International Journal of Medical Informatics最新文献

{"title":"A data-driven scoring framework for personalized employee health check-ups: Integrating historical laboratory trends and evidence-based prevalence","authors":"Saranya Thongsawaeng ,&nbsp;Siwapol Techaratsami ,&nbsp;Jidapa Hanvoravongchai ,&nbsp;Napatsorn Thewaran ,&nbsp;Piyawat Kantagowit ,&nbsp;Krit Pongpirul","doi":"10.1016/j.ijmedinf.2025.105974","DOIUrl":"10.1016/j.ijmedinf.2025.105974","url":null,"abstract":"<div><h3>Introduction</h3><div>Rising healthcare costs and growing demand for personalized preventive care have highlighted the need for data-driven approaches to optimize health check-ups, particularly in corporate settings. This study presents a scoring-based platform designed to prioritize laboratory tests for individual employees by integrating historical health data with condition prevalence, aiming to improve the precision and efficiency of routine health assessments.</div></div><div><h3>Methods</h3><div>The platform integrates two main components. First, a prevalence model was developed through a systematic review and <em>meta</em>-analysis of 266 studies (from an initial 28,558), providing prevalence estimates for various conditions detectable through laboratory testing. Second, the<!--> <!-->Individual Historical Lab Score (IHLS)<!--> <!-->model was built using employee health records. IHLS combines three metrics: (1) prevalence scores for each test, (2) abnormality scores based on current lab values relative to reference ranges, and (3) trend scores derived from linear trend estimation using least squares error across prior years. These components are heuristically combined to rank check-up items for each individual.</div></div><div><h3>Results</h3><div>The model was evaluated using six years (2016–2022) of longitudinal health check-up data from 3,198 employees across seven business entities (7,518 total records; mean follow-up: 3.4 years; mean age: 39.3 ± 9.6 years; 29.3 % male). Model performance was assessed using Receiver Operating Characteristic (ROC) curve analysis. IHLS achieved an Area Under the Curve (AUC) of 0.82, outperforming the prevalence-only model (AUC = 0.77) and random baseline (AUC = 0.50).</div></div><div><h3>Conclusions</h3><div>This prototype platform demonstrates the potential informatics-driven scoring systems to enhance personalized health check-up recommendations. By combining individual lab history with population-based prevalence data, the model supports early risk identification and cost-effective screening trategies, offering practical applications in workplace wellness programs and scalable integration into broader health systems.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"202 ","pages":"Article 105974"},"PeriodicalIF":3.7,"publicationDate":"2025-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144106378","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Some key challenges in providing online professionalism education for medical students and residents","authors":"Hongnan Ye","doi":"10.1016/j.ijmedinf.2025.105972","DOIUrl":"10.1016/j.ijmedinf.2025.105972","url":null,"abstract":"","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"202 ","pages":"Article 105972"},"PeriodicalIF":3.7,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144089877","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The accuracy of Machine learning in the prediction and diagnosis of diabetic kidney Disease: A systematic review and Meta-Analysis","authors":"Changmao Dai,&nbsp;Xiaolan Sun,&nbsp;Jia Xu,&nbsp;Maojun Chen,&nbsp;Wei Chen,&nbsp;Xueping Li","doi":"10.1016/j.ijmedinf.2025.105975","DOIUrl":"10.1016/j.ijmedinf.2025.105975","url":null,"abstract":"<div><h3>Purpose</h3><div>Machine learning (ML) has gained attention in diabetes management, particularly for predicting and diagnosing diabetic kidney disease (DKD). However, systematic evidence on its performance remains limited. This study evaluates the predictive and diagnostic accuracy of ML in DKD to support the development of tailored prevention strategies and non-invasive diagnostic tools.</div></div><div><h3>Methods</h3><div>A systematic search of PubMed, Embase, Web of Science, and Cochrane (up to April 14, 2024) identified relevant studies. Risk of bias was assessed using tools for predictive models, and <em>meta</em>-analysis included subgroup analyses based on task type, dataset, and model type.</div></div><div><h3>Results</h3><div>A total of 34 studies were included, with 19 on DKD risk prediction and 15 on diagnosis. For prediction, the pooled c-index was 0.81 (95% CI 0.79–0.83), sensitivity 0.81 (95% CI 0.74–0.86), and specificity 0.82 (95% CI 0.73–0.89). For diagnosis, the pooled c-index was 0.81 (95% CI 0.79–0.83), sensitivity 0.81 (95% CI 0.78–0.84), and specificity 0.75 (95% CI 0.72–0.79).</div></div><div><h3>Conclusions</h3><div>ML shows promising accuracy in DKD prediction and diagnosis, offering a viable tool for early screening and risk assessment.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"202 ","pages":"Article 105975"},"PeriodicalIF":3.7,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144105130","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Investigating the impact of sociodemographic factors on training efficacy and its correlation with technology usability in older adults: Lessons learned in Italian and Murcian pilots","authors":"Laura Fiorini ,&nbsp;Jasmine Pani ,&nbsp;Erika Rovini ,&nbsp;María-Victoria Bueno Delgado ,&nbsp;Salvador Pérez-Martos ,&nbsp;Sergio Russo ,&nbsp;Letizia Lorusso ,&nbsp;Giuseppina Iannacone ,&nbsp;Grazia D’Onofrio ,&nbsp;Francesco Giuliani ,&nbsp;Lara Toccafondi ,&nbsp;Gianna Vignani ,&nbsp;Filippo Cavallo","doi":"10.1016/j.ijmedinf.2025.105973","DOIUrl":"10.1016/j.ijmedinf.2025.105973","url":null,"abstract":"<div><h3>Introduction</h3><div>Technology training supports technology adoption among older adults. However, guidelines and insights into personal and sociodemographic factors affecting its effectiveness are lacking. This study explores how these factors influence training effectiveness in older adults and its impact on technology usability.</div></div><div><h3>Methods</h3><div>This paper focuses on two pilot sites of the Pharaon project that implemented similar health monitoring scenarios. A total of 114 older adults were recruited and trained on monitoring technologies following which they filled in sociodemographic and usability questionnaires.</div></div><div><h3>Results</h3><div>Our findings indicate that age, digital literacy, educational attainment, and perceived loneliness significantly affect training evaluation, while quality of life and gender do not show a significant impact. Training efficacy was also found to be connected to system usability (all p &lt; 0.005). Furthermore, the experience of professionals involved with providing training to older adults was elaborated highlighting the importance of tailored training approaches and continuous support mechanisms to enhance technology adoption among older populations.</div></div><div><h3>Discussion</h3><div>The results showed that training programs aimed at enhancing usability should consider tailoring the content to the user, as there are personal factors which can influence how the training is received. Finally, the results provide actionable recommendations for optimizing training protocols to facilitate the integration of digital health solutions across diverse environments.</div></div><div><h3>Conclusions</h3><div>The findings highlight the need for standardized yet adaptable training guidelines that address individual differences, offering practical direction for future implementations and policies to support long-term technology adoption in older adults.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"202 ","pages":"Article 105973"},"PeriodicalIF":3.7,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144116863","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Optimising design of clinical decision support systems and implementation strategies to improve radiological imaging appropriateness – A qualitative study in the emergency department","authors":"Yi Xiang Tay ,&nbsp;Jeremy C.P. Wee ,&nbsp;Marcus E.H. Ong ,&nbsp;Shane J. Foley ,&nbsp;Robert Chun Chen ,&nbsp;Lai Peng Chan ,&nbsp;Ronan Killeen ,&nbsp;Ivan S.Y. Chua ,&nbsp;Jonathan P. McNulty","doi":"10.1016/j.ijmedinf.2025.105966","DOIUrl":"10.1016/j.ijmedinf.2025.105966","url":null,"abstract":"<div><h3>Background</h3><div>There is existing evidence demonstrating the potential effectiveness of clinical decision support systems (CDSS) for appropriate imaging. However, the uptake of this technology and other similar evidence-based interventions has been highly variable. We explored the factors that influence clinicians’ acceptance of CDSS, while also examining strategies that can further support and promote appropriate imaging in the emergency department (ED), as perceived by ED physicians.</div></div><div><h3>Methods</h3><div>We adhered to established focus group (FG) frameworks and conducted interviews with a convenience sample of ED physicians in the largest acute tertiary hospital in Singapore. A semi-structured interview guide was developed through collaborative discussions with colleagues in radiology and emergency medicine, including radiographers, experienced ED physicians, and radiologists. Sessions were audio-recorded, transcribed, and analysed. We followed Braun and Clarke’s reflexive thematic analysis for data engagement, coding, and theme development.</div></div><div><h3>Results</h3><div>We interviewed 16 ED physicians (9 specialists and 7 non-specialists) in separate groups over five virtual sessions. We constructed two salient themes: (1) the features and design of a CDSS are crucial for increasing the uptake of appropriate imaging, and (2) strategies to support appropriate imaging in emergency medicine. Participants described personal experiences and expressed strategies and solutions to some of the challenges faced. Strategies exist at the human, technological, and organisational levels to promote appropriate imaging in the ED.</div></div><div><h3>Conclusion</h3><div>Participants expressed their expectations regarding the design and features of CDSS as end users. Many strategies to optimise the implementation of CDSS in the ED for appropriate imaging exist and can encourage acceptance by ED physicians.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"202 ","pages":"Article 105966"},"PeriodicalIF":3.7,"publicationDate":"2025-05-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144105133","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A scoping review of digital solutions in diabetes outpatient care: Functionalities and outcomes","authors":"Wenyong Wang ,&nbsp;Mahnaz Samadbeik ,&nbsp;Gaurav Puri ,&nbsp;Donald S.A. McLeod ,&nbsp;Elton Lobo ,&nbsp;Tuan Duong ,&nbsp;Jennifer Nguyen ,&nbsp;Mutian Ding ,&nbsp;Clair Sullivan","doi":"10.1016/j.ijmedinf.2025.105967","DOIUrl":"10.1016/j.ijmedinf.2025.105967","url":null,"abstract":"<div><h3>Background</h3><div>Digital interventions are increasingly used in outpatient diabetes care to address growing healthcare demands and workforce limitations. This study investigates the functionalities of digital solutions and their impact on Quadruple Aim outcomes: enhancing population health, improving patient experience, supporting clinician well-being, and reducing healthcare costs.</div></div><div><h3>Methods</h3><div>We followed Joanna Briggs Institute guidelines, searching PubMed, Embase, Cochrane, Scopus, and Web of Science (January 2019–February 2024). Included studies reported digital diabetes interventions with outcomes directly relevant to the Quadruple Aim. Each intervention was mapped to a digital solution horizon: Horizon 1 involves foundational digital workflows; Horizon 2 leverages real-time data to create analytics; Horizon 3 encompasses transformative uses, such as predictive analytics.</div></div><div><h3>Results</h3><div>We identified 4,397 articles with 56 meeting the inclusion criteria. Interventions included telehealth (n = 15), mobile health (mHealth) (n = 20), combined telehealth and mHealth (n = 14), robotics (n = 1), electronic medical records (n = 1), and artificial intelligence (n = 5). Most interventions (n = 51) were categorised as Horizon 1, with 10 adopting Horizon 2, 5 using Horizon 3, and 10 spanning multiple horizons. Regarding Quadruple Aim outcomes, 44 studies addressed population health (41 positive), 31 targeted patient experience (29 positive), 4 focused on clinician well-being (3 positive), and 6 on cost reduction (4 positive).</div></div><div><h3>Conclusion</h3><div>Digital solutions have demonstrated measurable benefits, particularly in population health and patient experience. Most interventions remain at Horizon 1. Advancing these digital solutions to Horizon 2 and 3 is essential for system-wide transformation. Future research should include cost efficiency and clinician experience alongside evaluations of population health and patient experience.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"202 ","pages":"Article 105967"},"PeriodicalIF":3.7,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143937215","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Getting to the information clinicians need quickly: Pharmacist evaluation of DynaMed and micromedex with watson (DynaMedex) using real-world questions","authors":"Heba H. Edrees ,&nbsp;Diane L. Seger ,&nbsp;Mary G. Amato ,&nbsp;Ania Syrowatka ,&nbsp;Pamela M. Garabedian ,&nbsp;Sevan Dulgarian ,&nbsp;Petra Schultz ,&nbsp;Gretchen Purcell Jackson ,&nbsp;David W. Bates","doi":"10.1016/j.ijmedinf.2025.105965","DOIUrl":"10.1016/j.ijmedinf.2025.105965","url":null,"abstract":"<div><h3>Introduction</h3><div>Digital point-of-care information resources are frequently used by clinicians to answer clinical questions. An evidence-based disease management database (DynaMed) was merged with a pharmaceutical knowledge base (Micromedex). We evaluated the ability of the combined solution, DynaMedex, to answer clinical questions.</div></div><div><h3>Methods</h3><div>Real-world questions were used for testing and were categorized by information type and specialty area. Two pharmacists independently performed 600 searches for 300 questions, using keyword search and Watson Assistant (WA). Search results were evaluated based on whether information was found (yes, no), relevance to the question (relevant, not relevant), difficulty in finding the answer (easy, medium, hard), and quality of the evidence (good, fair, poor).</div></div><div><h3>Results</h3><div>An answer was found 86.3% of the time using keyword search and 81.0% of the time using WA. In keyword searches, 86.0% of answers were considered relevant and 74.5% in WA. Most answers were easy to find (78.7% in keyword search, 94.4% in WA). The quality of evidence for answers was good, fair, or poor in 62.7%, 36.4%, and 0.9% for keyword search and 50.3%, 47.8%, and 1.9% for WA.</div></div><div><h3>Conclusion</h3><div>Pharmacists found answers to most clinical questions easily with good quality, evidence-based information and a high agreement rate. This resource could be further improved by recognizing different search terms, standardizing the location of drug and disease information in appropriate sections, providing citations, if available, with the highest quality evidence, and including access to content types that haven’t been incorporated.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"202 ","pages":"Article 105965"},"PeriodicalIF":3.7,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143948051","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evaluation of insulin sensitivity temporal prediction by using quantile regression combined with neural network model","authors":"Omer S. Alkhafaf ,&nbsp;J.Geoffrey Chase ,&nbsp;Balázs Benyó","doi":"10.1016/j.ijmedinf.2025.105964","DOIUrl":"10.1016/j.ijmedinf.2025.105964","url":null,"abstract":"<div><h3>Background</h3><div>Stress-induced hyperglycemia, a pathologically high blood glucose level, is a frequent complication in intensive care units. Blood glucose (BG) level control is crucial but challenging due to patient variability. The Stochastic TARgeted (STAR) protocol is clinically used for blood glucose control, which uses the current and predicted future patient insulin sensitivity (SI) parameter to assess BG outcomes of alternative treatment options.</div></div><div><h3>Objective</h3><div>Neural network (NN) models using quantile regression (QR) have enhanced SI prediction performance. However, remains a challenge in determining the optimal NN configuration to best predict SI. This study aims to find the NN configuration yielding the highest prediction accuracy to improve the STAR protocol and explores the behaviour of the QR method in predicting the percentiles of a non-Gaussian multi-mode distribution of a physiological parameter.</div></div><div><h3>Method</h3><div>Alternative NN architectures combined with QR were implemented and trained on a large dataset comprising 1,897 patients collected between 2011 and 2023 using five-fold cross-validation ensuring model robustness. Prediction performance was evaluated among NN configurations and compared using case-specific metrics across the global SI domain as well as within subdomains to analyse the models’ local performance.</div></div><div><h3>Results</h3><div>Outcomes indicate QR applied to simpler NN, consisting of one-hidden layer with four neurons, achieves a best prediction performance at a minimum network size. Using more complex NN did not improve the prediction performance significantly. However, at long prediction horizons, no compact network demonstrated improved outcomes. A more general methodological outcome of the study is that QR-based prediction does not need to be combined with complex NN to achieve the best prediction performance.</div></div><div><h3>Conclusion</h3><div>The QR-based method was found to be appropriate for the SI prediction problem in short-term predictions which may improve the STAR protocol’s clinical outcomes. Overall, the study provides a generalisable, empirical approach to network configuration optimisation for similar problems.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"202 ","pages":"Article 105964"},"PeriodicalIF":3.7,"publicationDate":"2025-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143937317","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Nutritional and lifestyle predictors of rectal bleeding in functional constipation: A machine learning approach","authors":"Joyeta Ghosh ,&nbsp;Jyoti Taneja ,&nbsp;Ravi Kant","doi":"10.1016/j.ijmedinf.2025.105963","DOIUrl":"10.1016/j.ijmedinf.2025.105963","url":null,"abstract":"<div><h3>Background</h3><div>Rectal bleeding among young adults is an increasingly common clinical concern often linked with chronic constipation and unhealthy lifestyle habits. Early identification of at-risk individuals through machine learning models-based approach may help in prevention and targeted intervention.</div></div><div><h3>Objectives</h3><div>We aim to identify dietary and lifestyle risk factors for rectal bleeding and to develop machine learning-based models for risk prediction.</div></div><div><h3>Methods</h3><div>A descriptive observational study was conducted on 875 Indian college going participants. A structured questionnaire assessed fiber intake, physical activity, constipation symptoms, and body mass index (BMI). Multiple machine learning algorithms were evaluated, and their performance was assessed using accuracy and area under the receiver operating characteristic curve (ROC-AUC).</div></div><div><h3>Results</h3><div>Low intake of boiled vegetables or oatmeal (&lt;50 g/day) was associated with a 43.92 % bleeding rate (p &lt; 0.001). Participants consuming inadequate whole grains (&gt;25 g/day) showed a 44.81 % bleeding rate. Overweight or obese individuals exhibited a significantly higher bleeding incidence (12.26 %) than those with normal BMI (5.55 %; p = 0.008). The KNeighbors Classifier showed the highest accuracy (98.86 %) and ROC-AUC (0.994). Variables related to symptoms had greater predictive importance than those related to lifestyle.</div></div><div><h3>Conclusions</h3><div>The findings support the role of dietary fiber and BMI in the development of rectal bleeding in constipated individuals. The predictive models demonstrate strong potential for identifying at-risk individuals and is considered a simple and useful tool for predicting rectal bleeding in functional constipation, suggesting preventive health strategies and dietary modifications. This novel algorithm might enable clinicians to perform personalized dietary strategies with improved clinical outcomes. Further validation across larger and more diverse populations is recommended.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"201 ","pages":"Article 105963"},"PeriodicalIF":3.7,"publicationDate":"2025-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143923992","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Translating the machine; An assessment of clinician understanding of ophthalmological artificial intelligence outputs","authors":"Oskar Wysocki ,&nbsp;Sammie Mak ,&nbsp;Hannah Frost ,&nbsp;Donna M. Graham ,&nbsp;Dónal Landers ,&nbsp;Tariq Aslam","doi":"10.1016/j.ijmedinf.2025.105958","DOIUrl":"10.1016/j.ijmedinf.2025.105958","url":null,"abstract":"<div><h3>Introduction</h3><div>Advances in artificial intelligence offer the promise of automated analysis of optical coherence tomography (OCT) scans to detect ocular complications from anticancer drug therapy. To explore how such AI outputs are interpreted in clinical settings, we conducted a survey-based interview study with 27 clinicians —comprising 10 ophthalmic specialists, 10 ophthalmic practitioners, and 7 oncologists. Participants were first introduced to core AI concepts and realistic clinical scenarios, then asked to assess AI-generated OCT analyses using standardized Likert-scale questions, allowing us to gauge their understanding, trust, and readiness to integrate AI into practice.</div></div><div><h3>Methods</h3><div>We developed a questionnaire through literature review and consultations with ophthalmologists, computer scientists, and AI researchers. A single investigator interviewed 27 clinicians across three specialties and transcribed their responses. Data were summarized as medians (ranges) and compared with Mann–Whitney U tests (α = 0.05).</div></div><div><h3>Results</h3><div>We noted important differences in the impact of various explainability methods on trust, depending on the clinical or AI scenario nature and the staff expertise. Explanations of AI outputs increased trust in the AI algorithm when outputs simply reflected ground truth expert opinion. When clinical scenarios were complex with incorrect AI outcomes, a mixed response to explainability led to correctly reduced trust in experienced clinicians but mixed feedback amongst less experienced clinicians. All clinicians had a general consensus on lack of current knowledge in interacting with AI and desire more training.</div></div><div><h3>Conclusions</h3><div>Clinicians’ trust in AI algorithms are affected by explainability methods and factors, including AI’s performance, personal judgments and clinical experience. The development of clinical AI systems should consider the above and these responses ideally be factored into real-world assessments. Use of this study’s findings could help improve the real world validity of medical AI systems by enhancing the human–computer interactions, with preferred explainability techniques tailored to specific situations.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"201 ","pages":"Article 105958"},"PeriodicalIF":3.7,"publicationDate":"2025-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143932033","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}