Background: AI agents built on large language models (LLMs) can plan tasks, use external tools, and coordinate with other agents. Unlike standard LLMs, agents can execute multi-step processes, access real-time clinical information, and integrate multiple data sources. There has been interest in using such agents for clinical and administrative tasks, however, there is limited knowledge on their performance and whether multi-agent systems function better than a single agent for healthcare tasks. Purpose: To evaluate the performance of AI agents in healthcare, compare AI agent systems vs. standard LLMs and catalog the tools used for task completion Data Sources: PubMed, Web of Science, and Scopus from October 1, 2022, through August 5, 2025. Study Selection: Peer-reviewed studies implementing AI agents for clinical tasks with quantitative performance comparisons. Data Extraction: Two reviewers (A.G., M.O.) independently extracted data on architectures, performance metrics, and clinical applications. Discrepancies were resolved by discussion, with a third reviewer (E.K.) consulted when consensus could not be reached. Data Synthesis: Twenty studies met inclusion criteria. Across studies, all agent systems outperformed their baseline LLMs in accuracy performance. Improvements ranged from small gains to increases of over 60 percentage points, with a median improvement of 53 percentage points in single-agent tool-calling studies. These systems were particularly effective for discrete tasks such as medication dosing and evidence retrieval. Multi-agent systems showed optimal performance with up to 5 agents, and their effectiveness was particularly pronounced when dealing with highly complex tasks. The highest performance boost occurred when the complexity of the AI agent framework aligned with that of the task. Limitations: Heterogeneous outcomes precluded quantitative meta-analysis. Several studies relied on synthetic data, limiting generalizability. Conclusions: AI agents consistently improve clinical task performance of Base-LLMs when architecture matches task complexity. Our analysis indicates a step-change over base-LLMs, with AI agents opening previously inaccessible domains. Future efforts should be based on prospective, multi-center trials using real-world data to determine safety, task matched and cost-effectiveness. Primary Funding Source: This work was supported in part through the computational and data resources and staff expertise provided by Scientific Computing and Data at the Icahn School of Medicine at Mount Sinai and supported by the Clinical and Translational Science Awards (CTSA) grant UL1TR004419 from the National Center for Advancing Translational Sciences. Research reported in this publication was also supported by the Office of Research Infrastructure of the National Institutes of Health under award number S10OD026880 and S10OD030463. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Registration: PROSPERO CRD420251120318

Background: Female gender is a known risk factor for long COVID. With the increasing number of COVID-19 cases, the corresponding number of survivors is also expected to rise. To the best of our knowledge, no systematic review has specifically addressed the gender differences in neurological symptoms of long COVID. Methods: We included studies on female individuals who presented with specific neurological symptoms at least 12 weeks after confirmed COVID-19 diagnosis from PubMed, Central, Scopus, and Web of Science. The search limit was put for after January 2020 until June 15, 2024. We excluded studies that did not provide sex-specific outcome data, those not in English, case reports, case series, and review articles Results: A total of 5,632 eligible articles were identified. This article provides relevant information from 12 studies involving 6,849 patients, of which 3,414 were female. The sample size ranged from 70 to 2,856, with a maximum follow-up period of 18 months. The earliest publication date was September 16, 2021, while the latest was June 11, 2024. The following neurological symptoms had a significant difference in the risk ratio (RR) for female gender: fatigue RR 1.40 (95% confidence interval [CI]: 1.22–1.60, p < 0.001), headache RR 1.37 (95% CI: 1.12–1.67, p = 0.002), brain-fog RR 1.38 (95% CI 1.08–1.76, p = 0.011) depression RR 1.49 (95% CI: 1.2–1.86, p < 0.001), and anosmia RR 1.61 (95% CI: 1.36–1.90, p < 0.001). High heterogenicity was found for fatigue, brain fog, and anxiety due to the diverse methodologies employed in the studies. Conclusion: Our findings suggest that women are at a higher risk for long-COVID neurological symptoms, including fatigue, headaches, brain fog, depression, and anosmia, compared to men. The prevalence of these symptoms decreases after 1 year, based on limited data from the small number of studies available beyond this period.

Individuals suffering from long-COVID can present with “brain fog”, which is characterized by a range of cognitive impairments, such as confusion, short-term memory loss, and difficulty concentrating. To date, several potential interventions for brain fog have been considered. Notably, no systematic review has comprehensively discussed the impact of each intervention type on brain fog symptoms. We included studies on adult (aged > 18 years) individuals with proven long- COVID brain-fog symptoms from PubMed, MEDLINE, Central, Scopus, and Embase. A search limit was set for articles published between 01/2020 and 31/12/2023. We excluded studies lacking an objective assessment of brain fog symptoms and patients with preexisting neurological diseases that affected cognition before COVID-19 infection. This review provided relevant information from 17 studies. The rehabilitation studies utilized diverse approaches, leading to a range of outcomes in terms of the effectiveness of the interventions. Six studies described noninvasive brain stimulation, and all showed improvement in cognitive ability. Three studies described hyperbaric oxygen therapy, all of which showed improvements in cognitive assessment tests and brain perfusion. Two studies showed that the use of Palmitoylethanolamide and Luteolin (PEA-LUT) improved cognitive impairment. Noninvasive brain stimulation and hyperbaric oxygen therapy showed promising results in the treatment of brain fog symptoms caused by long-COVID, with improved perfusion and cortical excitability. Furthermore, both rehabilitation strategies and PEA-LUT administration have been associated with improvements in symptoms of brain fog. Future studies should explore combinations of interventions and include longer follow-up periods to assess the long-term effects of these treatments.

Background and objectives: Accurate interpretation of electrodiagnostic (EDX) studies is essential for the diagnosis and management of neuromuscular disorders. Artificial intelligence (AI) based tools may improve consistency and quality of EDX reporting and reduce workload. The aim of this study is to evaluate the performance of an AI‐assisted, multi‐agent framework (INSPIRE) in comparison with standard physician interpretation in a randomized controlled trial (RCT). Methods: We prospectively enrolled 200 patients (out of 363 assessed for eligibility) referred for EDX. Patients were randomly assigned to either a control group (physician‐only interpretation) or an intervention group (physician–AI interpretation). Three board‐certified physicians, rotated across both arms. In the intervention group, an AI‐generated preliminary report was combined with the physician’s independent findings (human–AI integration). The primary outcome was EDX report quality using score we developed named AI-Generated EMG Report Score (AIGERS; score range, 0–1, with higher scores indicating more accurate or complete reports). Secondary outcomes included a physician‐reported AI integration rating score (PAIR) and a compliance survey evaluating ease of AI adoption. Results: Of the 200 enrolled patients, 100 were allocated to AI‐assisted interpretation and 100 to physician‐only reporting. While AI‐generated preliminary reports offered moderate consistency on the AIGERS metric, the integrated (physician–AI) approach did not significantly outperform physician‐only methods. Despite some anecdotal advantages such as efficiency in suggesting standardized terminology quantitatively, the AIGERS scores for physician–AI integration was nearly the same as those in the physician‐only arm and did not reach statistical significance (p > 0.05 for all comparisons). Physicians reported variable acceptance of AI suggestions, expressing concerns about the interpretability of AI outputs and workflow interruptions. Physician–AI collaboration scores showed moderate trust in the AI’s suggestions (mean 3.7/5) but rated efficiency (2.0/5), ease of use (1.7/5), and workload reduction (1.7/5) as poor, indicating usability challenges and workflow interruptions. Discussion: In this single‐center, randomized trial, AI‐assisted EDX interpretation did not demonstrate a significant advantage over conventional physician‐only interpretation. Nevertheless, the AI framework may help reduce workload and documentation burdens by handling simpler, routine EDX tests freeing physicians to focus on more complex cases that require greater expertise. Trial registration: ClinicalTrials.gov Identifier: NCT06902675