International Journal of Medical Informatics最新文献

{"title":"Investigating the role of large language models on questions about refractive surgery","authors":"Suleyman Demir","doi":"10.1016/j.ijmedinf.2025.105787","DOIUrl":"10.1016/j.ijmedinf.2025.105787","url":null,"abstract":"<div><h3>Background</h3><div>Large language models (LLMs) are becoming increasingly popular and are playing an important role in providing accurate clinical information to both patients and physicians. This study aimed to investigate the effectiveness of ChatGPT-4.0, Google Gemini, and Microsoft Copilot LLMs for responding to patient questions regarding refractive surgery.</div></div><div><h3>Methods</h3><div>The LLMs’ responses to 25 questions about refractive surgery, which are frequently asked by patients, were evaluated by two ophthalmologists using a 5-point Likert scale, with scores ranging from 1 to 5. Furthermore, the DISCERN scale was used to assess the reliability of the language models’ responses, whereas the Flesch Reading Ease and Flesch–Kincaid Grade Level indices were used to evaluate readability.</div></div><div><h3>Results</h3><div>Significant differences were found among all three LLMs in the Likert scores (p = 0.022). Pairwise comparisons revealed that ChatGPT-4.0′s Likert score was significantly higher than that of Microsoft Copilot, while no significant difference was found when compared to Google Gemini (p = 0.005 and p = 0.087, respectively). In terms of reliability, ChatGPT-4.0 stood out, receiving the highest DISCERN scores among the three LLMs. However, in terms of readability, ChatGPT-4.0 received the lowest score.</div></div><div><h3>Conclusions</h3><div>ChatGPT-4.0′s responses to inquiries regarding refractive surgery were more intricate for patients compared to other language models; however, the information provided was more dependable and accurate.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105787"},"PeriodicalIF":3.7,"publicationDate":"2025-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142959158","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":"Influencing factors: Unveiling patterns and reasons in telehealth care utilization and adoption/avoidance decisions","authors":"Hinpetch Daungsupawong ,&nbsp;Viroj Wiwanitkit","doi":"10.1016/j.ijmedinf.2025.105785","DOIUrl":"10.1016/j.ijmedinf.2025.105785","url":null,"abstract":"","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105785"},"PeriodicalIF":3.7,"publicationDate":"2025-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143140744","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 use of machine learning for the prediction of response to follow-up in spine registries","authors":"Alice Baroncini ,&nbsp;Andrea Campagner ,&nbsp;Federico Cabitza ,&nbsp;Francesco Langella ,&nbsp;Francesca Barile ,&nbsp;Pablo Bellosta-López ,&nbsp;Domenico Compagnone ,&nbsp;Riccardo Cecchinato ,&nbsp;Marco Damilano ,&nbsp;Andrea Redaelli ,&nbsp;Daniele Vanni ,&nbsp;Pedro Berjano","doi":"10.1016/j.ijmedinf.2024.105752","DOIUrl":"10.1016/j.ijmedinf.2024.105752","url":null,"abstract":"<div><h3>Background</h3><div>One of the main challenges in the maintenance of registries is to keep a high follow-up rate and a reliable strategy to limit dropout is currently lacking. Aim of this study was to utilize machine learning (ML) models to highlight the characteristics of patients who are most likely to drop out, and to evaluate the potential cost effectiveness of the implementation of a follow-up system based on the obtained data.</div></div><div><h3>Methods</h3><div>All patients recruited in the local spine surgery registry were included and demographic, peri- and postoperative data were collected. Five ML models were trained and evaluated for response to follow-up prediction. Explainable and Cautious AI were then implemented to increase the trustworthiness of the model. The efficacy and cost effectiveness of the current follow-up strategy (call everybody) were compared to a strategy based on the implemented model (call only patients with high dropout risk).</div></div><div><h3>Results</h3><div>Records from 4652 patients were available. The random forest (RF) outperformed all models in the prediction of response to follow-up. Among the considered variables, the ones that had the most weight were length of follow up, level of the main pathology and extent of surgery, SF-36 and BMI. Interpretable Decision Trees (IDT) and selective prediction models further increased the performance of the model. The cost reduction calculation predicted that implementing the developed ML model in the clinical practice would, over time, result in a reduction of costs by 31%, with only 2‰ missed calls.</div></div><div><h3>Conclusion</h3><div>ML models can effectively identify patients with high risk of dropout. The RF model outperformed all evaluated models, and was further improved with the use of Controllable AI. The application of ML to the follow-up strategy could reduce costs and limit missed responses.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105752"},"PeriodicalIF":3.7,"publicationDate":"2025-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142959159","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":"ATC-to-RxNorm mappings – A comparison between OHDSI Standardized Vocabularies and UMLS Metathesaurus","authors":"Rowdy de Groot ,&nbsp;Savannah Glaser ,&nbsp;Alexandra Kogan ,&nbsp;Stephanie Medlock ,&nbsp;Anna Alloni ,&nbsp;Matteo Gabetta ,&nbsp;Szymon Wilk ,&nbsp;Nicolette de Keizer ,&nbsp;Ronald Cornet","doi":"10.1016/j.ijmedinf.2024.105777","DOIUrl":"10.1016/j.ijmedinf.2024.105777","url":null,"abstract":"<div><h3>Introduction</h3><div>The World Health Organization global standard for representing drug data is the Anatomical Therapeutic Chemical (ATC) classification. However, it does not represent ingredients and other drug properties required by clinical decision support systems. A mapping to a terminology system that contains this information, like RxNorm, may help fill this gap. This work evaluates and compares the completeness of mappings from the chemical substance level (5th-level) ATC classes to RxNorm ingredient concepts in the OHDSI Standardized Vocabularies (OSV) and the Unified Medical Language System (UMLS) Metathesaurus.</div></div><div><h3>Methods</h3><div>To check the concordance between OSV and UMLS we compared the included contents of ATC and RxNorm not only in OSV and UMLS but also in BioPortal and the National Library of Medicine (NLM) repository. For each repository, we determined the number of 5th-level ATC concepts, RxNorm ingredient concepts, missing classes and concepts, and the ATC categories with the most missing concepts. The mappings from ATC to RxNorm in OSV and UMLS were compared, and we determined the number of mappings in common, and the mapping differences, which we categorized. We applied the mappings from OSV and UMLS on a sample of Electronic Health Record (EHR) data.</div></div><div><h3>Results</h3><div>NLM contained the most ATC and RxNorm concepts. UMLS contained more missing mappings (null mappings) than OSV, 1949 versus 916. Most mapping differences were in the “unknown ingredient in the ATC label” category, for which UMLS provided no mappings. UMLS had a higher coverage of mappings in the sample EHR data than OSV, 96.5% versus 91%.</div></div><div><h3>Discussion</h3><div>In conclusion, opting for OSV rather than UMLS is generally preferable for an ATC to RxNorm mapping since OSV provides more mappings. However, the results of the sample data show that UMLS can have fewer null mappings in concrete applications.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105777"},"PeriodicalIF":3.7,"publicationDate":"2024-12-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142928800","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Knowledge domain and frontier trends of artificial intelligence applied in solid organ transplantation: A visualization analysis","authors":"Miao Gong ,&nbsp;Yingsong Jiang ,&nbsp;Yingshuo Sun ,&nbsp;Rui Liao ,&nbsp;Yanyao Liu ,&nbsp;Zikang Yan ,&nbsp;Aiting He ,&nbsp;Mingming Zhou ,&nbsp;Jie Yang ,&nbsp;Yongzhong Wu ,&nbsp;Zhongjun Wu ,&nbsp;ZuoTian Huang ,&nbsp;Hao Wu ,&nbsp;Liqing Jiang","doi":"10.1016/j.ijmedinf.2024.105782","DOIUrl":"10.1016/j.ijmedinf.2024.105782","url":null,"abstract":"<div><h3>Background</h3><div>Solid organ transplantation (SOT) is vital for end-stage organ failure but faces challenges like organ shortage and rejection. Artificial intelligence (AI) offers potential to improve outcomes through better matching, success prediction, and automation. However, the evolution of AI in SOT research remains underexplored. This study uses bibliometric analysis to identify trends, hotspots, and key contributors in the field.</div></div><div><h3>Methods</h3><div>821 articles from the Web of Science Core Collection were exported for analysis. Microsoft Excel 2021 was used for descriptive statistics. VOSviewer, CiteSpace, Scimago Graphica, and Biblioshiny were used for bibliometric analysis. The ggalluvial package in R was utilized to create Sankey diagrams, and top articles were selected based on citation count.</div></div><div><h3>Results</h3><div>This analysis reveals the rapid expansion of AI in SOT. Key areas include robotic surgery, organ allocation, outcome prediction, immunosuppression management, and precision medicine. Robotic surgery has improved transplant outcomes. AI algorithms optimize organ matching and enhance fairness. Machine learning models predict outcomes and guide treatment, while AI-based systems advance personalized immunosuppression. AI in precision medicine, including diagnostics and imaging, is crucial for transplant success.</div></div><div><h3>Conclusion</h3><div>This study highlights AI’s transformative potential in SOT, with significant contributions from countries like the USA, Canada, and the UK. Key institutions such as the University of Toronto and the University of Pittsburgh have played vital roles. However, practical challenges like ethical issues, bias, and data integration remain. Fostering international and interdisciplinary collaborations is crucial for overcoming these challenges and accelerating AI’s integration into clinical practice, ultimately improving patient outcomes.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105782"},"PeriodicalIF":3.7,"publicationDate":"2024-12-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143140322","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Developing an AI-Based clinical decision support system for basal insulin titration in type 2 diabetes in primary Care: A Mixed-Methods evaluation using heuristic Analysis, user Feedback, and eye tracking","authors":"Camilla Heisel Nyholm Thomsen ,&nbsp;Thomas Kronborg ,&nbsp;Stine Hangaard ,&nbsp;Peter Vestergaard ,&nbsp;Morten Hasselstrøm Jensen","doi":"10.1016/j.ijmedinf.2024.105783","DOIUrl":"10.1016/j.ijmedinf.2024.105783","url":null,"abstract":"<div><h3>Background and aim</h3><div>The progressive nature of type 2 diabetes often, in time, necessitates basal insulin therapy to achieve glycemic targets. However, despite standardized titration algorithms, many people remain poorly controlled after initiating insulin therapy, leading to suboptimal glycemic control and complications. Both healthcare professionals and people with type 2 diabetes have expressed the need for novel tools to aid in this process. Traditional titration methods often lack the precision needed to address individual differences in glycemic response. Recent studies have highlighted the potential of AI-driven solutions, which can leverage large datasets to model patient-specific characteristics. Therefore, this study aims to develop a digital platform for an AI-based clinical decision support system to assist healthcare professionals in primary care with personalized and optimal basal insulin titration for people with type 2 diabetes.</div></div><div><h3>Methods</h3><div>An iterative mixed-method approach was used for system development, incorporating usability engineering principles. Initial requirements were gathered from domain experts and followed by heuristic evaluation and user-based evaluation. Data from these evaluations guided successive iterations of the prototype.</div></div><div><h3>Results</h3><div>The initial prototype featured a retrospective graph of insulin doses and fasting glucose levels and a dose adjustment simulation environment. Heuristic evaluation identified 92 issues, primarily related to minimalistic and aesthetic design. The second prototype addressed these concerns, but user-based evaluation found 66 additional usability problems, notably with HbA1c presentation and the need for more glucose measures. The final prototype showed high usability, with a median System Usability Scale score of 93.8. Task completion rates were high (task 1: 87.5%, task 2: 75.0%, and task 3: 100%). Eye-tracking data showed minimal distractions.</div></div><div><h3>Conclusions</h3><div>The AI-based Clinical Decision Support System shows promise in managing basal insulin titration for people with type 2 diabetes, addressing clinical inertia, and providing a user-friendly, efficient tool to improve glycemic control during insulin titration.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105783"},"PeriodicalIF":3.7,"publicationDate":"2024-12-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142959161","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Influencing factors: Unveiling patterns and reasons in telehealth care utilization and adoption/avoidance decisions","authors":"Ning Yang ,&nbsp;Xin Yang","doi":"10.1016/j.ijmedinf.2024.105781","DOIUrl":"10.1016/j.ijmedinf.2024.105781","url":null,"abstract":"<div><h3>Background</h3><div>The rapid expansion of telehealth, accelerated by the COVID-19 pandemic, has highlighted gaps in understanding demographic and health factors shaping its use. Exploring reasons behind individuals’ choices regarding telehealth can guide strategies to promote adoption among diverse populations.</div></div><div><h3>Methods</h3><div>Data from 5,119 participants in the 2022 Health Information National Trends Survey were analyzed. Dependent variables included telehealth usage and reasons for choosing or avoiding it. Independent variables included demographics, general health, and mental health. Associations were examined using multiple logistic regression models.</div></div><div><h3>Results</h3><div>Factors significantly associated with higher odds of telehealth use included education (college graduate: OR = 1.57, 95 % CI [1.19, 2.06]), gender (male: OR = 0.69, 95 % CI [0.55, 0.87]), rural residency (nonmetro: OR = 0.72, 95 % CI [0.53, 0.97]), depression (OR = 2.91, 95 % CI [2.29, 3.71]), age (e.g., 35–49: OR = 1.66, 95 % CI [1.2, 2.29]), and general health status (good: OR = 0.78, 95 % CI [0.61, 1], excellent or very good: OR = 0.74, 95 % CI [0.58, 0.95]). Older individuals preferred telehealth for convenience but inclined to avoid it in favor of in-person visits. Asian and other group were less likely to use telehealth for seeking advice and including others in visits.</div></div><div><h3>Conclusions</h3><div>Disparities in telehealth utilization were observed across gender, age, education, health status, and urbanization levels. Policymakers should focus on equitable delivery methods, updated regulatory frameworks, and reducing access disparities, especially in underserved communities.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105781"},"PeriodicalIF":3.7,"publicationDate":"2024-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142924116","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":"Towards secure and trusted AI in healthcare: A systematic review of emerging innovations and ethical challenges","authors":"Muhammad Mohsin Khan ,&nbsp;Noman Shah ,&nbsp;Nissar Shaikh ,&nbsp;Abdulnasser Thabet ,&nbsp;Talal alrabayah ,&nbsp;Sirajeddin Belkhair","doi":"10.1016/j.ijmedinf.2024.105780","DOIUrl":"10.1016/j.ijmedinf.2024.105780","url":null,"abstract":"<div><h3>Introduction</h3><div>Artificial Intelligence is in the phase of health care, with transformative innovations in diagnostics, personalized treatment, and operational efficiency. While having potential, critical challenges are apparent in areas of safety, trust, security, and ethical governance. The development of these challenges is important for promoting the responsible adoption of AI technologies into healthcare systems.</div></div><div><h3>Methods</h3><div>This systematic review of studies published between 2010 and 2023 addressed the applications of AI in healthcare and their implications for safety, transparency, and ethics. A comprehensive search was performed in PubMed, IEEE Xplore, Scopus, and Google Scholar. Those studies that met the inclusion criteria provided empirical evidence, theoretical insights, or systematic evaluations addressing trust, security, and ethical considerations.</div></div><div><h3>Results</h3><div>The analysis brought out both the innovative technologies and the continued challenges. Explainable AI (XAI) emerged as one of the significant developments. It made it possible for healthcare professionals to understand AI-driven recommendations, by this means increasing transparency and trust. Still, challenges in adversarial attacks, algorithmic bias, and variable regulatory frameworks remain strong. According to several studies, more than 60 % of healthcare professionals have expressed their hesitation in adopting AI systems due to a lack of transparency and fear of data insecurity. Moreover, the 2024 WotNot data breach uncovered weaknesses in AI technologies and highlighted the dire requirement for robust cybersecurity.</div></div><div><h3>Discussion</h3><div>Full understanding of the potential of AI will be possible only with putting into practice of ethical and technical maintains in healthcare systems. Effective strategies would include integrating bias mitigation methods, strengthening cybersecurity protocols to prevent breaches. Also by adopting interdisciplinary collaboration with the goal of forming transparent regulatory guidelines. These are very important steps toward earning trust and ensuring that AI systems are safe, reliable, and fair.</div></div><div><h3>Conclusion</h3><div>AI can bring transformative opportunities to improve healthcare outcomes, but successful implementation will depend on overcoming the challenges of trust, security, and ethics. Future research should focus on testing these technologies in multiple real-world settings, enhance their scalability, and fine-tune regulations to facilitate accountability. Only by combining technological innovations with ethical principles and strong governance can AI reshape healthcare, ensuring at the same time safety and trustworthiness.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105780"},"PeriodicalIF":3.7,"publicationDate":"2024-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142928831","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Predicting the likelihood of readmission in patients with ischemic stroke: An explainable machine learning approach using common data model data","authors":"Yu Seong Hwang ,&nbsp;Seongheon Kim ,&nbsp;Inhyeok Yim ,&nbsp;Yukyoung Park ,&nbsp;Seonguk Kang ,&nbsp;Heui Sug Jo","doi":"10.1016/j.ijmedinf.2024.105754","DOIUrl":"10.1016/j.ijmedinf.2024.105754","url":null,"abstract":"<div><h3>Background</h3><div>Ischemic stroke affects 15 million people worldwide, causing five million deaths annually. Despite declining mortality rates, stroke incidence and readmission risks remain high, highlighting the need for preventing readmission to improve the quality of life of survivors. This study developed a machine-learning model to predict 90-day stroke readmission using electronic medical records converted to the common data model (CDM) from the Regional Accountable Care Hospital in Gangwon state in South Korea.</div></div><div><h3>Methods</h3><div>We retrospectively analyzed data from 1,136 patients with ischemic stroke admitted between August 2003 and August 2021 after excluding cases with missing blood test values. Demographics, blood test results, treatments, and comorbidities were used as key features. Six machine learning models and three deep learning models were used to predict 90-day readmission using the synthetic minority over-sampling technique to address class imbalance. Models were evaluated using threefold cross-validation, and SHapley Additive exPlanations (SHAP) values were calculated to interpret feature importance.</div></div><div><h3>Results</h3><div>Among 1,136 patients, 196 (17.2 %) were readmitted within 90 days. Male patients were significantly more likely to experience readmission (<em>p</em> = 0.02). LightGBM achieved an area under the curve of 0.94, demonstrating that analyzing stroke and stroke-related conditions provides greater predictive accuracy than predicting stroke alone or all-cause readmissions. SHAP analysis highlighted renal and metabolic variables, including creatinine, blood urea nitrogen, calcium, sodium, and potassium, as key predictors of readmission.</div></div><div><h3>Conclusion</h3><div>Machine-learning models using electronic health record-based CDM data demonstrated strong predictive performance for 90-day stroke readmission. These results support personalized post-discharge management and lay the groundwork for future multicenter studies.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105754"},"PeriodicalIF":3.7,"publicationDate":"2024-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142928827","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":"Identifying potential medical aid beneficiaries using machine learning: A Korean Nationwide cohort study","authors":"Junmo Kim ,&nbsp;Su Hyun Park ,&nbsp;Hyesu Lee ,&nbsp;Su Kyoung Lee ,&nbsp;Jihye Kim ,&nbsp;Suhyun Kim ,&nbsp;Yong Jin Kwon ,&nbsp;Kwangsoo Kim","doi":"10.1016/j.ijmedinf.2024.105775","DOIUrl":"10.1016/j.ijmedinf.2024.105775","url":null,"abstract":"<div><h3>Objective</h3><div>To identify potential medical aid beneficiaries using demographic and medical history of individuals and analyzing important features qualitatively.</div></div><div><h3>Methods</h3><div>This retrospective, national cohort, case-control study included data from the National Health Insurance Service (NHIS) in Korea between January 1, 2002 and December 31, 2019. Potential medical aid beneficiaries were classified using several machine learning models (linear models and tree-based models). Demographic data such as age, sex, region, insurance type, insurance fee, and medical history such as diagnosis, operation, statement, visits, and costs were collected. Those data were transformed into a one-dimensional vector for each individual, allowing machine learning models to learn. For feature importance calculation, we used the average gain across all splits for each feature.</div></div><div><h3>Results</h3><div>274,635 individuals were finally included in the study population, and 62,501 were classified as potential medical aid beneficiaries. XGBoost successfully classified potential medical aid beneficiaries with an AUROC of around 0.891. Assuming predicting before two years, the performance was still significant with an AUROC of around 0.832. Economic variables, such as insurance fees and several costs, turned out to be the most important, but variables regarding medical status, such as the results of blood tests and history of chronic diseases, were also important.</div></div><div><h3>Conclusion</h3><div>Machine learning-based models successfully screened potential medical aid beneficiaries. Qualitative analysis of important features well reflected prior knowledge regarding public health. These findings can contribute to the soundness of healthcare finance and the improvement of public health.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"195 ","pages":"Article 105775"},"PeriodicalIF":3.7,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142904016","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}