European heart journal. Digital health最新文献

Introduction: Integrating machine learning (ML) models into wearable or connected devices to deliver early warning alerts prior to atrial fibrillation (AF) onset may represent an effective preventive strategy. Machine learning algorithms applied to two-lead Holter electrocardiogram (ECG) recordings can support the development of predictive models capable of detecting imminent paroxysmal AF episodes within short-term windows. This approach could facilitate a more targeted 'pill-in-the-pocket' (PITP)-like intervention strategy, potentially enhancing timely therapeutic actions and improving patient outcomes.

Aim: This study aimed to identify patients currently in sinus rhythm who will experience an AF episode within the subsequent hours by analysing 24-h Holter ECG recordings with ML.

Methods: We established a novel database comprising 95 871 manually analysed Holter ECG recordings, identifying 1319 episodes of paroxysmal AF from 872 patients. Among these, 835 AF episodes from 506 recordings had more than 60 min of normal sinus rhythm prior to AF onset and more than 10 min of sustained AF following onset. Patients were stratified into five age groups: all patients combined, under 60 years, 60-70 years, 70-80 years, and over 80 years. Additionally, 365 recordings from 347 patients without rhythm abnormalities were identified and classified, from which two ECG segments were selected. Two deep learning (DL) models were trained on raw ECG data to predict AF onset. To compare DL models with traditional ML approaches using heart rate variability (HRV) parameters, we employed a random forest classifier and a gradient-boosted decision tree model (XGBoost, XGB).

Results: The decision trees models trained on HRV parameters delivered the best predictive performance. The most significant results were observed for episodes lasting more than 5 min of AF, achieving an area under the receiver operating characteristic curve of 0.919 (95% CI: 0.879-0.958) and an area under the precision-recall curve of 0.919 (95% CI: 0.879-0.958) for XGB. At a decision threshold of 0.5, accuracy was 84.5% (81.2-87.8), sensitivity was 83.0% (79.5-86.4), specificity was 86.6% (79.3-93.9), positive predictive value was 90.2% (85.5-94.9), negative predictive value was 78.4% (74.7-82.1), and the F1 score was 86.2% (83.5-89.0).

Conclusion: These findings indicate that HRV parameters contain crucial information for the short-term prediction of AF onset, supporting preventive strategies. Integration of such predictive models into wearable mHealth technologies could facilitate a PITP-like preventive approach, potentially reducing AF-related morbidity. Prospective studies are warranted to validate these promising results further.

Ambulatory cardiac monitoring (ACM) allows long-term electrocardiogram (ECG) monitoring to detect arrhythmias with different modalities, ranging from short-term Holter monitoring (up to 48 h) to long-term continuous patch ECG monitors (up to 14 days), external event recorders (up to 30 days), and implantable loop recorders (ILRs). Access and reimbursement for ACM across Europe are not well understood. We performed a systematic review and analysis to understand ACM reimbursement across Europe, including a review of the reimbursement systems in each country and a detailed inspection of clinical coding and provider reimbursement. Level of reimbursement is dependent on many factors, including clinical setting (inpatient, outpatient, and day case), hospital length of stay, diagnosis, complications/severity, geographical location, hospital type, and device model and manufacturer. In most countries, reimbursement is performed for the monitoring procedure itself, without considering the time extension of monitoring and the specific type of device used for monitoring. The monetary value of reimbursement varies by country for both ACM and ILR [for Holter from €17.49 to €939.78 and for ILR from €416.14 (provider reimbursement only) to €18,718 (provider reimbursement bundled with ILR device)]. Holter and ILR are universally reimbursed, but newer ACM technologies with extended duration of monitoring, including long-term continuous monitoring and event recorders, are not. Across Europe, we found large variation in monetary values for reimbursement for ACM and ILR. We also found limited reimbursement and access to longer-duration ACM technologies. These findings suggest heterogeneous and problematic access to evidence-based tools for longer-duration monitoring.

Aims: Familial hypercholesterolaemia (FH) is the most prevalent autosomal dominant disorder, affecting about 1 in 200-250 individuals. It is the leading cause of early and aggressive coronary artery disease.

Methods and results: We analysed patients with genetically confirmed FH or a score >8 on the Dutch Lipid Clinics Network criteria from the National Registry of the Spanish Atherosclerosis Society, including individuals enrolled from January 2010 to December 2017. The model utilized a dataset incorporating family history, clinical characteristics, laboratory results, genetic data, imaging studies, and lipid-lowering treatment details. Eighty per cent of the population was allocated for training the AI algorithm and 20% was used for testing. A Histogram-based Gradient Boosting Classification Tree was used. The stability of the AI system was assessed using K-fold cross-validation. Shapley additive explanations methodology analysed the influence of different variables by sex. Youden's J statistic established the optimal cut-off point. A total of 1764 patients were included (51.8% women), among whom 264 experienced major adverse cardiovascular events (MACEs), with 8% being women. The final model incorporated 82 variables, achieving metrics of precision for MACE accuracy (0.92), recall (0.89), F1-score (0.91), and receiver operating characteristic (0.88; 95% confidence interval, 0.85-0.90). In the model, age, gamma-glutamyl transferase levels, and subclinical disease significantly impacted risk for women, while year of birth, age at initiation of statin treatment, and HbA1c levels were more influential for men. The optimal risk threshold was 0.25.

Conclusion: Artificial intelligence-machine learning algorithms are promising tools for enhancing vascular risk stratification, revealing critical sex-based differences.

Aims: The potential application of wearable technology solutions for detecting out-of-hospital cardiac arrest (OHCA) is increasingly explored to enhance survival outcomes, but questions related to device accuracy, psychological well-being, privacy, and equal access need to be sorted out before implementation in clinical care and society. This qualitative interview study investigates patients' and physicians' perspectives on end-user needs, preferences, and potential barriers to smartwatch-based OHCA detection.

Methods and results: During the first cycle, 17 patients with elevated OHCA risk were interviewed individually (n = 8) or with their partner (n = 9). The second cycle consisted of interviews with 18 physicians, including cardiologists (n = 9), and other physicians involved in the clinical care of OHCA: general physicians (n = 3), intensivists (n = 3), and neurologists (n = 3). Verbatim interview transcripts were inductively coded for thematic analysis. Five overarching themes were derived: (1) acceptance, use, and optimal informed consent; (2) identifying the target population; (3) technology-related barriers, such as false alarms, localization, and locked doors; (4) design preferences related to privacy, comfort, and hardware alternatives; and (5) public-private partnerships, costs, and equitable access.

Conclusion: This study is the first to explore the perspectives of patients and physicians on smartwatch-based OHCA detection using qualitative analysis of interview data. The results provide important building blocks for the ethically and psychologically sound development and implementation of smartwatch-based OHCA detection in clinical practice, taking the social context into account. The availability of OHCA detection using wearable devices to a wide range of people requires further attention, with emphasis on populations at elevated risk of cardiac arrhythmias.

Aims: We aimed to develop and externally validate a convolutional neural network (CNN) using sinus rhythm electrocardiograms (ECGs) and CHARGE-AF features to predict incident paroxysmal atrial fibrillation (AF), benchmarking its performance against the CHARGE-AF score.

Methods and results: We curated 157 192 sinus ECGs from 76 986 patients within the New York University (NYU) Langone Health system, splitting data into training, validation, and test sets. Two cohorts, from suburban US outpatient practices and Greek tertiary hospitals, were used for external validation. The model utilizing the sinus ECG signal and all CHARGE-AF features achieved the highest test set area under the receiver operator curve (AUC) (0.89) and area under the precision recall curve (AUPRC) (0.69), outperforming the CHARGE-AF score alone. Model robustness was maintained in the external US cohort (AUC 0.90, AUPRC 0.67) and the European cohort (AUC 0.85, AUPRC 0.78). Subgroup analyses confirmed consistent performance across age, sex, and race strata. A CNN using ECG signals alone retained strong predictive ability, particularly when simulating missing or inaccurate clinical data.

Conclusion: Our CNN integrating sinus rhythm ECGs and CHARGE-AF features demonstrated superior predictive performance over traditional risk scoring alone for detecting incident paroxysmal AF. The model maintained accuracy across geographically and clinically diverse external validation cohorts, supporting its potential for broad implementation in AF screening strategies.

Aims: Patients with abdominal aortic aneurysms and peripheral arterial disease (arterial vascular diseases) carry a high disease burden and are likely to experience cardiovascular events. Novel strategies using artificial intelligence could identify which patients with arterial vascular diseases are at high risk of cardiovascular disease progression. Structured data dictionaries are needed to ensure high-quality, unbiased, and ethically sound data input for artificial intelligence models. The aim of this study was to obtain expert consensus-based data dictionaries that adhere to applicable ethical guidelines to support research on arterial vascular diseases.

Methods and results: The data dictionaries were created through a modified Delphi approach to achieve consensus among key opinion leaders in the cardiovascular field. First, data requirements were defined and variable longlists were created per disease through a literature review. Secondly, written feedback rounds were held. Lastly, face-to-face meetings were held to establish consensus on the final data dictionaries. During the whole process, ethical and legal experts on trustworthy artificial intelligence were involved to ensure adherence to corresponding guidelines and laws. The aneurysm data dictionary contains 312 variables, and the peripheral arterial disease data dictionary contains 325 variables. A total of 16 clinical experts were involved in the creation, including 12 vascular surgeons, two vascular medicine specialists, one cardiologist, and one gastroenterology surgeon and digital health expert.

Conclusion: Two expert consensus-based data dictionaries for use in clinical and artificial intelligence research on arterial vascular diseases were created, developed for application in research on predicting disease progression and cardiovascular risk.

Aims: Early detection of the need for coronary revascularization and timely intervention may reduce fatal events, but limited screening tools often leads to underdiagnosis. The aim of this study is to use a deep learning model (DLM) that utilizes electrocardiography (ECG) and the eXtreme Gradient Boosting (XGBoost) model to predict risk of coronary revascularization in the general population.

Methods and results: This study included patients with at least one ECG per patient. The development set comprised 113 451 patients for training a DLM. After excluding patients with elevated troponin I levels and those without follow-up records, the internal validation set consisted of 66 680 patients. The external validation was conducted using data from a community hospital. XGBoost predicted events based on demographic data and ECG features. The primary endpoint was coronary revascularization within 1 year. Model performance was evaluated using the C-index. The DLM stratified patients by risk of coronary revascularization within 1 year. The study included 51% males with a mean age of 53 years, 10% with diabetes, and a revascularization rate of 2.6%. High-risk patients had a hazard ratio of 9.77 (95% CI: 7.63-12.51) compared with low-risk patients. The C-index was 0.825 (95% CI: 0.81-0.84). Combining demographic and AI-ECG data, XGBoost achieved a C-index of 0.884 (95% CI: 0.87-0.89). Comparative C-index analysis revealed significantly different discriminative performance between models (P = 1.110223e-15).

Conclusions: The DLM demonstrates ECG's potential as a screening tool for coronary revascularization, enabling opportunistic detection and prompting further evaluation of high-risk patients.

Aims: Cardiovascular diseases are the leading global cause of mortality, with ischaemic heart disease contributing significantly to the burden. Primary and secondary prevention strategies are essential to reducing the incidence and recurrence of acute myocardial infarction. Healthcare professionals are no longer the sole source of health education; the Internet, including tools powered by artificial intelligence, is also widely utilized. This study evaluates the accuracy and the readability of large language model (LLM)-generated information on cardiovascular primary and secondary prevention.

Methods and results: An observational study assessed LLM's responses to two tailored questions about acute myocardial infarction risk prevention. The LLM used was ChatGPT (4o version). Expert cardiologists evaluated the accuracy of each response using a Likert scale, while readability was assessed with the Flesch Reading Ease Score (FRES). ChatGPT-4o provided comprehensive and accurate responses for 15 out of 20 (75%) of the items. Readability scores were low, with median FRES indicating that both primary and secondary prevention content were difficult to understand. Specialized clinical topics exhibited lower accuracy and readability compared to the other topics.

Conclusion: The current study demonstrated that ChatGPT-4o provided accurate information on primary and secondary prevention, although its readability was assessed as difficult. However, clinical oversight still remains critical to bridge gaps in accuracy and readability and ensure optimal patient outcomes.