European heart journal. Digital health最新文献

{"title":"Prediction of white-coat hypertension and white-coat uncontrolled hypertension using machine learning algorithm.","authors":"Ling-Chieh Shih,&nbsp;Yu-Ching Wang,&nbsp;Ming-Hui Hung,&nbsp;Han Cheng,&nbsp;Yu-Chieh Shiao,&nbsp;Yu-Hsuan Tseng,&nbsp;Chin-Chou Huang,&nbsp;Shing-Jong Lin,&nbsp;Jaw-Wen Chen","doi":"10.1093/ehjdh/ztac066","DOIUrl":"https://doi.org/10.1093/ehjdh/ztac066","url":null,"abstract":"<p><strong>Aims: </strong>The detection of white-coat hypertension/white-coat uncontrolled hypertension (WCH/WUCH) with out-of-office blood pressure (BP) monitoring is time- and resource-consuming. We aim to develop a machine learning (ML)-derived prediction model based on the characteristics of patients from a single outpatient visit.</p><p><strong>Methods and results: </strong>Data from two cohorts in Taiwan were used. Cohort one (970 patients) was used for development and internal validation, and cohort two (464 patients) was used for external validation. WCH/WUCH was defined as an office BP of ≥140/90 mmHg and daytime ambulatory BP of <135/85 mmHg in treatment-naïve or treated individuals. Logistic regression, random forest (RF), eXtreme Gradient Boosting, and artificial neural network models were trained using 26 patient parameters. We used SHapley Additive exPlanations values to provide explanations for the risk factors. All models achieved great area under the receiver operating characteristic curve (AUROC), specificity, and negative predictive value in both validations (AUROC = 0.754-0.891; specificity = 0.682-0.910; negative predictive value = 0.831-0.968). The RF model was the best performing (AUROC = 0.884; sensitivity = 0.619; specificity = 0.887; negative predictive value = 0.872; accuracy = 0.819). The five most influential features of the RF model were office diastolic BP, office systolic BP, current smoker, estimated glomerular filtration rate, and fasting glucose level.</p><p><strong>Conclusion: </strong>Our prediction models achieved good performance, underlining the feasibility of applying ML models to outpatient populations for the diagnosis of WCH and WUCH. Further validation with other prospective data sets should be considered in the future.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"3 4","pages":"559-569"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ftp.ncbi.nlm.nih.gov/pub/pmc/oa_pdf/4f/0b/ztac066.PMC9779877.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"10734974","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Circulating biomarker-based risk stratifications individualize arch repair strategy of acute Type A aortic dissection via the XGBoosting algorithm.","authors":"Hong Liu,&nbsp;Si-Chong Qian,&nbsp;Lu Han,&nbsp;Ying-Yuan Zhang,&nbsp;Ying Wu,&nbsp;Liang Hong,&nbsp;Ji-Nong Yang,&nbsp;Ji-Sheng Zhong,&nbsp;Yu-Qi Wang,&nbsp;Dong-Kai Wu,&nbsp;Guo-Liang Fan,&nbsp;Jun-Quan Chen,&nbsp;Sheng-Qiang Zhang,&nbsp;Xing-Xing Peng,&nbsp;Zhi-Wei Tang,&nbsp;Al-Wajih Hamzah,&nbsp;Yong-Feng Shao,&nbsp;Hai-Yang Li,&nbsp;Hong-Jia Zhang","doi":"10.1093/ehjdh/ztac068","DOIUrl":"https://doi.org/10.1093/ehjdh/ztac068","url":null,"abstract":"<p><strong>Aims: </strong>The incremental usefulness of circulating biomarkers from different pathological pathways for predicting mortality has not been evaluated in acute Type A aortic dissection (ATAAD) patients. We aim to develop a risk prediction model and investigate the impact of arch repair strategy on mortality based on distinct risk stratifications.</p><p><strong>Methods and results: </strong>A total of 3771 ATAAD patients who underwent aortic surgery retrospectively included were randomly divided into training and testing cohorts at a ratio of 7:3 for the development and validation of the risk model based on multiple circulating biomarkers and conventional clinical factors. Extreme gradient boosting was used to generate the risk models. Subgroup analyses were performed by risk stratifications (low vs. middle-high risk) and arch repair strategies (proximal vs. extensive arch repair). Addition of multiple biomarkers to a model with conventional factors fitted an ABC risk model consisting of platelet-leucocyte ratio, mean arterial pressure, albumin, age, creatinine, creatine kinase-MB, haemoglobin, lactate, left ventricular end-diastolic dimension, urea nitrogen, and aspartate aminotransferase, with adequate discrimination ability {area under the receiver operating characteristic curve (AUROC): 0.930 [95% confidence interval (CI) 0.906-0.954] and 0.954, 95% CI (0.930-0.977) in the derivation and validation cohort, respectively}. Compared with proximal arch repair, the extensive repair was associated with similar mortality risk among patients at low risk [odds ratio (OR) 1.838, 95% CI (0.559-6.038); <i>P</i> = 0.316], but associated with higher mortality risk among patients at middle-high risk [OR 2.007, 95% CI (1.460-2.757); <i>P</i> < 0.0001].</p><p><strong>Conclusion: </strong>In ATAAD patients, the simultaneous addition of circulating biomarkers of inflammatory, cardiac, hepatic, renal, and metabolic abnormalities substantially improved risk stratification and individualized arch repair strategy.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"3 4","pages":"587-599"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ftp.ncbi.nlm.nih.gov/pub/pmc/oa_pdf/70/b5/ztac068.PMC9779759.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"10747228","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Non-hyperaemic assessment of coronary ischaemia: application of machine learning techniques.","authors":"James N Cameron,&nbsp;Andrea Comella,&nbsp;Nigel Sutherland,&nbsp;Adam J Brown,&nbsp;Thanh G Phan","doi":"10.1093/ehjdh/ztac050","DOIUrl":"https://doi.org/10.1093/ehjdh/ztac050","url":null,"abstract":"<p><strong>Aims: </strong>Hyperaemic and non-hyperaemic pressure ratios (NHPR) are routinely used to identify significant coronary lesions. Machine learning (ML) techniques may help better understand these indices and guide future practice. This study assessed the ability of a purpose-built ML algorithm to classify coronary ischaemia during non-hyperaemia compared with the existing gold-standard technique (fractional flow reserve, FFR). Further, it investigated whether ML could identify components of coronary and aortic pressure cycles indicative of ischaemia.</p><p><strong>Methods and results: </strong>Seventy-seven coronary vessel lesions (39 FFR defined ischaemia, 53 patients) with proximal and distal non-hyperaemic pressure waveforms and FFR values were assessed using supervised and unsupervised learning techniques in combination with principal component analysis (PCA). Fractional flow reserve measurements were obtained from the right coronary artery (13), left anterior descending (46), left circumflex (11), left main (1), obtuse marginal (2), and diagonal (4). The most accurate supervised learning classification utilized whole-cycle aortic with diastolic distal blood pressure waveforms, yielding a classification accuracy of 86.9% (sensitivity 86.8%, specificity 87.2%, positive predictive value 86.8%, negative predictive value 87.2%). Principal component analysis showed subtle variations in coronary pressures at the start of diastole have significant relation to ischaemia, and whole-cycle aortic pressure data are important for determining ischaemia.</p><p><strong>Conclusions: </strong>Our ML algorithm classifies significant coronary lesions with accuracy similar to previous studies comparing time-domain NHPRs with FFR. Further, it has identified characteristics of pressure waveforms that relate to function. These results provide an application of ML to ischaemia requiring only standard data from non-hyperaemic pressure measurements.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"3 4","pages":"505-515"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ftp.ncbi.nlm.nih.gov/pub/pmc/oa_pdf/2b/41/ztac050.PMC9779890.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"10747232","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Variational auto-encoders improve explainability over currently employed heatmap methods for deep learning-based interpretation of the electrocardiogram.","authors":"Rutger R van de Leur,&nbsp;Rutger J Hassink,&nbsp;René van Es","doi":"10.1093/ehjdh/ztac063","DOIUrl":"https://doi.org/10.1093/ehjdh/ztac063","url":null,"abstract":"We appreciate the opportunity to address Higaki and Yamaguchi and their detailed commentary on our study. 1 In the referenced study, we show that variational auto-encoders (VAEs), which use deep neural networks (DNNs) to learn the underlying factors of variation in the median beat electrocardiogram (ECG), can be used to provide improved explainability over previous attempts to open the ‘black box’ of ECG-based DNNs using saliency-based heatmaps. There are currently conflicting definitions of explainability and interpretability in the literature and both are used interchangeably","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"3 4","pages":"502-504"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ftp.ncbi.nlm.nih.gov/pub/pmc/oa_pdf/7b/bc/ztac063.PMC9779792.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"10747230","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Corrigendum.","authors":"","doi":"10.1093/ehjdh/ztac048","DOIUrl":"https://doi.org/10.1093/ehjdh/ztac048","url":null,"abstract":"<p><p>[This corrects the article DOI: 10.1093/ehjdh/ztaa012.][This corrects the article DOI: 10.1093/ehjdh/ztaa018.][This corrects the article DOI: 10.1093/ehjdh/ztab009.][This corrects the article DOI: 10.1093/ehjdh/ztab011.][This corrects the article DOI: 10.1093/ehjdh/ztab032.][This corrects the article DOI: 10.1093/ehjdh/ztab034.][This corrects the article DOI: 10.1093/ehjdh/ztab038.][This corrects the article DOI: 10.1093/ehjdh/ztab051.][This corrects the article DOI: 10.1093/ehjdh/ztab082.][This corrects the article DOI: 10.1093/ehjdh/ztab108.].</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"3 4","pages":"658"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9779761/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"10586478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Automated interpretation of stress echocardiography reports using natural language processing.","authors":"Chengyi Zheng,&nbsp;Benjamin C Sun,&nbsp;Yi-Lin Wu,&nbsp;Maros Ferencik,&nbsp;Ming-Sum Lee,&nbsp;Rita F Redberg,&nbsp;Aniket A Kawatkar,&nbsp;Visanee V Musigdilok,&nbsp;Adam L Sharp","doi":"10.1093/ehjdh/ztac047","DOIUrl":"https://doi.org/10.1093/ehjdh/ztac047","url":null,"abstract":"<p><strong>Aims: </strong>Stress echocardiography (SE) findings and interpretations are commonly documented in free-text reports. Reusing SE results requires laborious manual reviews. This study aimed to develop and validate an automated method for abstracting SE reports in a large cohort.</p><p><strong>Methods and results: </strong>This study included adult patients who had SE within 30 days of their emergency department visit for suspected acute coronary syndrome in a large integrated healthcare system. An automated natural language processing (NLP) algorithm was developed to abstract SE reports and classify overall SE results into normal, non-diagnostic, infarction, and ischaemia categories. Randomly selected reports (<i>n</i> = 140) were double-blindly reviewed by cardiologists to perform criterion validity of the NLP algorithm. Construct validity was tested on the entire cohort using abstracted SE data and additional clinical variables. The NLP algorithm abstracted 6346 consecutive SE reports. Cardiologists had good agreements on the overall SE results on the 140 reports: Kappa (0.83) and intraclass correlation coefficient (0.89). The NLP algorithm achieved 98.6% specificity and negative predictive value, 95.7% sensitivity, positive predictive value, and <i>F</i>-score on the overall SE results and near-perfect scores on ischaemia findings. The 30-day acute myocardial infarction or death outcomes were highest among patients with ischaemia (5.0%), followed by infarction (1.4%), non-diagnostic (0.8%), and normal (0.3%) results. We found substantial variations in the format and quality of SE reports, even within the same institution.</p><p><strong>Conclusions: </strong>Natural language processing is an accurate and efficient method for abstracting unstructured SE reports. This approach creates new opportunities for research, public health measures, and care improvement.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"3 4","pages":"626-637"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ftp.ncbi.nlm.nih.gov/pub/pmc/oa_pdf/97/ff/ztac047.PMC9779789.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"10734966","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Artificial intelligence-derived cardiac ageing is associated with cardiac events post-heart transplantation.","authors":"Ilke Ozcan,&nbsp;Takumi Toya,&nbsp;Michal Cohen-Shelly,&nbsp;Hyun Woong Park,&nbsp;Ali Ahmad,&nbsp;Alp Ozcan,&nbsp;Peter A Noseworthy,&nbsp;Suraj Kapa,&nbsp;Lilach O Lerman,&nbsp;Zachi I Attia,&nbsp;Sudhir S Kushwaha,&nbsp;Paul A Friedman,&nbsp;Amir Lerman","doi":"10.1093/ehjdh/ztac051","DOIUrl":"https://doi.org/10.1093/ehjdh/ztac051","url":null,"abstract":"<p><strong>Aims: </strong>An artificial intelligence algorithm detecting age from 12-lead electrocardiogram (ECG) has been suggested to reflect 'physiological age'. An increased physiological age has been associated with a higher risk of cardiac mortality in the non-transplant population. We aimed to investigate the utility of this algorithm in patients who underwent heart transplantation (HTx).</p><p><strong>Methods and results: </strong>A total of 540 patients were studied. The average ECG ages within 1 year before and after HTx were used to represent pre- and post-HTx ECG ages. Major adverse cardiovascular event (MACE) was defined as any coronary revascularization, heart failure hospitalization, re-transplantation, and mortality. Recipient pre-transplant ECG age (mean 63 ± 11 years) correlated significantly with recipient chronological age (mean 49 ± 14 years, <i>R</i> = 0.63, <i>P</i> < 0.0001), while post-transplant ECG age (mean 54 ± 10 years) correlated with both the donor (mean 32 ± 13 years, <i>R</i> = 0.45, <i>P</i> < 0.0001) and the recipient ages (<i>R</i> = 0.38, <i>P</i> < 0.0001). During a median follow-up of 8.8 years, 307 patients experienced MACE. Patients with an increase in ECG age post-transplant showed an increased risk of MACE [hazard ratio (HR): 1.58, 95% confidence interval (CI): (1.24, 2.01), <i>P</i> = 0.0002], even after adjusting for potential confounders [HR: 1.58, 95% CI: (1.19, 2.10), <i>P</i> = 0.002].</p><p><strong>Conclusion: </strong>Electrocardiogram age-derived cardiac ageing after transplantation is associated with a higher risk of MACE. This study suggests that physiological age change of the heart might be an important determinant of MACE risk post-HTx.</p>","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"3 4","pages":"516-524"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ftp.ncbi.nlm.nih.gov/pub/pmc/oa_pdf/db/48/ztac051.PMC9779895.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"9310426","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"COP27 Climate Change Conference: urgent action needed for Africa and the world: Wealthy nations must step up support for Africa and vulnerable countries in addressing past, present and future impacts of climate change.","authors":"Lukoye Atwoli,&nbsp;Gregory E Erhabor,&nbsp;Aiah A Gbakima,&nbsp;Abraham Haileamlak,&nbsp;Jean-Marie Kayembe Ntumba,&nbsp;James Kigera,&nbsp;Laurie Laybourn-Langton,&nbsp;Bob Mash,&nbsp;Joy Muhia,&nbsp;Fhumulani Mavis Mulaudzi,&nbsp;David Ofori-Adjei,&nbsp;Friday Okonofua,&nbsp;Arash Rashidian,&nbsp;Maha El-Adawy,&nbsp;Siaka Sidibé,&nbsp;Abdelmadjid Snouber,&nbsp;James Tumwine,&nbsp;Mohammad Sahar Yassien,&nbsp;Paul Yonga,&nbsp;Lilia Zakhama,&nbsp;Chris Zielinski","doi":"10.1093/ehjdh/ztac056","DOIUrl":"https://doi.org/10.1093/ehjdh/ztac056","url":null,"abstract":"Editor-in-Chief, East African Medical Journal; Editor-in-Chief, West African Journal of Medicine; Editor-in-Chief, Sierra Leone Journal of Biomedical Research; Editor-in-Chief, Ethiopian Journal of Health Sciences; Chief Editor, Annales Africaines de Medecine; Editor-in-Chief, Annals of African Surgery; University of Exeter, UK; Editor-in-Chief, African Journal of Primary Health Care & Family Medicine; London School of Medicine and Tropical Hygiene; Editor-in-Chief, Curationis; Editor-in-Chief, Ghana Medical Journal; Editorin-Chief, African Journal of Reproductive Health; Executive Editor, Eastern Mediterranean Health Journal; Director of Health Promotion, Eastern Mediterranean Health Journal; Director of Publication, Mali Médical; Managing Editor, Journal de la Faculté de Médecine d’Oran; Editor-in-Chief, African Health Sciences; Editor-in-Chief, Evidence-Based Nursing Research; Managing Editor, East African Medical Journal; Editor-in-Chief, La Tunisie Médicale; and University of Winchester, UK","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"3 4","pages":"496-498"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9779924/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"10584383","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Correspondence to the European Heart Journal-digital health in response to the paper by Attia <i>et al.</i> 2022.","authors":"Nishil Patel,&nbsp;Salaheldin Agamy,&nbsp;Mahmood Ahmad","doi":"10.1093/ehjdh/ztac053","DOIUrl":"https://doi.org/10.1093/ehjdh/ztac053","url":null,"abstract":"We were interested to read the paper by Attia et al. 1 which demon-strated the value of electrocardiogram enabled stethoscopes (ECG-Scope). Their findings show potential in the utilization of artificial intelligence (AI) algorithms in conjunction with a single lead ECG-Scope to identify left ventricular dysfunction (LVSD). A clinical pathway such as this may speed up diagnosis and potentially improve patient outcomes.","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"3 4","pages":"499"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ftp.ncbi.nlm.nih.gov/pub/pmc/oa_pdf/3e/53/ztac053.PMC9779798.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"10734970","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Validation of risk prediction models applied to longitudinal electronic health record data for the prediction of major cardiovascular events in the presence of data shifts.","authors":"Yikuan Li,&nbsp;Gholamreza Salimi-Khorshidi,&nbsp;Shishir Rao,&nbsp;Dexter Canoy,&nbsp;Abdelaali Hassaine,&nbsp;Thomas Lukasiewicz,&nbsp;Kazem Rahimi,&nbsp;Mohammad Mamouei","doi":"10.1093/ehjdh/ztac061","DOIUrl":"https://doi.org/10.1093/ehjdh/ztac061","url":null,"abstract":"Abstract Aims Deep learning has dominated predictive modelling across different fields, but in medicine it has been met with mixed reception. In clinical practice, simple, statistical models and risk scores continue to inform cardiovascular disease risk predictions. This is due in part to the knowledge gap about how deep learning models perform in practice when they are subject to dynamic data shifts; a key criterion that common internal validation procedures do not address. We evaluated the performance of a novel deep learning model, BEHRT, under data shifts and compared it with several ML-based and established risk models. Methods and results Using linked electronic health records of 1.1 million patients across England aged at least 35 years between 1985 and 2015, we replicated three established statistical models for predicting 5-year risk of incident heart failure, stroke, and coronary heart disease. The results were compared with a widely accepted machine learning model (random forests), and a novel deep learning model (BEHRT). In addition to internal validation, we investigated how data shifts affect model discrimination and calibration. To this end, we tested the models on cohorts from (i) distinct geographical regions; (ii) different periods. Using internal validation, the deep learning models substantially outperformed the best statistical models by 6%, 8%, and 11% in heart failure, stroke, and coronary heart disease, respectively, in terms of the area under the receiver operating characteristic curve. Conclusion The performance of all models declined as a result of data shifts; despite this, the deep learning models maintained the best performance in all risk prediction tasks. Updating the model with the latest information can improve discrimination but if the prior distribution changes, the model may remain miscalibrated.","PeriodicalId":72965,"journal":{"name":"European heart journal. Digital health","volume":"3 4","pages":"535-547"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9779795/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"10233201","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}