JMIR Medical Informatics最新文献

Unlabelled: Retrieval-augmented generation (RAG) systems have emerged as a powerful technique to enhance the capabilities of large language models by enabling them to access external, up-to-date knowledge in real time, and RAG systems are being increasingly adopted by researchers in the medical field. In this viewpoint article, we explore the ethical imperatives for implementing RAG systems in clinical nursing environments, with particular attention to how these technologies affect patient care quality and safety. The purpose of this paper is to examine the ethical risks introduced by RAG-enhanced large language models in clinical nursing and to propose strategic guidelines for their responsible implementation. Key considerations include ensuring accuracy, fairness, transparency, and accountability, as well as maintaining essential human oversight, as discussed through a structured analysis. We argue that robust data governance, explainable artificial intelligence (AI) techniques, and continuous monitoring are critical components of a responsible RAG implementation strategy. Ultimately, realizing the benefits of RAG while mitigating ethical concerns requires sustained collaboration among health care professionals, AI developers, and policymakers, fostering a future where AI supports patient safety, reduces disparities, and improves the quality of nursing care.

Background: Considering sex and gender improves research quality, innovation, and social equity, while ignoring them leads to inaccuracies and inefficiency in study results. Despite increasing attention on sex- and gender-sensitive medicine, challenges remain with accurately representing gender due to its dynamic and context-specific nature.

Objective: This work aims to contribute to the implementation of a standard for collecting and assessing gender-specific data in German university hospitals and associated research facilities.

Methods: We carried out a review to identify and categorize state-of-the-art gender scores. We systematically assessed 22 publications regarding the applicability and practicability of their proposed gender scores. Specifically, we evaluated the use of these gender scores on German research data from routine clinical practice, using the Medical Informatics Initiative core dataset (MII CDS).

Results: Different methods for assessing gender have been proposed, but no standardized and validated gender score is available for health research. Most gender scores target epidemiological or public health research where questions about social aspects and life habits are already part of the questionnaires. However, it is challenging to apply concepts for gender scoring on clinical data. The MII CDS, for example, lacks all variables currently being recorded in gender scores. Although some of the required variables are indeed present in routine clinical data, they need to become part of the MII CDS.

Conclusions: To enable gender-specific retrospective analysis of routine clinical data, we recommend updating and expanding the MII CDS by including more gender-relevant information. For this purpose, we provide concrete action steps on how gender-related variables can be captured in routine clinical practice and represented in a machine-readable way.

Background: Minimally invasive posterior lumbar interbody fusion (MIS-PLIF) is commonly performed to treat degenerative lumbar spinal conditions. Patients of advanced age often present with multiple comorbidities and reduced physiological reserves, influencing surgical risks and recovery. The growing aging population has led to a rising demand for care for older adults, posing significant challenges for health care systems worldwide.

Objective: This study aimed to identify the associations between different age groups and MIS-PLIF outcomes.

Methods: This study retrospectively analyzed data from the United States Nationwide Inpatient Sample collected between 2016 and 2020. Patients aged ≥60 years who underwent MIS-PLIF were eligible for inclusion in this study. Patients were categorized into age groups (60-69, 70-79, and ≥80 y). Logistic and linear regressions were used to determine the associations between the study variables and outcomes, including in-hospital mortality, complications, nonroutine discharge, and length of stay.

Results: A total of 785 patients aged ≥60 (mean age 69.4, SD 0.2) years who underwent MIS-PLIF were included in the analysis, and 18.7% (147/785) experienced at least one complication. After adjustment, compared with patients aged 60 to 69 years, the risk of nonroutine discharge was significantly increased in patients aged 70 to 79 years (adjusted odds ratio 2.33, 95% CI 1.57-3.46; P<.001) and ≥80 years (adjusted odds ratio 4.79, 95% CI 2.64-8.67; P<.001). No significant differences in the risk of complications or length of hospital stay were observed across the age groups.

Conclusions: In older patients undergoing MIS-PLIF, advanced age is an independent predictor of nonroutine discharge. Furthermore, our findings suggest that age alone is not an independent risk factor for complications or extended hospital stays among older patients. These findings underscore that MIS-PLIF is a viable option for older patients, for whom extra attention may still be needed for postoperative care. Implementing age-stratified management for older patients undergoing MIS-PLIF may have important clinical policy implications.

Background: Large language models (LLMs), such as GPT-3.5 and GPT-4 (OpenAI), have been transforming virtual patient systems in medical education by providing scalable and cost-effective alternatives to standardized patients. However, systematic evaluations of their performance, particularly for multimorbidity scenarios involving multiple coexisting diseases, are still limited.

Objective: This systematic review aimed to evaluate LLM-based virtual patient systems for medical history-taking, addressing four research questions: (1) simulated patient types and disease scope, (2) performance-enhancing techniques, (3) experimental designs and evaluation metrics, and (4) dataset characteristics and availability.

Methods: Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020, 9 databases were searched (January 1, 2020, to August 18, 2025). Nontransformer LLMs and non-history-taking tasks were excluded. Multidimensional quality and bias assessments were conducted.

Results: A total of 39 studies were included, screened by one computer science researcher under supervision. LLM-based virtual patient systems mainly simulated internal medicine and mental health disorders, with many addressing distinct single disease types but few covering multimorbidity or rare conditions. Techniques like role-based prompts, few-shot learning, multiagent frameworks, knowledge graph (KG) integration (top-k accuracy 16.02%), and fine-tuning enhanced dialogue and diagnostic accuracy. Multimodal inputs (eg, speech and imaging) improved immersion and realism. Evaluations, typically involving 10-50 students and 3-10 experts, demonstrated strong performance (top-k accuracy: 0.45-0.98, hallucination rate: 0.31%-5%, System Usability Scale [SUS] ≥80). However, small samples, inconsistent metrics, and limited controls restricted generalizability. Common datasets such as MIMIC-III (Medical Information Mart for Intensive Care-III) exhibited intensive care unit (ICU) bias and lacked diversity, affecting reproducibility and external validity.

Conclusions: Included studies showed moderate risk of bias, inconsistent metrics, small cohorts, and limited dataset transparency. LLM-based virtual patient systems excel in simulating multiple disease types but lack multimorbidity patient representation. KGs improve top-k accuracy and support structured disease representation and reasoning. Future research should prioritize hybrid KG-chain-of-thought architectures integrated with open-source KGs (eg, UMLS [Unified Medical Language System] and SNOMED-CT [Systematized Nomenclature of Medicine - Clinical Terms]), parameter-efficient fine-tuning, dialogue compression, multimodal LLMs, standardized metrics, larger cohorts, and open-access multimodal datasets to further enhance realism, diagnostic accuracy, fairness, and educational utility.

Background: Accurately predicting left ventricular ejection fraction (LVEF) recovery after percutaneous coronary intervention (PCI) in patients with chronic coronary syndrome (CCS) is crucial for clinical decision-making.

Objective: This study aimed to develop and compare multiple machine learning (ML) models to predict LVEF recovery and identify key contributing features.

Methods: We retrospectively analyzed 520 patients with CCS from the Clinical Deep Data Accumulation System database. Patients were categorized into 4 binary classification tasks based on baseline LVEF (≥50% or <50%) and degree of recovery: (1) good recovery, defined as an LVEF increase of >10% compared with ≤0%; and (2) normal recovery, defined as an LVEF increase of 0% to 10% compared with ≤0%. For each task, 3 feature selection strategies (all features, least absolute shrinkage and selection operator [LASSO] regression, and recursive feature elimination [RFE]) were combined with 4 ML algorithms (extreme gradient boosting [XGBoost], categorical boosting, light gradient boosting machine, and random forest), resulting in 48 models. Models were evaluated using 10-fold cross-validation and assessed by the area under the curve (AUC), decision curve analysis, and calibration plots.

Results: The highest AUCs were achieved by RFE combined with XGBoost (AUC=0.93) for preserved LVEF with good recovery, LASSO combined with XGBoost (AUC=0.79) for preserved LVEF with normal recovery, LASSO combined with XGBoost (AUC=0.88) for reduced LVEF with good recovery, and RFE combined with XGBoost (AUC=0.84) for reduced LVEF with normal recovery. Shapley Additive Explanation analysis identified uric acid, platelets, hematocrit, brain natriuretic peptide, glycated hemoglobin, glucose, creatinine, baseline LVEF, left ventricular end-diastolic internal diameter, heart rate, R wave amplitude in V5, and R wave amplitude in V6 as important predictive factors of LVEF recovery.

Conclusions: ML models incorporating feature selection strategies demonstrated strong predictive performance for LVEF recovery after PCI. These interpretable models may support clinical decision-making and can improve the management of patients with CCS after PCI.