BMC Medical Informatics and Decision Making最新文献

Background: Viral hepatitis is a major global public health problem that affects millions of people; therefore, accurate and accessible information is essential for both the general public and non-specialist healthcare providers to correctly understand, prevent, and manage the disease. This study evaluated four large language models (LLMs)-Gemini-2.0, Claude-3.5-sonnet, ChatGPT-4.5, and ChatGPT-4-and compared their responses to viral hepatitis-related questions to assess differences in performance across models.

Methods: This comparative evaluation study, conducted at Nanjing Drum Tower Hospital from March to April 2025, examined 52 questions pertaining to viral hepatitis. Four large language models were assessed based on their responses to these 52 questions which encompassed four domains: concepts, risk factors, diagnosis, and prevention and treatment. Initial evaluation used a three-point scale of good, borderline, and poor. Further evaluation criteria included relevance, comprehensiveness, accuracy, safety, and readability, with each response scored on a scale of 1 to 5.

Results: ChatGPT-4.5 achieved the highest performance, with 89.1% of its responses rated as good, significantly outperforming Claude-3.5-sonnet (71.15% good), Gemini-2.0 (62.82% good), and ChatGPT-4 (50.64% good). Statistical analysis confirmed superior performance of ChatGPT-4.5 in all evaluated dimensions. Consistently, ChatGPT-4.5 scored the highest across all five criteria: relevance, comprehensiveness, accuracy, safety, and readability.

Conclusions: ChatGPT-4.5 demonstrates superior performance in addressing viral hepatitis queries compared to other three models. Its high reliability makes it a valuable tool for patients and medical professionals not specializing in viral hepatitis by improving information accessibility.

Background: Rare disease diagnosis often involves complex, lengthy, and costly procedures. Traditional cost-effectiveness analyses typically rely on static diagnostic workflow models that apply uniform diagnostic strategies across heterogeneous patient populations. With recent advancements in artificial intelligence (AI) and a growing emphasis on personalized medicine, there is a pressing need for dynamic frameworks that assess diagnostic cost-effectiveness at the individual patient level.

Methods: We introduce the PRICE analysis framework, a novel, tree-based model designed to evaluate the cost-effectiveness of diagnostic strategies, accommodating both expert-alone and AI-delegated decision-making modes. The model computes the expected cost of a diagnostic process via a back-propagation algorithm and quantifies effectiveness through a utility-based approach (i.e., Quality Adjusted Life Years). Parameters such as disease prevalence, test costs, test performance metrics, and turnaround time are incorporated to enable individualized assessments.

Results: We demonstrat the utility of this novel framework in a proof-of-concept study by evaluating four diagnostic strategies for developmental delay (DD) and multiple congenital anomalies (MCA). The results highlight how PRICE can support personalized decision-making by modeling outcomes under varying parameters such as cost, prevalence, yield, and AI accuracy. To better visualize and interpret this framework, we developed an interactive web-based tool to demonstrate how to build PRICE pathways and conduct cost-effectiveness analysis in real time.

Conclusion: PRICE is a novel cost-effective analysis framework that captures the sequential and recursive nature of real-world diagnostic workflows, with the ability to be extended to future AI-integrated clinical practice. It enables personalized evaluations of diagnostic strategies from both economic and clinical perspectives, promoting more informed and individualized decision-making for rare disease diagnosis.

Background: As populations age, policymakers and researchers are investigating innovative systems to support healthy brain ageing and prevent dementia to reduce the personal, community and fiscal burden of age-related disease. Primary care is a routinely accessed part of global healthcare systems and could act as an avenue to introduce platforms to support healthy ageing. However, research suggests that many primary care professionals feel underprepared, lack confidence, knowledge, and resources for age-related practice. This exploratory study aims to explore the current state of healthy brain ageing management in primary care and investigate how digital technologies, particularly brain health dashboards, could be implemented and used to increase clinician and patient outcomes.

Methods: Healthcare professionals (N = 9) with a range of primary healthcare expertise (general practice, specialist, nursing, health promotion and health administration) were interviewed on current practices used to support healthy ageing in older adults and use of digital technologies for brain health management. Semi-structured interviews were audio-recorded, transcribed, and analysed thematically.

Results: Three core themes were identified: current management practices, facilitators and barriers to adoption and general acceptability. Core facilitators such as IT prevalence and clinician openness and barriers, including resource allocation, technological issues, breakdown of clinician-patient rapport and skills acquisition necessary for usage were identified by the healthcare professionals.

Conclusions: This study provides initial findings on the barriers and facilitators to adopting technologies for brain health management in primary care, which could be used to create systems to support clinicians' management of brain ageing and improve patient outcomes.

Clinical trial number: Not applicable.

Background: Voice biomarkers hold potential for early cognitive disorder detection, but variations in recording conditions across different environments present challenges for accurate diagnosis using artificial intelligence (AI) models. This study aims to develop a robust, generalizable model for reliably diagnosing cognitive impairments across varied datasets.

Methods: We implemented a domain generalization approach using an adapted Deep Domain-Adversarial Image Generation (DDAIG) framework. This method transforms input data to reduce center-specific characteristics and emphasizes domain-invariant features, allowing the model to focus on cognitive impairment indicators.

Results: Before applying domain generalization, both cognitive impairment (CI) and center classification models achieved accuracies of 0.96. After implementing domain generalization, the CI classification accuracy decreased to 0.90, while the center classification model's accuracy dropped to 0.64. This reduction in the center classification metrics reflects the model's reduced dependence on center-specific features, indicating effective domain generalization.

Conclusion: The adapted DDAIG framework effectively reduced center-specific learning, enhancing the model's ability to generalize cognitive impairment classifications across different centers. These findings suggest the role of domain generalization in developing reliable AI diagnostic tools for cognitive disorder detection.