Capturing Physiological Data in Children With Heart Failure Using Wearable Digital Technology: Lessons From Pilot Project

Abstract

We conducted a proof-of-concept pilot project to evaluate the feasibility of collecting and describing physiological and activity data from pediatric patients with heart failure using an Apple Watch and a specially designed app. Five patients (12-17 years old) with significant heart failure, including 2 patients with left ventricular assist device, were enrolled. Heart rate and activity data from each patient were obtained and presented along with corresponding clinical status and trajectory. We believe that sharing these early, detailed data can help generate hypotheses and spur further study of this growing technology and its applicability in pediatric conditions including, but not limited to, heart failure.

Pediatric heart failure is a significant, burdensome, and complex condition associated with high morbidity and mortality. The estimated incidence of heart failure is 0.9-7.4 per 100,000 children.¹ Each year in the United States, tens of thousands of children are admitted for heart failure, with an in-hospital mortality of 7%.^2,3 In children with congenital heart disease admitted with severe heart failure, in-hospital mortality climbs to 26%.⁴ Of the sickest children on the waiting list for a heart transplant, 22% will die before a donor organ becomes available.⁵ Managing end-stage heart failure in children is particularly challenging due to the paucity of pediatric-specific data and evidence. Few scoring systems and tools are available to monitor and assess disease severity and functional status in pediatric heart failure.^6,7 Historically, a patient’s clinical status has been assessed by unreliable, subjective recall and brief, infrequent testing and encounters.

Wearable technologies have rapidly advanced in recent years. Reasonable-cost and portable on-person devices can now capture high-fidelity physiological and activity data, facilitate interventions, and may present opportunities to improve heart failure outcomes.^8,9 Adult heart failure studies have demonstrated that digital biomarkers from wearables may help identify patients with heart failure at greater risk for clinical decompensation.^10,11 Average step counts measured from wearables correlate with hemodynamic exercise testing performance,¹² patient-reported outcomes and health status,¹³ and all-cause mortality and cardiovascular events.¹⁴

Recently highlighted as a significant knowledge and research gap,¹⁵ it is unknown if and how wearables can be helpful in children with heart failure. To address this gap, we conducted a proof-of-concept pilot project to evaluate the feasibility of collecting and describing physiological and activity data from pediatric patients with severe heart failure using an Apple Watch.

Methods

We conducted a pilot feasibility study through the Advanced Cardiac Therapies Improving Outcomes Network (ACTION). ACTION is an international learning health system aimed at improving outcomes in children with heart failure.^16,17 The protocol was approved by ACTION’s centralized institutional review board at Cincinnati Children’s Hospital, which houses the data coordinating center.

We sought to include patients aged 12-19 years who were enrolled in the ACTION HF registry¹⁸ (admitted with acute decompensated heart failure) and/or the ACTION VAD registry (supported with a ventricular assist device).¹⁹ Patients in the intensive care unit at the time of potential enrollment were excluded. ACTION site investigators and research coordinators screened patients for eligibility. Consent and assent were obtained for each participant.

Subjects were asked to wear Apple Watches daily for at least 3 months. An app, “My ACTION Tracker,” was programmed to securely collect and transmit Apple HealthKit data from a subject-worn Apple Watch to the ACTION data coordinating center. From HealthKit, we primarily evaluated the subjects’ daily step count (measured number of steps the user has taken), daily resting heart rate (estimated by analyzing sedentary heart rate samples throughout the day, in beats per minute, bpm), and heart rate variability (standard deviation of the RR intervals of regular heartbeats, in milliseconds), which were obtained through the app. Apple Watch measures heart rate via photoplethysmography (based on fluctuations in light absorption), and its accuracy has recently been validated in pediatric patients.²⁰ Resting heart rate data obtained from the Apple Watch were compared with those recorded in the medical record.

Cases

Eight patients consented to the study across 4 ACTION centers. Three subjects had technical connectivity issues and were unable to transmit data successfully.

Five patients were successfully enrolled. Patient descriptions and important clinical events are summarized below. Their transmitted wearable data are presented in Figure 1, Figure 2, Figure 3. Importantly, the timing of clinical events is correlated with the overall month of study (x-axis of Figure 1, Figure 2, Figure 3).

Figure 1.

Daily step counts over time.

Figure 2.

Daily resting heart rate over time.

Figure 3.

Heart rate variability over time.

Patient 1

Patient 1 was a 15-year-old (118 kg) girl with newly diagnosed dilated cardiomyopathy and was admitted for acute decompensated heart failure. After being weaned off continuous intravenous milrinone and transferred to the step-down unit, she was enrolled in the study (at 0.4 months in Figure 1, Figure 2, Figure 3) and started wearing the Apple Watch. Over the first 2 weeks of wearing the watch, she showed gradual clinical improvement, was placed on guideline-directed medical therapies, and was discharged home (at 0.8 months). She has remained well compensated and has not been readmitted. Her step counts and heart rate variability increased over time as her clinical status improved. Despite inconsistent wearing of the watch over time, she continued to periodically have higher-than-previous step count days (>4000 steps/day) as an outpatient.

Patient 2

Patient 2 was a 17-year-old (87 kg) girl with known dilated cardiomyopathy who had been lost to follow-up and presented with acute decompensated heart failure. She was enrolled while receiving continuous intravenous milrinone (at 2.2 months in Figure 1, Figure 2, Figure 3) but continued to decline clinically. She underwent an Impella 5.5 implant (at 2.6 months) and a Heartmate 3 implant (at 3.8 months). Despite inconsistent wear, her daily step count peaks increased, especially around the time of discharge on Heartmate 3 (at 3.8 months in Fig. 1).

Patient 3

Patient 3 was a 17-year-old (95 kg) boy presenting with newly diagnosed cardiomyopathy, which required Heartmate 3 placement, and 1 month later, he was enrolled in the study (at 2 months in Figure 1, Figure 2, Figure 3). He recovered and was discharged home (at 3.1 months). Figure 1 shows the step counts associated with this patient’s clinical status and trajectory. He remained well supported on Heartmate 3, participating in a structured outpatient cardiac rehabilitation program, which correlated with step counts from 8000 to 14,000 steps/day. He subsequently underwent heart transplant (at 6.3 months in Figure 1, Figure 2, Figure 3), with a brief lapse in wear, and was discharged home (at 6.7 months) with a gradual increase in activity through an uncomplicated post-transplant course. There was a notable increase in resting heart rate (Fig. 2) and a decrease in heart rate variability (Fig. 3) after heart transplant, presumably from the denervated graft.

Patient 4

Patent 4 was a 16-year-old (54 kg) boy admitted with decompensated Fontan circulatory failure with reduced ejection fraction and was medically managed. He was enrolled in the study during his second week of admission (at 2.6 months in Figure 1, Figure 2, Figure 3). He underwent an uncomplicated heart transplant (at 3 months), was discharged home 10 days later, and has had an uncomplicated post-transplant course. Compared with his pretransplant admission data, daily step counts increased significantly after the transplant (Fig. 1).

Patient 5

Patient 5 was a 12-year-old (37 kg) girl with Fontan circulatory failure with reduced ejection fraction. She was enrolled as an outpatient (at 4.8 months in Figure 1, Figure 2, Figure 3) and remained compensated on oral medications. There were no hospitalizations or significant clinical events during this time frame (Figure 1, Figure 2, Figure 3).

In patients 1, 4, and 5 (without an left ventricular assist device [LVAD]) and patient 3 (previously implanted with an LVAD), the resting heart rate recorded from the Apple Watch closely matched (within several bpm) resting heart rates documented in the medical record.

Before LVAD implant, the Apple Watch heart rate data of patient 2 also closely matched heart rates in the medical record. However, after LVAD implant, the Apple Watch heart rates were initially significantly lower than the documented inpatient (electrical) heart rate. For instance, on postimplant day 10 (the first day the Apple Watch was worn after Heartmate 3), the Apple Watch recorded a resting heart rate of 55 bpm, though the lowest electrical heart rate in the medical record was 108 bpm. Of note, as the patient’s electrical heart rate decreased and the patient reached outpatient status, Apple Watch heart rates more closely matched electrical heart rates (similar to patient 3).

Discussion

In this report, we present novel pilot wearable data from pediatric patients with heart failure. Using a specially designed app, we securely collected digital health data from pediatric patients wearing an Apple Watch and temporarily associated them with each patient’s clinical status. We believe that sharing these novel, detailed data can help generate hypotheses and spur further study of this growing technology and its applicability in pediatric conditions including, but not limited to, heart failure.

Apple HealthKit allows users to collect and derive dozens of different “health data” from their devices. In collecting these wearable variables, we thought that daily step counts could potentially help us better understand our patients. This aligns with recently published data showing that daily step counts are associated with patient-reported health outcomes and hemodynamics during exercise testing in adult heart failure patients.^12,13 Our experience supports the potential for activity tracking as a research and clinical endpoint.²¹

As previously demonstrated,¹³ high or increased activity detected by wearables is likely to be a more specific and significant marker for improving health and functional capacity than low activity. Decreased step count is inherently much less specific and could reflect many other factors not directly or necessarily related to cardiovascular health such as wear time, poor weather, or orthopedic injury to name a few.

In addition to step counts, we found face validity in the wearable-measured resting heart rate and heart rate variability. The accuracy of Apple Watch heart rate data has recently been validated in a wide range of pediatric ages.²⁰ More data are needed to see if these data can be clinically valuable and correlate with functional status, medication compliance and tolerance, and heart failure outcomes.

Not surprisingly, we observed that in patients with a continuous flow LVAD and nonpulsatile circulation (patient 2), Apple Watch heart rate data measured via photoplethysmography were inaccurate and cannot be relied upon. However, if patients regain more native ventricular function and ventricular pulsatility, the Apple Watch heart rate data may better match the electrical heart rate. We hypothesize that the Apple Watch heart rate data of patient 3 correlated more with his electrical heart rate because they were farther out from LVAD and had more native contractility and pulsatility.

We primarily evaluated the step counts and heart rate data given their face validity, simplicity, and interpretability. Other derived Apple HealthKit propriety data, such as energy burned and running/walking speed, were transmittable and available. However, because these variables are complexly derived and nonstandard, they were outside the scope of this feasibility project.

This pilot study highlighted essential challenges and limitations to using wearables in pediatric heart failure. Data from wearables are most useful when patients wear their devices consistently. Nevertheless, even inconsistent, intermittent data from wearables may provide insights not otherwise afforded. For instance, high daily step counts, even if only sporadically obtained, may be a pertinent positive finding for good functional capacity and reassuring that the patient is not in a significantly decompensated heart failure state. Furthermore, heart rate variability and resting heart rate data can still be obtained even when the patient does not wear their device continuously.

It is important to consider the potential impact of wearable data on health inequity. Theoretically, wearables could provide important data and connectivity for patients living remotely with poor access, language barriers, or low health literacy. Still, wearables can potentially worsen health disparities, especially in the most at-risk population that, for instance, cannot afford a smartphone or a reliable connection. Equitable access for all patients must be of the highest priority for any future implementation of these technologies.

Low enrollment in this pilot study was multifactorial. Site investigators had limited bandwidth, given competing demands within the ACTION collaborative. In this pilot study, we could not offer additional incentives to subjects for participation. The app was solely developed for the Apple Watch; thus, we were only able to enroll patients or parents who had an Apple iPhone. The Apple Watch devices provided to centers required operating system updates, which took several hours and created a major hurdle. Even more cumbersome, institutional firewalls often block updates onsite. Three patients were enrolled but quickly lost connection and did not transmit wearable data due to the instability and failure of the custom app.

This pilot study has highlighted key areas for improvement. To start, larger studies should include incentives for participation. All stakeholders will need to derive value in the wearing of devices for more widespread, consistent, and lasting use. Positive feedback and rewards will need to be developed. For its next wearables project, ACTION will use a proven, more stable application/platform²² to incorporate data from a range of wearable products from Apple, Google/Fitbit, Garmin, and others. Additional qualitative data on patient experience and barriers to wear will be collected to inform strategies to improve adherence. With larger numbers, wearable data should be validated against other pediatric heart failure outcomes, including patient-reported outcomes. As we work toward collecting larger amounts of wearable data from patients, artificial intelligence and machine learning should be leveraged to make sense of the vast data. Extensive work is required to standardize measures, terminology, and reporting of wearable data.