From mobile ECGs to tele-rehabilitation to defibrillator delivery drones, a range of new data was on display at the American Heart Association Scientific Sessions 2019.

Roundup: 16 digital health studies, pilots and reviews from AHA 2019

By Dave Muoio
Share

There are few clinical areas more in need of innovation than heart disease. It’s the leading killer among men and women in the US, accounting for one in every four deaths by the CDC’s numbers, and yet it’s also a body of health conditions that could be substantially mitigated through a number of lifestyle and behavior changes.

So of course it’s little surprise that cardiovascular health and care has long been in the crosshairs of digital health, whether it be in the form of fitness trackers, tele-rehabilitation and coaching programs, mobile ECG hardware, disease detection algorithms or any number of other novel approaches.

Much like any other healthcare tools though, these novel approaches all need to win the approval of the clinical community before they’ll begin to see meaningful use at scale — and what better time to do so than one of the year’s largest gatherings of cardiovascular clinicians, researchers, educators, program managers and vendors?

MobiHealthNews was on-site last weekend at the American Heart Association Scientific Sessions 2019 in Philadelphia to check out posters, oral presentations and news from this year’s show. Read on below for a scattershot of 16 such studies from both academics and digital cardio health players, and also check out some of the other features and HIMSS TV videos straight from the show floor.


Atrial Fibrillation Progression in a Novel Digital Health Cohort

by Zohreh Karbaschi et al.

WHAT: A longitudinal characterization of AliveCor’s connected ECG users based on the company’s de-identified ECG result database.

TOPLINE DATA: Based on inclusion criteria, researches reviewed a final cohort of 28,736 regular users with 9,222,967 ECGs logged over a median follow up of 26 months. Among these, they identified 940 users who had progressed from paroxysmal to persistent atrial fibrillation during the course of the study period. This group was significantly older, and had significantly higher average heart rates than others less affected by atrial fibrillation.

TAKEAWAY: This investigation offers a blueprint of how digitally collected user data can be used to monitor or identify large-scale cardiac disease trends, (with AliveCor Chief Medical Officer Dr. Jacqueline Shreibati promising that more explorations of the companies data will be cropping up at future academic conferences).


Assessment of Atrial Fibrillation Burden by Smartwatch Photoplethysmography

by Anthony Faranesh et al.

WHAT: An accuracy assessment of photoplethysmography (PPG)-based atrial fibrillation detection on Fitbit smartwatches, using a single-lead ECG patch as the standard. All readings interpretations were confirmed by a cardiologist.

TOPLINE DATA: Among 96 patients with atrial fibrillation history who wore both devices for seven days, 61 had 0% burden while seven had 100% burden as characterized by ECG and PPG. Those 28 patients in between had an average ECG burden of 4% and an average PPG burden of 5%. There was no significant difference in atrial fibrillation burden between the two methods, although there were five false positives and two false negatives among the smartwatches’ readings.

TAKEAWAY: Smartwatch PPG detection of atrial fibrillation appears accurate, but longer study periods and broader sample populations should be investigated as well.


Provider Perspectives on Smartwatch Monitoring for Atrial Fibrillation

by Eric Ding et al.

WHAT: A patient and provider survey gauging comfort with smartwatch cardiovascular monitoring and automated atrial fibrillation alerts.

TOPLINE DATA: Among 144 patients asked to wear a smartwatch and report their perspective, 86% said they were comfortable with sending heart rate data to their provider. Surveys from 36 providers were split on whether to schedule another device or a clinical visit to confirm smartwatch ECG and pulse alerts. However, 91% said they were willing to prescribe anticoagulants based on the smartwatch’s readings.

TAKEAWAY: Older patients are interested in using these devices, but healthcare providers are not entirely sold on their value for diagnosis.


Prevalence of Mobile Health Access and Use Among Individuals With or at Risk for Cardiovascular Disease: 2018 Health Information National Trends Survey (HINTS)

by Rongzi Shan et al.

WHAT: A review of national survey data to determine whether mobile health tools are within reach for those affected or at risk of cardiovascular disease.

TOPLINE DATA: The analysis landed on 3,248 survey respondents — 1,903 of whom had or were at risk for heart disease and 1,345 who were not. Smartphone or tablet owner ship was high among all respondents, but lower than those in the risk group (80% versus 92%). These patients less often reported health app ownership (48% versus 55%) and digital health tracking (43% versus 51%), and yet were more likely to report sharing their digitally-collected health data with a provider (23% versus 13%).

TAKEAWAY: There’s certainly room for digital health growth among those most at risk for heart disease, although it’s worth exploring why these individuals are less exposed to the technologies than their peers.


Prospective Analysis of Utility of Signals From an ECG-Enabled Stethoscope to Automatically Detect a Low Ejection Fraction Using Neural Network Techniques Trained From the Standard 12-Lead Ecg

by Zachi Attia et al.

WHAT: A prospective study applying an algorithm for 12-lead low ejection fraction detection to Eko’s single-lead digital stethoscope device.

TOPLINE DATA: Both transthoracic ECG and supine and/or sitting stethoscope ECG readings were obtained for 100 hundred patients referred for ECG. When comparing the latter’s best recordings to the former, the algorithm yielded an AUC of 0.906 for ejection fraction less than 35%, and an AUC of 0.841 for ejection fraction under 50%.

TAKEAWAY: Use of the algorithm with Eko’s digital stethoscope was fairly reliable, providing a potential opportunity to introduce screening during routine care.


Beyond Apple 4 Watch and Af — Is Artificial Intelligence Ready to Take on ST Elevation Mi Diagnosis?

by Sameer Mehta et al.

WHAT: An algorithm capable of detecting ECG heart rate variability based on ST segment rather than peak intervals (RR).

TOPLINE DATA: Researchers tested their algorithm among a sample of 5,087 anonymized ECG STEMI records — 2,543 of which included confirmed cases that were 50% “normal” and 50% “abnormal.” After training on 90% of this sample and testing on the final 10%, the model achieved accuracy, sensitivity and specificity greater than 95%.

TAKEAWAY: ECG-based detection algorithms using this metric can be reliable, and according to the authors it might be possible to integrate this approach into wearables.


Feasibility of, and Adherence to, a Novel, Home-based Cardiac Tele-rehabilitation Program for Heart Attack Survivors: The MI-PACE Study

by Emily Erisom et al.

WHAT: A 12-week pilot of a smartwatch-informed tele-rehabilitation program for 20 acute myocardial infarction survivors.

TOPLINE DATA: Eighteen participants completed the full program. These patients wore their smartwatch a median 86% of study days, for a median 12.7 hours per day. Participants completed 86% of their overall exercise goals, and maintained a gradual increase in average number of walking minutes with each week. A median 91% of telephone cardiac rehabilitation sessions were completed. Compared to baseline, participants were significantly more adherent to exercise recommendations and reported greater angina stability at week 12.

TAKEAWAY: In this pilot, wearable-driven tele-rehabilitation appears to be a feasible stand in for those unable to receive cardiac rehabilitation on site.


Heart Rate Variability for the Detection of Myocardial Ischemia in Subjects Without Known Coronary Artery Disease: The HRV-DETECT Study

by Ilan Goldenberg et al.

WHAT: A multicenter study of whether Lev El Diagnostics’ HeartTrends algorithm for heart rate variability is correlated with the presence of myocardial ischemia in 1,043 patients without known coronary artery disease.

TOPLINE DATA: Researchers detected myocardial ischemia in 6.3% of study participants. After removing other risk factors, analysis showed an independent association between low heart rate variability values and higher risk of myocardial ischemia (odds ratio = 2.11; < .001). Low heart rate variability values were also tied to an increased risk for cardiovascular events over an average follow-up of 3.5 years (adjusted hazard ratio = 1.72; p = .008).

TAKEAWAY: "This study introduces a sensitive, non-invasive test for the early detection of subclinical or early ischemia in individuals with low-to-intermediate pre-test probability for [coronary artery disease], providing an incremental risk assessment and re-stratification tool for physicians,” Goldenberg, director of clinical cardiovascular research at the University of Rochester Medical Center and the study’s principal investigator, said in a statement.


Enhancing Mobile Health to Achieve Both Hypertension and Diabetes Goals

by Mansur Shomali et al.

WHAT: A review of BlueStar user data after the inclusion of a blood pressure tracking feature.

TOPLINE DATA: The review included data from 969 users, who were active for an average of 117 days. Although optional, 12% of users logged 1,868 blood pressure readings in the first 30 days, for a mean systolic blood pressure of 127 mmHg. Of note, 35% of entries were above the American Heart Association’s recommended blood pressure of 130 mmHg, and 16% were above 140 mmHg, in which cases the user and care team would be notified.

TAKEAWAY: Introducing voluntary blood pressure monitoring into a diabetes management digital therapeutic can highlight opportunities for care intervention.


More Frequent Remote Home Monitoring Decreases Blood Pressure in an Unselected Population of People With Diabetes

by Karen Xu et al.

WHAT: A study examining a cohort of 1,690 Livongo program members with a diabetes diagnosis and hypertension over 12 weeks.

TOPLINE DATA: After 12 weeks of Livongo’s testing and coaching program, 1,065 participants with whose hypertension was uncontrolled at baseline recorded decreases of 7.2 mmHg for systolic blood pressure, and 4.0 mmHg for diastolic blood pressure (P < .001). Across the full cohort, a multivariable regression analysis found age, baseline blood pressure, and the number of at-home blood pressure monitor uses to each be significantly associated with blood pressure declines.

TAKEAWAY: “This study shows that by providing people with connected technology, insights, access to coaching, and the ability manage their conditions outside of the four walls of the doctor’s office, we can drive sustained outcomes,” Dr. Bimal R. Shah, Livongo’s chief medical officer and the senior author of the study, said in a statement.


Screening for Atrial Fibrillation in Native Americans Using Smartphone-Based ECG

by Khaled Elkholey et al.

WHAT: An opportunistic screening investigation among 507 Native Americans aged 50 years or older with no atrial fibrillation history that used AliveCor’s Kardia Mobile single-lead smartphone ECG device.

TOPLINE DATA: Eight of the patients screened were flagged for possible atrial fibrillation, had their diagnosis confirmed by a cardiologist and were referred for further management. Six of these patients younger than 65 years, which is the minimum age recommended for screening in clinical guidelines. The researchers saw now significant differences in age or other risk factors when comparing these eight patients with the rest of the cohort.

TAKEAWAY: Native Americans appear to develop atrial fibrillation at a younger age, and could likely benefit from earlier digital atrial fibrillation screening in tribal clinics.


Fitbit-derived Physical Activity in Atrial Fibrillation Patients

by Sarah Semaan et al.

WHAT: An observation of step count’s relationship to atrial fibrillation within the Health eHeart Study cohort, as measured by Fitbit devices.

TOPLINE DATA: Among the 3,333 Health eHeart participants that connected their Fitbit, 7% had atrial fibrillation. Adjusted analysis found those with the condition took 8.5% fewer steps than those without (< .0001), and higher participant scores on the validated Atrial Fibrillation Effect on Quality of Life Survey (AFEQT) could be tied to high daily recorded steps ( p = .017).

TAKEAWAY: Daily activity is directly associated with atrial fibrillation measures, and connected fitness trackers provided an objective means of highlighting this relationship.


A Feasibility Study to Assess Delivery of Automated External Defibrillators via Drones in Simulated Cardiac Arrest — Users’ Experiences and the Human-Drone Interaction

by Jessica Zegre-Hemsey et al.

WHAT: A quantitative and qualitative survey on whether the public is comfortable receiving automatic external defibrillator deliveries from a drone in a simulated emergency situation.

TOPLINE DATA: Surveys from 64 participants and interviews with 15 participants suggest an overall positive experience receiving the tool from the drone. They often cited the ability to stay with the cardiac arrest victim while awaiting the delivery as a major benefit and, according to the presenter, noted a feeling of relief when seeing the drones arriving.

TAKEAWAY: With some fine-tuning of the approach, delivery of defibrillators via drone during out-of-hospital cardiac arrest appears to be feasible.


Home Blood Pressure Monitoring Plus a Mobile Phone-Based Hypertension Health Coaching App or a Blood Pressure Tracking App: A Randomized Controlled Clinical Trial

by Stephen Persell et al.

WHAT: A comparison between a hypertension control program employing an Omron connected blood pressure monitor and coaching app, and a control program providing a monitor and tracking app.

TOPLINE DATA: Researchers enrolled 333 participants, 297 of whom completed a six-month follow up. Adjusted mean systolic blood pressure was 2 mmHg lower among those randomized to the intervention, although this difference was not statistically significant. Similarly, adjusted self-reported physical activity was improved by 26.7 minutes per week, but this was also not significant between the groups. Of note, a subgroup analysis suggests the intervention’s effects may differ due to patient age, sex, BMI and race.

TAKEAWAY: The coaching app program has weak evidence for its superiority, but a larger effort and deeper look into patient subgroups could provide more substantial recommendations.


Text Message Medication Adherence Reminders Automated and Delivered at Scale Across Two Institutions: Pilot Testing the “Nudge” System

by Phat Luong et al.

WHAT: A proof-of-concept automated texting system that encourages patients to refill their prescriptions if nonadherence is detected.

TOPLINE DATA: Among 344 hypertension and/or diabetes patients from the VA and the Denver Health and Hospital Authority (DHHA), Via an EHR integration, the system identified seven-or-more-day gap in medication refill among 61% of patients, and sent them a text message. Of these, 37% from the VA and 12% from the DHHA responded to the texts saying they had refilled their medications, and only four patients across the entire study replied to the system asking it to stop sending reminders.

TAKEAWAY: Medication adherence detection and text message-based intervention can be integrated in the EHRs of large health system. The next step is to scale and measure impact.


Artificial Intelligence Finds Myocardial Infaction in ECG More Accurately Than Cardiologists

by Hisaki Makimoto et al.

WHAT: A direct comparison of a AI neural network’s ability to spot myocardial infarction patients from nine-lead ECG images compared to nine human cardiologists.

TOPLINE DATA: The neural network’s accuracy went as high as 84%, which was significantly higher than that of the cardiologists’ 64% accuracy (< .001). Receiver-operator-characteristic analyses found a 0.89 AOC for the AI. Cutting the resolution of the ECG images did not greatly affect the neural network’s accuracy compared to raw images.

TAKEAWAY: Don’t count out the possibility of AI-assisted ECG interpretation any time soon.