BMC Medical Informatics and Decision Making最新文献

{"title":"Digital monitoring of medication safety in children: an investigation of ADR signalling techniques in Malaysia.","authors":"Beldona Hema Rekha, Shairyzah Ahmad Hisham, Izyan A Wahab, Norleen Mohamed Ali, Khang Wen Goh, Long Chiau Ming","doi":"10.1186/s12911-024-02801-y","DOIUrl":"https://doi.org/10.1186/s12911-024-02801-y","url":null,"abstract":"<p><strong>Background: </strong>Digital solutions can help monitor medication safety in children who are often excluded in clinical trials. The lack of reliable safety data often leads to either under- or over-dose of medications during clinical management which make them either not responding well to treatment or susceptible to adverse drug reactions (ADRs).</p><p><strong>Aim: </strong>This study investigated ADR signalling techniques to detect serious ADRs in Malaysian children aged from birth to 12 years old using an electronic ADRs' database.</p><p><strong>Methods: </strong>Four techniques (Proportional Reporting Ratio (PRR), Reporting Odds Ratio (ROR), Bayesian Confidence Propagation Neural Network (BCPNN) and Multi-item Gamma Poisson Shrinker (MGPS)) were tested on ADR reports submitted to the National Pharmaceutical Regulatory Agency between 2016 and 2020. Sensitivity, Specificity, Positive Predictive Value (PPV) and Negative Predictive Value (NPV) of the techniques were compared.</p><p><strong>Results: </strong>A total of 31 medicine-Important Medical Event pairs were found and examined among the 3152 paediatric ADR reports. Three techniques (PRR, ROR, MGPS) signalled oculogyric crisis and dystonia for metoclopramide. BCPNN and MGPS signalled angioedema for paracetamol, amoxicillin and ibuprofen. Similar performances were found for PRR, ROR and BCPNN (sensitivity of 12%, specificity of 100%, PPV of 100% and NPV of 21%). MGPS revealed the highest sensitivity (20%) and NPV (23%), as well as similar specificity and PPV (100%).</p><p><strong>Conclusions: </strong>This study suggests that medication safety signalling techniques could be applied on electronic health records to monitor medication safety issues in children. Clinicians and medication safety specialist could prioritise the signals for further clinical consideration and prompt response.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"24 1","pages":"395"},"PeriodicalIF":3.3,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142852890","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Uncertainty-aware automatic TNM staging classification for [¹⁸F] Fluorodeoxyglucose PET-CT reports for lung cancer utilising transformer-based language models and multi-task learning.","authors":"Stephen H Barlow, Sugama Chicklore, Yulan He, Sebastien Ourselin, Thomas Wagner, Anna Barnes, Gary J R Cook","doi":"10.1186/s12911-024-02814-7","DOIUrl":"https://doi.org/10.1186/s12911-024-02814-7","url":null,"abstract":"<p><strong>Background: </strong>[¹⁸F] Fluorodeoxyglucose (FDG) PET-CT is a clinical imaging modality widely used in diagnosing and staging lung cancer. The clinical findings of PET-CT studies are contained within free text reports, which can currently only be categorised by experts manually reading them. Pre-trained transformer-based language models (PLMs) have shown success in extracting complex linguistic features from text. Accordingly, we developed a multi-task 'TNMu' classifier to classify the presence/absence of tumour, node, metastasis ('TNM') findings (as defined by The Eight Edition of TNM Staging for Lung Cancer). This is combined with an uncertainty classification task ('u') to account for studies with ambiguous TNM status.</p><p><strong>Methods: </strong>2498 reports were annotated by a nuclear medicine physician and split into train, validation, and test datasets. For additional evaluation an external dataset (n = 461 reports) was created, and annotated by two nuclear medicine physicians with agreement reached on all examples. We trained and evaluated eleven publicly available PLMs to determine which is most effective for PET-CT reports, and compared multi-task, single task and traditional machine learning approaches.</p><p><strong>Results: </strong>We find that a multi-task approach with GatorTron as PLM achieves the best performance, with an overall accuracy (all four tasks correct) of 84% and a Hamming loss of 0.05 on the internal test dataset, and 79% and 0.07 on the external test dataset. Performance on the individual TNM tasks approached expert performance with macro average F1 scores of 0.91, 0.95 and 0.90 respectively on external data. For uncertainty an F1 of 0.77 is achieved.</p><p><strong>Conclusions: </strong>Our 'TNMu' classifier successfully extracts TNM staging information from internal and external PET-CT reports. We concluded that multi-task approaches result in the best performance, and better computational efficiency over single task PLM approaches. We believe these models can improve PET-CT services by assisting in auditing, creating research cohorts, and developing decision support systems. Our approach to handling uncertainty represents a novel first step but has room for further refinement.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"24 1","pages":"396"},"PeriodicalIF":3.3,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142853031","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A potential predictive model based on machine learning and CPD parameters in elderly patients with aplastic anemia and myelodysplastic neoplasms.","authors":"Yuxiang Qi, Xu Liu, Zhishan Ding, Ying Yu, Zhenchao Zhuang","doi":"10.1186/s12911-024-02781-z","DOIUrl":"https://doi.org/10.1186/s12911-024-02781-z","url":null,"abstract":"<p><strong>Background: </strong>Aplastic anemia (AA) and myelodysplastic neoplasms (MDS) have similar peripheral blood manifestations and are clinically characterized by reduced hematological triad. It is challenging to distinguish and diagnose these two diseases. Hence, utilizing machine learning methods, we employed and validated an algorithm that used cell population data (CPD) parameters to diagnose AA and MDS.</p><p><strong>Methods: </strong>In this study, CPD parameters were obtained from the Beckman Coulter DxH800 analyzer for 160 individuals diagnosed with AA or MDS through a comprehensive retrospective analysis. The individuals were unselectively assigned to a training cohort (77%) and a testing cohort (23%). Additionally, an external validation cohort consisting of eighty-six elderly patients with AA and MDS from two additional centers was established. The discriminative parameters were carefully analyzed through univariate analysis, and the most predictive variables were selected using least absolute shrinkage and selection operator (LASSO) regression. Six machine learning algorithms were utilized to compare the performance of forecasting AA and MDS patients. The area under the curves (AUCs), calibration curves, decision curves analysis (DCA), and shapley additive explanations (SHAP) plots were employed to interpret and assess the model's predictive accuracy, clinical utility, and stability.</p><p><strong>Results: </strong>After the comparative evaluation of various models, the logistic regression model emerged as the most suitable machine learning model for predicting the probability of AA and MDS, which utilized five principal variables (age, MNVLY, SDVLY, MNLALSEGC, and MNCEGC) to accurately estimate the risk of these diseases. The best model delivered an AUC of 0.791 in the testing cohort and had a high specificity (0.850) and positive predictive value (0.818). Furthermore, the calibration curve indicated excellent agreement between actual and predicted probabilities. The DCA curve further supported the clinical utility of our model and offered significant clinical advantages in guiding treatment decisions. Moreover, the model's performance was consistent in an external validation group, with an AUC of 0.719.</p><p><strong>Conclusions: </strong>We developed a novel model that effectively distinguished elderly patients with AA and MDS, which had the potential to provide physicians assistance in early diagnosis and the proper treatment for the elderly.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"24 1","pages":"379"},"PeriodicalIF":3.3,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142853021","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A prior-knowledge-guided dynamic attention mechanism to predict nocturnal hypoglycemic events in type 1 diabetes.","authors":"Xia Yu, Zi Yang, Xinzhuo Wang, Xiaoyu Sun, Ruiting Shen, Hongru Li, Mingchen Zhang","doi":"10.1186/s12911-024-02761-3","DOIUrl":"https://doi.org/10.1186/s12911-024-02761-3","url":null,"abstract":"<p><p>Nocturnal hypoglycemia is a critical problem faced by diabetic patients. Failure to intervene in time can be dangerous for patients. The existing early warning methods struggle to extract crucial information comprehensively from complex multi-source heterogeneous data. In this paper, a deep learning framework with an innovative dynamic attention mechanism is proposed to predict nocturnal hypoglycemic events for type 1 diabetes patients. Features related to nocturnal hypoglycemia are extracted from multi-scale and multi-dimensional data, which enables comprehensive information extraction from diverse sources. Then, we propose a prior-knowledge-guided attention mechanism to enhance the network's learning capability and interpretability. The method was evaluated on a public available clinical dataset, which successfully warned 94.91% of nocturnal hypoglycemic events with an F1-score of 96.35%. By integrating our proposed framework into the nocturnal hypoglycemia early warning model, issues related to feature redundancy and incompleteness were mitigated. Comparative analysis demonstrates that our method outperforms existing approaches, offering superior accuracy and practicality in real-world scenarios.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"24 1","pages":"378"},"PeriodicalIF":3.3,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142853022","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The perception of facilitators and barriers to the use of e-health solutions in Poland: a qualitative study.","authors":"Paulina Smoła, Iwona Młoźniak, Monika Wojcieszko, Urszula Zwierczyk, Mateusz Kobryn, Elżbieta Rzepecka, Mariusz Duplaga","doi":"10.1186/s12911-024-02791-x","DOIUrl":"https://doi.org/10.1186/s12911-024-02791-x","url":null,"abstract":"<p><strong>Background: </strong>E-health entails the use of information and communication technologies in support of health and health-related activities. E-health increased significantly during the COVID-19 pandemic in Poland. The pandemic showed that the e-health environment may be an important element of the response to epidemiological challenges. Polish citizens were provided with an array of e-health tools supporting the provision of health services.</p><p><strong>Methods: </strong>The main aim of the study was to assess the knowledge, use, and opinions about e-health solutions in Polish society. Fifty participants representing the general population took part in in-depth interviews. The interviews were conducted face-to-face with participants in their homes or via a teleconferencing platform from November 2023 to January 2024. At first, the interviewees were recruited by convenience, and at a later stage, a snowballing approach was applied. A semi-structured guide covered the knowledge about and use of e-health solutions, attitudes toward new technologies, and opinions about artificial intelligence and robots in healthcare. The interviewers interviewed 50 participants, of whom 26 were females. The interview transcriptions were analyzed with MAXQDA Analytics Pro 2022 (Release 22.7.0). An approach based on thematic analysis was employed to evaluate the interviews' content.</p><p><strong>Results: </strong>Thematic analysis of the interviews resulted in the identification of three main themes: (1) knowledge about e-health, (2) barriers, and (3) facilitators of e-health use. Recognition of the term 'e-health' was limited among study participants, although they used e-health solutions frequently. The main barriers included limited digital skills and unfavorable attitudes to new technologies. Some of the participants complained about technical difficulties, e.g., poor Internet access. The main facilitators identified based on the interviews include saving time and reducing costs, as well as the ability to access medical records in one repository, as in the case of the Internet Patient Account. Some people believed e-health to be an element of progress. Overall, the study participants supported sharing their medical data for research.</p><p><strong>Conclusions: </strong>Implementing e-health solutions seems to be perceived as an inevitable consequence of technological progress. However, a lack of adequate technical skills remains one of the major obstacles to efficiently utilizing e-health's potential.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"24 1","pages":"381"},"PeriodicalIF":3.3,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142853029","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development of transient ischemic attack risk prediction model suitable for initializing a learning health system unit using electronic medical records.","authors":"Jian Wen, Tianmei Zhang, Shangrong Ye, Cheng Li, Ruobing Han, Ran Huang, Bairong Shen, Anjun Chen, Qinghua Li","doi":"10.1186/s12911-024-02767-x","DOIUrl":"https://doi.org/10.1186/s12911-024-02767-x","url":null,"abstract":"<p><strong>Background: </strong>Patients with transient ischemic attack (TIA) face a significantly increased risk of stroke. However, TIA screening and early detection rates are low, especially in developing countries. This study aims to develop an inclusive and practical TIA risk prediction model using machine learning (ML) that performs well in both hospital and resource-limited clinic settings. This model is essential for initiating the first ML-enabled learning health system (LHS) unit designed for routine and equitable TIA screening and early detection across broad populations.</p><p><strong>Methods: </strong>Employing a novel protocol, this study first standardized data from a hospital's electronic medical records (EMR) to construct inclusive TIA risk prediction ML models using a data-centric approach. Subsequently, a quantitative distribution of TIA risk factors was applied in feature engineering to reduce the number of variables for a practical ML model. This refined model initiated a TIA ML-LHS unit that is capable of continuously updating with new EMR data from hospitals and clinics. Additionally, the practical model underwent external validation using data from another hospital.</p><p><strong>Results: </strong>The inclusive 150-variable ML models, derived from all available EMR variables for TIA, achieved a recall of 0.868 and an accuracy of 0.886 in predicting TIA risk. Further feature engineering produced a practical XGBoost model with 20 variables, maintaining acceptable performance of 0.855 recall and 0.796 accuracy. The initialized TIA ML-LHS unit, based on the practical model, achieved performance metrics of 0.830 recall, 0.726 precision, 0.816 ROC-AUC, and 0.812 accuracy. The model also performed well in external validation, confirming its effectiveness with patient data from different clinical settings.</p><p><strong>Conclusions: </strong>This study developed the first inclusive and practical TIA XGBoost model from full hospital EHR and initiated the first TIA risk prediction ML-LHS unit. This TIA model, which requires only 20 variables, enables the ML-LHS to serve not only patients in hospitals but also those in resource-limited clinics. These results have significant implications for expanding risk-based TIA screening in community and rural clinics, thereby enhancing early detection of TIA among underserved populations and improving health equity. The novel protocol used in this study is also applicable for initiating ML-LHS units for various preventable diseases, providing a new system-level approach to responsible AI development and applications.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"24 1","pages":"392"},"PeriodicalIF":3.3,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142852896","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Reduced-dose deep learning iterative reconstruction for abdominal computed tomography with low tube voltage and tube current.","authors":"Shumeng Zhu, Baoping Zhang, Qian Tian, Ao Li, Zhe Liu, Wei Hou, Wenzhe Zhao, Xin Huang, Yao Xiao, Yiming Wang, Rui Wang, Yuhang Li, Jian Yang, Chao Jin","doi":"10.1186/s12911-024-02811-w","DOIUrl":"https://doi.org/10.1186/s12911-024-02811-w","url":null,"abstract":"<p><strong>Background: </strong>The low tube-voltage technique (e.g., 80 kV) can efficiently reduce the radiation dose and increase the contrast enhancement of vascular and parenchymal structures in abdominal CT. However, a high tube current is always required in this setting and limits the dose reduction potential. This study investigated the feasibility of a deep learning iterative reconstruction algorithm (Deep IR) in reducing the radiation dose while improving the image quality for abdominal computed tomography (CT) with low tube voltage and current.</p><p><strong>Methods: </strong>Sixty patients (male/female, 36/24; Age, 57.72 ± 10.19 years) undergoing the abdominal portal venous phase CT were randomly divided into groups A (100 kV, automatic exposure control [AEC] with reference tube-current of 213 mAs) and B (80 kV, AEC with reference of 130 mAs). Images were reconstructed via hybrid iterative reconstruction (HIR) and Deep IR (levels 1-5). The mean CT and standard deviation (SD) values of four regions of interest (ROI), i.e. liver, spleen, main portal vein and erector spinae at the porta hepatis level in each image serial were measured, and the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated. The image quality was subjectively scored by two radiologists using a 5-point criterion.</p><p><strong>Results: </strong>A significant reduction in the radiation dose of 69.94% (5.09 ± 0.91 mSv vs. 1.53 ± 0.37 mSv) was detected in Group B compared with Group A. After application of the Deep IR, there was no significant change in the CT value, but the SD gradually increased. Group B had higher CT values than group A, and the portal vein CT values significantly differed between the groups (P < 0.003). The SNR and CNR in Group B with Deep IR at levels 1-5 were greater than those in Group A and significantly differed when HIR and Deep IR were applied at levels 1-3 of HIR and Deep IR (P < 0.003). The subjective scores (distortion, clarity of the portal vein, visibility of small structures and overall image quality) with Deep IR at levels 4-5 in Group B were significantly higher than those in group A with HIR (P < 0.003).</p><p><strong>Conclusion: </strong>Deep IR algorithm can meet the clinical requirements and reduce the radiation dose by 69.94% in portal venous phase abdominal CT with a low tube voltage of 80 kV and a low tube current. Deep IR at levels 4-5 can significantly improve the image quality of the abdominal parenchymal organs and the clarity of the portal vein.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"24 1","pages":"389"},"PeriodicalIF":3.3,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142853026","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Engaging adolescents and young adults in decisions about return of genomic research results: study protocol for a mixed-methods longitudinal clinical trial protocol.","authors":"Amy A Blumling, Michelle L McGowan, Cynthia A Prows, Kristin Childers-Buschle, Lisa J Martin, John A Lynch, Kevin R Dufendach, Ellen A Lipstein, Melinda Butsch Kovacic, William B Brinkman, Melanie F Myers","doi":"10.1186/s12911-024-02784-w","DOIUrl":"10.1186/s12911-024-02784-w","url":null,"abstract":"<p><strong>Background: </strong>To protect minors' future autonomy, professional organizations have historically discouraged returning predictive adult-onset genetic test results and carrier status to children. Recent clinical guidance diverges from this norm, suggesting that when minors have genomic sequencing performed for clinical purposes, parents and children should have the opportunity to learn secondary findings, including for some adult-onset conditions. While parents can currently opt in or out of receiving their child's secondary findings, the American Society of Human Genetics Workgroup on Pediatric Genetic and Genomic Testing suggests including adolescents in the decision-making process. However, it is not clear what factors young people consider when given the opportunity to learn genetic findings for themselves. In this manuscript, we report on the methods for a clinical trial that examines adolescents', young adults', and their parents' decisions about learning genomic information for the adolescent or young adult.</p><p><strong>Methods: </strong>We are enrolling assenting (ages 13-17) adolescents and consenting (ages 18-21) young adults in a prospective genomic screening study to assess the choices they make about receiving individual genomic results. Participants use an online tool to indicate whether they want to learn their personal genetic risk for specific preventable, treatable, and adult-onset conditions, as well as carrier status for autosomal recessive conditions. We are examining (1) how choices differ between adolescent and young adult cohorts (as well as between adolescents/young adults and parents) and (2) decisional conflict and stability across study timepoints. Results are returned based on participants' choices. Qualitative interviews with a subset of participants explore decisional stability, adolescent/young adult engagement with parents in decision-making, and the impact of learning pathogenic/likely pathogenic and autosomal recessive carrier results.</p><p><strong>Discussion: </strong>This study explores decision making and decision stability between adolescents and parents (where applicable), as well as the ethical implications and impact of return of clinical-grade genetic research results to adolescents and young adults. The results of this study will contribute empirical evidence to support best practices and guidance on engaging young people in genomic research studies and clinical care that offer return of results.</p><p><strong>Trial registration: </strong>ClinicalTrials.gov Identifier: NCT04481061. Registered 22 July 2020.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"24 1","pages":"391"},"PeriodicalIF":3.3,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142852926","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhanced forecasting of emergency department patient arrivals using feature engineering approach and machine learning.","authors":"Bruno Matos Porto, Flavio Sanson Fogliatto","doi":"10.1186/s12911-024-02788-6","DOIUrl":"https://doi.org/10.1186/s12911-024-02788-6","url":null,"abstract":"<p><strong>Background: </strong>Emergency department (ED) overcrowding is an important problem in many countries. Accurate predictions of ED patient arrivals can help management to better allocate staff and medical resources. In this study, we investigate the use of calendar and meteorological predictors, as well as feature-engineered variables, to predict daily patient arrivals using datasets from eleven different EDs across three countries.</p><p><strong>Methods: </strong>Six machine learning (ML) algorithms were tested on forecasting horizons of 7 and 45 days. Three of them - Light Gradient Boosting Machine (LightGBM), Support Vector Machine with Radial Basis Function (SVM-RBF), and Neural Network Autoregression (NNAR) - were never before reported for predicting ED patient arrivals. Algorithms' hyperparameters were tuned through a grid-search with cross-validation. Prediction performance was assessed using fivefold cross-validation and four performance metrics.</p><p><strong>Results: </strong>The eXtreme Gradient Boosting (XGBoost) was the best-performing model on both prediction horizons, also outperforming results reported in past studies on ED arrival prediction. XGBoost and NNAR achieved the best performance in nine out of the eleven analyzed datasets, with MAPE values ranging from 5.03% to 14.1%. Feature engineering (FE) improved the performance of the ML algorithms.</p><p><strong>Conclusion: </strong>Accuracy in predicting ED arrivals, achieved through the FE approach, is key for managing human and material resources, as well as reducing patient waiting times and lengths of stay.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"24 1","pages":"377"},"PeriodicalIF":3.3,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142852942","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Improving event prediction using general practitioner clinical judgement in a digital risk stratification model: a pilot study.","authors":"Emma Parry, Kamran Ahmed, Elizabeth Guest, Vijay Klaire, Abdool Koodaruth, Prasadika Labutale, Dawn Matthews, Jonathan Lampitt, Alan Nevill, Gillian Pickavance, Mona Sidhu, Kate Warren, Baldev M Singh","doi":"10.1186/s12911-024-02797-5","DOIUrl":"https://doi.org/10.1186/s12911-024-02797-5","url":null,"abstract":"<p><strong>Background: </strong>Numerous tools based on electronic health record (EHR) data that predict risk of unscheduled care and mortality exist. These are often criticised due to lack of external validation, potential for low predictive ability and the use of thresholds that can lead to large numbers being escalated for assessment that would not have an adverse outcome leading to unsuccessful active case management. Evidence supports the importance of clinical judgement in risk prediction particularly when ruling out disease. The aim of this pilot study was to explore performance analysis of a digitally driven risk stratification model combined with GP clinical judgement to identify patients with escalating urgent care and mortality events.</p><p><strong>Methods: </strong>Clinically risk stratified cohort study of 6 GP practices in a deprived, multi-ethnic UK city. Initial digital driven risk stratification into Escalated and Non-escalated groups used 7 risk factors. The Escalated group underwent stratification using GP global clinical judgement (GCJ) into Concern and No concern groupings.</p><p><strong>Results: </strong>3968 out of 31,392 patients were data stratified into the Escalated group and further categorised into No concern (n = 3450 (10.9%)) or Concern (n = 518 (1.7%)) by GPs. The 30-day combined event rate (unscheduled care or death) per 1,000 was 19.0 in the whole population, 67.8 in the Escalated group and 168.0 in the Concern group (p < 0.001). The de-escalation effect of GP assessment into No Concern versus Concern was strongly negatively predictive (OR 0.25 (95%CI 0.19-0.33; p < 0.001)). The whole population ROC for the global approach (Non-escalated, GP No Concern, GP Concern) was 0.614 (0.592-0.637), p < 0.001, and the increase in the ROC area under the curve for 30-day events was all focused here (+ 0.4% (0.3-0.6%, p < 0.001), translating into a specific ROC c-statistic for GP GCJ of 0.603 ((0.565-0.642), p < 0.001).</p><p><strong>Conclusions: </strong>The digital only component of the model performed well but adding GP clinical judgement significantly improved risk prediction, particularly by adding negative predictive value.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"24 1","pages":"382"},"PeriodicalIF":3.3,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142852954","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}