Search

Background: Timely diagnosis of structural heart disease improves patient outcomes, yet many remain underdiagnosed. While population screening with echocardiography is impractical, ECG-based prediction models can help target high-risk patients. We developed a novel ECG-based machine learning approach to predict multiple structural heart conditions, hypothesizing that a composite model would yield higher prevalence and positive predictive values to facilitate meaningful recommendations for echocardiography. Methods: Using 2 232 130 ECGs linked to electronic health records and echocardiography reports from 484 765 adults between 1984 to 2021, we trained machine learning models to predict the presence or absence of any of 7 echocardiography-confirmed diseases within 1 year. This composite label included the following: moderate or severe valvular disease (aortic/mitral stenosis or regurgitation, tricuspid regurgitation), reduced ejection fraction 15 mm. We tested various combinations of input features (demographics, laboratory values, structured ECG data, ECG traces) and evaluated model performance using 5-fold cross-validation, multisite validation trained on 1 site and tested on 10 independent sites, and simulated retrospective deployment trained on pre-2010 data and deployed in 2010. Results: Our composite rECHOmmend model used age, sex, and ECG traces and had a 0.91 area under the receiver operating characteristic curve and a 42% positive predictive value at 90% sensitivity, with a composite label prevalence of 17.9%. Individual disease models had area under the receiver operating characteristic curves from 0.86 to 0.93 and lower positive predictive values from 1% to 31%. Area under the receiver operating characteristic curves for models using different input features ranged from 0.80 to 0.93, increasing with additional features. Multisite validation showed similar results to cross-validation, with an aggregate area under the receiver operating characteristic curve of 0.91 across our independent test set of 10 clinical sites after training on a separate site. Our simulated retrospective deployment showed that for ECGs acquired in patients without preexisting structural heart disease in the year 2010, 11% were classified as high risk and 41% (4.5% of total patients) developed true echocardiography-confirmed disease within 1 year. Conclusions: An ECG-based machine learning model using a composite end point can identify a high-risk population for having undiagnosed, clinically significant structural heart disease while outperforming single-disease models and improving practical utility with higher positive predictive values. This approach can facilitate targeted screening with echocardiography to improve underdiagnosis of structural heart disease.

Background Heart failure and sleep-disordered breathing have been increasingly recognized as co-occurring conditions. Their bidirectional relationship warrants investigation into whether heart failure therapy improves sleep and sleep-disordered breathing. We sought to explore the effect of treatment with sacubitril/valsartan on sleep-related endpoints from the AWAKE-HF study. Methods and Results AWAKE-HF was a randomized, double-blind study conducted in 23 centers in the United States. Study participants with heart failure with reduced rejection fraction and New York Heart Association class II or III symptoms were randomly assigned to receive treatment with either sacubitril/valsartan or enalapril. All endpoints were assessed at baseline and after 8 weeks of treatment. Portable sleep-monitoring equipment was used to measure the apnea-hypopnea index, including obstructive and central events. Total sleep time, wake after sleep onset and sleep efficiency were exploratory measures assessed using wrist actigraphy. The results were as follows 140 patients received treatment in the double-blind phase (sacubitril/valsartan, n = 70; enalapril, n = 70). At baseline, 39% and 40% of patients randomly assigned to receive sacubitril/valsartan or enalapril, respectively, presented with undiagnosed, untreated, moderate-to-severe sleep-disordered breathing (≥ 15 events/h), and nearly all had obstructive sleep apnea. After 8 weeks of treatment, the mean 4% apnea-hypopnea index changed minimally from 16.3/h to 15.2/h in the sacubitril/valsartan group and from 16.8/h to 17.6/h in the enalapril group. Mean total sleep time was long at baseline and decreased only slightly in both treatment groups at week 8 (–14 and –11 minutes for sacubitril/valsartan and enalapril, respectively), with small changes in wake after sleep onset and sleep efficiency in both groups. Conclusions In a cohort of patients with heart failure with reduced rejection fraction who met prescribing guidelines for sacubitril/valsartan, one-third had undiagnosed moderate-to-severe obstructive sleep apnea. The addition of sacubitril/valsartan therapy did not significantly improve sleep-disordered breathing or sleep duration or efficiency. Patients who meet indications for treatment with sacubitril/valsartan should be evaluated for sleep-disordered breathing.

Context: There is an ongoing established need to develop engaging pain assessment strategies to provide more effective individualized care to pediatric patients with serious illnesses. This study explores the acceptability of wireless devices as one option. Objective: To evaluate the ability of wrist-wearable technology to collect physiological data from children with serious illnesses. Methods: Single-site prospective observational study conducted between September 2017 and September 2018 at Rady Children's Hospital, San Diego, California, inpatient wards. Pediatric patients with diagnoses of cancer and sickle cell disease admitted to the hospital for acute-on-chronic pain and taking opioid pain medications were asked to complete two 24-hour continuous monitoring periods with the Empatica E4 wristband. Results: Data collected from the device correlated with manually obtained vital signs. Children responded favorably to wearing the device. Participants with reported subjective pain versus no pain had average heart rate increased by 16.4 bpm, skin temperature decreased by 3.5°C, and electrodermal activity decreased by 0.27. Conclusions: This study shows the possibility of collecting continuous biophysical data in a nonobtrusive manner in seriously ill children experiencing acute-on-chronic pain using wearable devices. It provides the framework for larger studies to explore the utility of such data in relation to metrics of pain and suffering in this patient population.

The use of information and communication technology (ICT) in medical and healthcare services goes beyond everyday life. Ex pectations of a new medical environment, not previously experienced by ICT, exist in the near future. In particular, chronic meta bolic diseases such as diabetes and obesity, have a high prevalence and high social and economic burden. In addition, the continu ous evaluation and monitoring of daily life is important for effective treatment and management. Therefore, the wide use of ICT-based digital health systems is required for the treatment and management of these diseases. In this article, we compiled a variety of digital health technologies introduced to date in the field of diabetes and metabolic diseases.

Background Screening for asymptomatic, undiagnosed atrial fibrillation (AF) has the potential to allow earlier treatment, possibly resulting in prevention of strokes, but also to increase medical resource utilization. Objective To compare healthcare utilization rates during the year following initiation of screening among participants screened for AF by electrocardiogram (ECG) sensor patch compared with a matched observational control group. Methods A total of 1718 participants recruited from a health care plan based on age and comorbidities who were screened with an ECG patch (actively monitored group) as part of a prospective, pragmatic research trial were matched by age, sex, and CHA2DS2-VASc score with 3371 members from the same health plan (observational control group). Healthcare utilization, including visits, prescriptions, procedures, and diagnoses, during the 1 year following screening was compared between the groups using health plan claims data. Results Overall, the actively monitored group had significantly higher rates of cardiology visits (adjusted incidence rate ratio [aIRR] [95% confidence interval (CI)]: 1.43 [1.27, 1.60]), no difference in primary care provider visits (aIRR [95% CI]: 1.0 [0.95, 1.05]), but lower rates of emergency department (ED) visits and hospitalizations (aIRR [95% CI]: 0.80 [0.69, 0.92]) compared with controls. Among those with newly diagnosed AF, the reduction in ED visits and hospitalizations was even greater (aIRR [95% CI]: 0.27 [0.17, 0.43]). Conclusion AF screening in an asymptomatic, moderate-risk population with an ECG patch was associated with an increase in cardiology outpatient visits but also significantly lower rates of ED visits and hospitalizations over the 1 year following screening.

Aims Screening for asymptomatic atrial fibrillation (AF) could prevent strokes and save lives, but the AF burden of those detected can impact prognosis. New technologies enable continuous monitoring or intermittent electrocardiogram (ECG) snapshots, however, the relationship between AF detection rates and the burden of AF found with intermittent strategies is unknown. We simulated the likelihood of detecting AF using real-world 2-week continuous ECG recordings and developed a generalizable model for AF detection strategies. Methods and results From 1738 asymptomatic screened individuals, ECG data of 69 individuals (mean age 76.3, median burden 1.9%) with new AF found during 14 days continuous monitoring were used to simulate 30 seconds ECG snapshots one to four times daily for 14 days. Based on this simulation, 35–66% of individuals with AF would be detected using intermittent screening. Twice-daily snapshots for 2 weeks missed 48% of those detected by continuous monitoring, but mean burden was 0.68% vs. 4% in those detected (P Conclusion Using twice-daily ECG snapshots over 2 weeks would detect only half of individuals discovered to have AF by continuous recordings, but AF burden of those missed was low. A model predicting AF detection, validated using real-world data, could assist development of optimized AF screening programmes.

AWAKE-HF evaluated the effect of the initiation of sacubitril/valsartan versus enalapril on activity and sleep using actigraphy in patients who have heart failure with reduced ejection fraction (HFrEF). In this randomized, double-blind study, patients with HFrEF (n = 140) were randomly assigned to sacubitril/valsartan or enalapril for 8 weeks, followed by an 8-week open-label phase with sacubitril/valsartan. Primary endpoint was change from baseline in mean activity counts during the most active 30 min/day at week 8. The key secondary endpoint was change in mean nightly activity counts/minute from baseline to week 8. Kansas City Cardiomyopathy Questionnaire-23 (KCCQ-23) was an exploratory endpoint. There were no detectable differences between groups in geometric mean ratio of activity counts during the most active 30 min/day at week 8 compared with baseline (0.9456 [sacubitril/valsartan:enalapril]; 95% confidence interval [CI] 0.8863–1.0088; P = 0.0895) or in mean change from baseline in activity during sleep (difference: 2.038 counts/min; 95% CI − 0.062 to 4.138; P = 0.0570). Change from baseline to week 8 in KCCQ-23 was 2.89 for sacubitril/valsartan and 4.19 for enalapril, both nonsignificant. In AWAKE-HF, no detectable differences in activity and sleep were observed when comparing sacubitril/valsartan with enalapril in patients with HFrEF using a wearable biosensor. ClinicalTrials.gov, NCT02970669.

Background Diabetes is an independent risk factor for atrial fibrillation (AF), which is associated with increases in mortality and morbidity, as well as a diminished quality of life. Renal involvement in diabetes is common, and since chronic kidney disease (CKD) shares several of the same putative mechanisms as AF, it may contribute to its increased risk in individuals with diabetes. The objective of this study is to identify the relationship between CKD and the rates of newly-diagnosed AF in individuals with diabetes taking part in a screening program using a self-applied wearable electrocardiogram (ECG) patch. Materials and methods The study included 608 individuals with a diagnosis of diabetes among 1738 total actively monitored participants in the prospective mHealth Screening to Prevent Strokes (mSToPS) trial. Participants, without a prior diagnosis of AF, wore an ECG patch for 2 weeks, twice, over a 4-months period and followed clinically through claims data for 1 year. Definitions of CKD included ICD-9 or ICD-10 chronic renal failure diagnostic codes, and the Health Profile Database algorithm. Individuals requiring dialysis were excluded from trial enrollment. Results Ninety-six (15.8%) of study participants with diabetes also had a diagnosis of CKD. Over 12 months of follow-up, 19 new cases of AF were detected among the 608 participants. AF was newly diagnosed in 7.3% of participants with CKD and 2.3% in those without (P Conclusion Among individuals with diabetes, CKD significantly increases the risk of incident AF. Identification of AF prior to clinical symptoms through active ECG screening could help to improve the clinical outcomes in individuals with CKD and diabetes.

Background:Pregnant women living in rural locations in the USA have higher rates of maternal and infant mortality compared to their urban counterparts. One factor contributing to this disparity may be lack of representation of rural women in traditional clinical research studies of pregnancy. Barriers to participation often include transportation to research facilities, which are typically located in urban centers, childcare, and inability to participate during nonwork hours.Methods:POWERMOM is a digital research app which allows participants to share both survey and sensor data during their pregnancy. Through non-targeted, national outreach a study population of 3612 participants (591 from rural zip codes and 3021 from urban zip codes) have been enrolled so far in the study, beginning on March 16, 2017, through September 20, 2019.Results:On average rural participants in our study were younger, had higher pre-pregnancy weights, were less racially diverse, and were more likely to plan a home birth compared to the urban participants. Both groups showed similar engagement in terms of week of pregnancy when they joined, percentage of surveys completed, and completion of the outcome survey after they delivered their baby. However, rural participants shared less HealthKit or sensor data compared to urban participants.Discussion:Our study demonstrated the feasibility and effectiveness of enrolling pregnant women living in rural zip codes using a digital research study embedded within a popular pregnancy app. Future efforts to conduct remote digital research studies could help fill representation and knowledge gaps related to pregnant women.

Traditional viral illness surveillance relies on in-person clinical or laboratory data, paper-based data collection, and outdated technology for data transfer and aggregation. We aimed to assess whether continuous sensor data can provide an early warning signal for COVID-19 activity as individual physiological and behavioural changes might precede symptom onset, care seeking, and diagnostic testing.This multivariable, population-based, modelling study recruited adult (aged ≥18 years) participants living in the USA who had a smartwatch or fitness tracker on any device that connected to Apple HealthKit or Google Fit and had joined the DETECT study by downloading the MyDataHelps app. In the model development cohort, we included people who had participated in DETECT between April 1, 2020, and Jan 14, 2022. In the validation cohort, we included individuals who had participated between Jan 15 and Feb 15, 2022. When a participant joins DETECT, they fill out an intake survey of demographic information, including their ZIP code (postal code), and surveys on symptoms, symptom onset, and viral illness test dates and results, if they become unwell. When a participant connects their device, historical sensor data are collected, if available. Sensor data continue to be collected unless a participant withdraws from the study. Using sensor data, we collected each participant's daily resting heart rate and step count during the entire study period and identified anomalous sensor days, in which resting heart rate was higher than, and step count was lower than, a specified threshold calculated for each individual by use of their baseline data. The proportion of users with anomalous data each day was used to create a 7-day moving average. For the main cohort, a negative binomial model predicting 7-day moving averages for COVID-19 case counts, as reported by the Centers for Disease Control and Prevention (CDC), in real time, 6 days in the future, and 12 days in the future in the USA and California was fitted with CDC-reported data from 3 days before alone (Hsub0/sub) or in combination with anomalous sensor data (Hsub1/sub). We compared the predictions with Pearson correlation. We then validated the model in the validation cohort.Between April 1, 2020, and Jan 14, 2022, 35 842 participants enrolled in DETECT, of whom 4006 in California and 28 527 in the USA were included in our main cohort. The Hsub1/submodel significantly outperformed the Hsub0/submodel in predicting the 7-day moving average COVID-19 case counts in California and the USA. For example, Pearson correlation coefficients for predictions 12 days in the future increased by 32·9% in California (from 0·70 [95% CI 0·65-0·73] to 0·93 [0·92-0·94]) and by 12·2% (from 0·82 [0·79-0·84] to 0·92 [0·91-0·93]) in the USA from the Hsub0/submodel to the Hsub1/submodel. Our validation model also showed significant correlations for predictions in real time, 6 days in the future, and 12 days in the future.Our study showed that passively collected sensor data from consenting participants can provide real-time disease tracking and forecasting. With a growing population of wearable technology users, these sensor data could be integrated into viral surveillance programmes.The National Center for Advancing Translational Sciences of the US National Institutes of Health, The Rockefeller Foundation, and Amazon Web Services.

Introduction: Timely diagnosis of structural heart disease improves patient outcomes, yet millions remain undiagnosed. ECG-based prediction models can help identify high-risk patients for targeted screening, but existing individual disease models often have low positive predictive values (PPV) and limited clinical utility. Hypothesis: An ECG-based composite model can predict one of multiple, actionable structural heart conditions and yield higher prevalence and PPVs than individual models. Methods: Using 2,141,366 ECGs linked to echocardiography and EHR records from 461,466 adults from 1984-2021, we trained machine learning models to predict any of 7 echocardiography-confirmed diseases within 1 year. This composite label included: moderate or severe valvular disease (aortic stenosis or regurgitation, mitral stenosis or regurgitation, tricuspid regurgitation), reduced ejection fraction of 15mm. We tested various combinations of inputs and evaluated model performance with 1) cross-validation and 2) a simulated retrospective deployment. We measured area under the receiver operating characteristic curve (AUROC) and precision-recall curve (AUPRC), and PPV at 90% sensitivity. Results: Our composite "rECHOmmend" model using age, sex and ECG traces had an AUROC of 0.91, AUPRC of 0.78, and PPV of 52% at 90% sensitivity and 23% disease prevalence. Individual disease models had similar AUROCs (0.88-0.93), but lower AUPRCs (0.07-0.71) and PPVs (2%-41%; Figure). Across inputs, model AUROCs ranged from 0.85 to 0.93. Our simulated deployment model classified 22% of at-risk patients in 2010 as high-risk, of whom 40% developed true, echo-confirmed disease within 1 year. Conclusions: An ECG-based machine learning model using a composite endpoint can predict undiagnosed structural heart disease, outperforming single disease models with higher PPVs to facilitate targeted screening with echocardiography.

Background: Several large trials have employed age or clinical features to select patients for atrial fibrillation (AF) screening to reduce strokes. We hypothesized that a deep neural network (DNN) model risk prediction based on ECG would be superior to age and clinical variables at selecting a population at high risk for AF and AF-related stroke. Methods: We retrospectively included all patients with an ECG at Geisinger without a prior history of AF. Incidence of AF and AF-related strokes were identified as outcomes within 1 and 3 years after the ECG, respectively. AF-related stroke was defined as a stroke where AF was diagnosed at the time of stroke or within a year after the stroke. We selected a high-risk cohort for AF screening based on five risk stratification methods - criteria from four clinical trials (mSToPS, STROKESTOP, GUARD-AF and SCREEN-AF) and the DNN model at the qualifying ECG. We simulated patient selection and evaluated outcomes for twenty 1-year periods between 2010-2014 centered around the ECG encounter. For the clinical trials, the patients were considered eligible if they met the criteria before or within the period unless they satisfied exclusion criteria at the time of ECG. Results: The DNN model achieved optimal sensitivity (65%), PPV (10%), NNS for AF (10) within this population compared with all other risk models with a NNS for AF-related stroke of 160. Total screening number, sensitivity, positive predictive value (PPV) and number needed to screen (NNS) to capture AF and AF-related stroke are summarized in Table 1. The number of additional screens for the DNN model was slightly higher for two of the other models (SCREEN-AF and STROKESTOP) but lower than the other two (mSToPS and GUARD-AF). Conclusions: A DNN ECG-based risk prediction model is superior to contemporary AF-screening criteria based on age alone or age and clinical features in selecting a population for additional screening due to high risk for future AF and potential AF-related strokes.

Introduction: Initiation of QTc-prolonging medications may lead to the rare but potentially catastrophic event, torsades de pointes (TDP). At present, no adequate, generalizable tools exist to predict drug-induced long QTc (LQT); machine learning from ECG data is a promising approach. Hypothesis: Prediction of drug-induced LQT using an ECG-based machine learning model is feasible, and outperforms a model trained on baseline QTc, age, and sex alone. Methods: We identified baseline 12-lead ECGs with QTc values < 500 ms for patients who had not received any known, conditional, or possible QTc prolonging medication per CredibleMeds at the time of ECG or within the past 90 days. We matched these with ECGs from the same patients while they were taking at least one CredibleMeds drug (“on-drug” ECGs). Using 5-fold cross-validation, we trained and tested two machine learning models using the baseline ECGs of the 92,848 resulting pairs to predict drug-induced LQT (≥500 ms) in the on-drug ECGs: a deep neural network using ECG voltage data, and a gradient-boosted tree using the baseline QTc. Age and sex were also inputs to both models. Results: On-drug LQT prevalence was 16%. The ECG model demonstrated superior performance in predicting on-drug LQT (area under the receiver operating characteristic curve (AUC) = 0.756) compared to the QTc model (0.710). At a potential operating point (Figure), the ECG model had 89% sensitivity and 95% negative predictive value. Even in the subset of patients with baseline QTc < 470/480 ms (male/female; post-drug LQT prevalence = 14%), the ECG model demonstrated good performance (AUC = 0.736). Conclusions: An ECG-based machine learning model can stratify patients by risk of developing drug-induced LQT better than a model using baseline QTc alone. This model may have clinical value to identify high-risk drug starts that would benefit from closer monitoring and others who are at low risk of drug-induced LQT.

BackgroundEarly diagnosis of structural heart disease improves patient outcomes, yet many remain underdiagnosed. While population screening with echocardiography is impractical, electrocardiogram (ECG)-based prediction models can help target high-risk patients. We developed a novel ECG-based machine learning approach to predict multiple structural heart conditions, hypothesizing that a composite model would yield higher prevalence and positive predictive values (PPVs) to facilitate meaningful recommendations for echocardiography.MethodsUsing 2,232,130 ECGs linked to electronic health records and echocardiography reports from 484,765 adults between 1984-2021, we trained machine learning models to predict the presence of any of seven echocardiography-confirmed diseases within one year. This composite label included: moderate or severe valvular disease (aortic/mitral stenosis or regurgitation, tricuspid regurgitation), reduced ejection fraction 15mm. We tested various combinations of input features (demographics, labs, structured ECG data, ECG traces) and evaluated model performance using 5-fold cross-validation, multi-site validation trained on one clinical site and tested on 11 other independent sites, and simulated retrospective deployment trained on pre-2010 data and deployed in 2010.FindingsOur composite “rECHOmmend” model using age, sex and ECG traces had an area under the receiver operating characteristic curve (AUROC) of 0.91 and a PPV of 42% at 90% sensitivity at a prevalence of 17.9% for our composite label. Individual disease models had AUROCs ranging from 0.86-0.93 and lower PPVs from 1%-31%. The AUROC for models using different input features ranged from 0.80-0.93, increasing with additional features. Multi-site validation showed similar results to the cross-validation, with an aggregate AUROC of 0.91 across our independent test set of 11 clinical sites after training on a separate site. Our simulated retrospective deployment showed that for ECGs acquired in patients without pre-existing known structural heart disease in a single year, 2010, 11% were classified as high-risk, of which 41% developed true, echocardiography-confirmed disease within one year.InterpretationAn ECG-based machine learning model using a composite endpoint can predict previously undiagnosed, clinically significant structural heart disease while outperforming single disease models and improving practical utility with higher PPVs. This approach can facilitate targeted screening with echocardiography to improve under-diagnosis of structural heart disease.

Background Atrial fibrillation (AF) is common, often without symptoms, and is an independent risk factor for mortality, stroke and heart failure. It is unknown if screening asymptomatic individuals for AF can improve clinical outcomes. Methods mSToPS was a pragmatic, direct-to-participant trial that randomized individuals from a single US-wide health plan to either immediate or delayed screening using a continuous-recording ECG patch to be worn for two weeks and 2 occasions, ~3 months apart, to potentially detect undiagnosed AF. The 3-year outcomes component of the trial was designed to compare clinical outcomes in the combined cohort of 1718 individuals who underwent monitoring and 3371 matched observational controls. The prespecified primary outcome was the time to first event of the combined endpoint of death, stroke, systemic embolism, or myocardial infarction among individuals with a new AF diagnosis, which was hypothesized to be the same in the two cohorts but was not realized. Results Over the 3 years following the initiation of screening (mean follow-up 29 months), AF was newly diagnosed in 11.4% (n = 196) of screened participants versus 7.7% (n = 261) of observational controls (p Conclusions At 3 years, screening for AF was associated with a lower rate of clinical events and improved outcomes relative to a matched cohort, although the influence of earlier diagnosis of AF via screening on this finding is unclear. These observational data, including the high event rate surrounding a new clinical diagnosis of AF, support the need for randomized trials to determine whether screening for AF will yield a meaningful protection from strokes and other clinical events. Trail registration The mHealth Screening To Prevent Strokes (mSToPS) Trial is registered on ClinicalTrials.gov with the identifier NCT02506244.

Background: The availability of a wide range of innovative wearable sensor technologies today allows for the ability to capture and collect potentially important health-related data in ways not previously possible. These sensors can be adopted in digitalized clinical trials, i.e., clinical trials conducted outside the clinic to capture data about study participants in their day-to-day life. However, having participants activate, charge, and wear the digital sensors for long hours may prove to be a significant obstacle to the success of these trials. Objective: This study explores a broad question of wrist-wearable sensor effectiveness in terms of data collection as well as data that are analyzable per individual. The individuals who had already consented to be part of an asymptomatic atrial fibrillation screening trial were directly sent a wrist-wearable activity and heart rate tracker device to be activated and used in a home-based setting. Methods: A total of 230 participants with a median age of 71 years were asked to wear the wristband as frequently as possible, night and day, for at least a 4-month monitoring period, especially to track heart rhythm during sleep. Results: Of the individuals who received the device, 43% never transmitted any data. Those who used the device wore it a median of ∼15 weeks (IQR 2–24) and for 5.3 days (IQR 3.2–6.5) per week. For rhythm detection purposes, only 5.6% of all recorded data from individuals were analyzable (with beat-to-beat intervals reported). Conclusions: This study provides some important learnings. It showed that in an older population, despite initial enthusiasm to receive a consumer-quality wrist-based fitness device, a large proportion of individuals never activated the device. However, it also found that for a majority of participants it was possible to successfully collect wearable sensor data without clinical oversight inside a home environment, and that once used, ongoing wear time was high. This suggests that a critical barrier to overcome when incorporating a wearable device into clinical research is making its initiation of use as easy as possible for the participant.

Aims Limited data are available regarding the ability of sacubitril/valsartan to provide clinically meaningful health‐related quality of life (HRQoL) improvements among individuals with heart failure (HF). Objective measurement of physical activity and sleep using actigraphy can provide insight into daily functioning and HRQoL. Methods and results We designed an 18 week, multicenter, randomized, double‐blind, double‐dummy, parallel‐group study to objectively assess changes in function and HRQoL directly after initiating sacubitril/valsartan vs. enalapril in participants with HF in their home environments. A total of 136 outpatient, ambulatory participants with New York Heart Association Class II or III HF with reduced ejection fraction (HFrEF) will be included in the study. Patients will undergo a 2 week baseline observational phase (continuing current HF treatment); data from the second week of this phase will be the baseline value for comparison with those of subsequent periods. Patients will then enter an 8 week blinded‐treatment phase (randomly assigned 1:1 to sacubitril/valsartan or enalapril), followed by an 8 week open‐label extension phase (treatment with only sacubitril/valsartan). The primary efficacy endpoint is the change in mean activity counts during the most active 30 min of the participant's day between baseline and the final randomized treatment phase measurement. Secondary endpoints include the change in mean sleep activity during the randomized and open‐label phases; questionnaires will also assess HRQoL measures. Rather than analysing pooled actigraphy data, the researchers are considering each participant to be acting as his or her own control. Conclusions This will be the first study to assess the effects of sacubitril/valsartan on objective measures of sleep and activity in individuals with HFrEF within the context of their daily lives. Wearable accelerometer devices will be used to gain insight into how the medication affects physical activity and sleep.

Digital medicine has the capacity to affect all aspects of medicine, including disease prediction, prevention, diagnosis, treatment, and post-treatment management. In the field of thyroidology, researchers are also investigating potential applications of digital technology for the thyroid disease. Recent studies using artificial intelligence (AI)/machine learning (ML) have reported reasonable performance for the classification of thyroid nodules based on ultrasonographic (US) images. AI/ML-based methods have also shown good diagnostic accuracy for distinguishing between benign and malignant thyroid lesions based on cytopathologic findings. Assistance from AI/ML methods could overcome the limitations of conventional thyroid US and fine-needle aspiration cytology. A web-based database has been developed for thyroid cancer care. In addition to its role as a nationwide registry of thyroid cancer, it is expected to serve as a clinical platform to facilitate better thyroid cancer care and as a research platform providing comprehensive disease-specific big data. Evidence has been found that biosignal monitoring with wearable devices may predict thyroid dysfunction. This real-world thyroid function monitoring could aid in the management and early detection of thyroid dysfunction. In the thyroidology field, research involving the range of digital medicine technologies and their clinical applications is expected to be even more active in the future.

In this study, we propose a post-hoc explainability framework for deep learning models applied to quasi-periodic biomedical time-series classification. As a case study, we focus on the problem of atrial fibrillation (AF) detection from electrocardiography signals, which has strong clinical relevance. Starting from a state-of-the-art pretrained model, we tackle the problem from two different perspectives: global and local explanation. With global explanation, we analyze the model behavior by looking at entire classes of data, showing which regions of the input repetitive patterns have the most influence for a specific outcome of the model. Our explanation results align with the expectations of clinical experts, showing that features crucial for AF detection contribute heavily to the final decision. These features include R-R interval regularity, absence of the P-wave or presence of electrical activity in the isoelectric period. On the other hand, with local explanation, we analyze specific input signals and model outcomes. We present a comprehensive analysis of the network facing different conditions, whether the model has correctly classified the input signal or not. This enables a deeper understanding of the network's behavior, showing the most informative regions that trigger the classification decision and highlighting possible causes of misbehavior.

Introduction: Deep learning (DL) has proved effective for automatic identification of atrial fibrillation (AF) using single-lead ECG. Adoption and trust of DL by clinicians is limited by its black box nature. Hypothesis: Post hoc explanations can elucidate what part of ECG signal is used by the black box DL algorithm, quantifying the importance of clinically relevant features in the classification decision. Making DL decision process transparent will help its integration into clinical practice. Methods: 8,528 single-lead ECG recordings collected using AliveCor devices (PhysioNet) were used. Each signal was labeled as normal sinus rhythm, AF, other arrhythmia or noise. DL automatic classification involves a lightweight convolutional neural network architecture - MobileNet - whose performance is analyzed with an explanation method for DL. Results: Each RR interval is divided into 8 equal segments, where segment 1 follows each R peak, 4 and 5 correspond to the isoelectric baseline, and 7 to the P wave. The explanation method substitutes one of these segments with a straight line, and the corresponding change in sensitivity highlights its importance for the DL algorithm decision. MobileNet achieved a sensitivity of 92.5% to identify AF (9.4% of ECGs were in AF). Sensitivity increases by 2.5% when Segment 7 is removed, indicating that the absence of P wave leads the network to classify more frequently samples as AF.(Figure) When Segments 4 and 5 are removed, the sensitivity decreases by 2.5% and 5.0%, and by 26.7% when removed together. When all RR intervals are normalized to the same value (RR in the Figure), sensitivity for AF drops by 78.3%, showing that RR intervals are key for AF detection by DL algorithm. Conclusions: Post hoc explanations for AF detection by DL from single-lead ECG show the importance of common morphological features used for classifying AF. These methods can be used to understand the decision-making process of DL and motivate its clinical adoption.