100), and per-PII sensitivity of 99.82% (95%-CI 99.76; 99.88). Critical clinical preservation was 99.60% (95%-CI 98.80; 100) per-patient, and clinical preservation was 99.61% (95%-CI 99.51; 99.71) per-file. All modalities achieved at least 98.30% sensitivity (lower bound 95%-CI). On our local data, the system achieved a deidentification sensitivity of 100% per-patient and per-PII; and a critical clinical preservation of 100% per-patient as well as a clinical preservation of 99.97% (95%-CI 99.91; 100) per-file. When comparing orchestrators, the leading local models were similar to proprietary models (GPT-5.2) in deidentification sensitivity while showing higher deidentification specificity. The Multimodal Anonymizer outperformed previous tools on most modalities. Conclusion: Near-complete, utility-preserving deidentification of multimodal clinical data is achievable with a unified, locally deployable multi-agent system, enabling safer large-scale reuse of hospital data for research and AI development. Graphical Highlights: Framework for deidentification of multimodal clinical data. Multimodal deidentification with preservation of clinically relevant content. On-premises plug-and-play deployment for local data processing. Evaluation of 16 model configurations and comparison with existing tools. Assessment on external, multilingual and site-specific datasets. Short Description: The Multimodal Anonymizer is a fully local, multi-agent system that prepares multimodal clinical records for privacy-preserving reuse by coordinating multimodal large language model reasoning, specialist neural networks, rule-based transformations, and iterative verification. Across benchmarks spanning text, tables, PDFs, imaging, metadata, filenames, audio, and handwriting, its best configuration using a local open-source multimodal large language model achieved 98.80% patient-level deidentification sensitivity and 99.60% preservation of clinically critical content, performing comparably to proprietary models and outperforming established deidentification tools across most modalities.

Introduction: Turning is a complex locomotor task that can accentuate gait impairments and precipitate freezing of gait (FOG) in Parkinson’s disease (PD). However, gait-derived spatiotemporal parameters during inter-episodic turning remain incompletely characterized. We compared turning-related gait patterns among healthy older adults, patients with PD without FOG, and patients with PD and a history of FOG. Methods: Ten healthy older controls, 10 patients with PD without FOG (PD-P), and 10 patients with PD and FOG (FOG-P) performed 90°, 180°, and 360° turns while undergoing optical motion capture. Spatiotemporal gait parameters were analyzed using Bayesian generalized linear models. Trials containing observable FOG episodes were excluded; consequently, the analyses represent inter-episodic turning rather than freezing episodes themselves. Results: Compared with controls, PD-P participants demonstrated reduced stride width of the internal leg and increased stride width of the external leg, whereas FOG-P participants demonstrated shorter step and stride lengths. Both PD groups predominantly used step-out or mixed turning strategies and required more time and steps to complete larger-amplitude turns. Discussion: Gait-derived spatial parameters during turning differed between the PD groups. PD-P participants showed more pronounced mediolateral foot-placement differences, whereas FOG-P participants showed more pronounced reductions in anteroposterior step and stride length. Because the groups differed in overall motor severity and the analyses were intentionally restricted to inter-episodic turning, these findings should be interpreted as descriptive turning gait patterns in patients with a history of FOG rather than as evidence of FOG-specific mechanisms.