Efficacy of an Internet of Things-based system for cardiac rehabilitation monitoring: insights from the IntellIoT pivotal trial in heart failure patients

Abstract

Aims

Digital health solutions targeted to remote clinical monitoring are constantly gaining ground in cardiovascular care. However, evidence regarding their impact on cardiac rehabilitation efficiency in heart failure (HF) patients is relatively limited. In this study, conducted in the context of the IntellIoT project, we evaluated the effect of a purpose-designed Internet of Things (IoT)-based patient monitoring system on cardiac rehabilitation outcomes in a cohort of HF patients.

Methods and results

Nineteen clinically stable HF patients were enrolled in the study, which consisted of a 12-month standard-of-care run-in phase and a remote follow-up phase of equal duration, whereby an IoT-based e-health system was provided to study subjects. Device-derived data transmission was facilitated by a mobile phone application, coupled with a web-based platform accessible to study physicians. Study endpoints were (i) patient adherence rates to e-health system use and their associations to key clinical parameters, (ii) the degree of change in physical activity, and (iii) total time dedicated by physicians to enrolled patients’ care with and without the aid of the e-health system. Baseline-to-peak increase in daily step count was calculated at 23.34%. System use was associated with a decrease in time dedicated by physicians to enrolled patients’ care. A significant negative correlation was observed between age and progressive drop-in adherence rate to system use (r = −0.5722, P = 0.02).

Conclusion

Internet of Things-based healthcare constitutes a promising approach in HF patients’ rehabilitation, whereas elderly patients might constitute the population most likely to benefit. However, larger, randomized studies are required to confirm our findings.

Graphical Abstract

Graphical Abstract.

Introduction

Regular physical activity deserves particular focus in patients with heart failure (HF) and exercise training carries a Class I indication in international guidelines.^1,2 Nonetheless, satisfactory adherence rates remain suboptimal, which is largely attributable to the unavailability or suboptimal utilization of comprehensive cardiac rehabilitation (CR) programmes.³

In this context, digital health applications targeted to remote clinical monitoring and to motivate patients to adopt lifestyle recommendations are constantly gaining ground in cardiovascular care.⁴ Among technologies that facilitate remote interaction with attending physicians, wearable devices could deliver one or more components of a CR programme to patients in a home-based setting. However, evidence regarding the effectiveness of this approach in terms of an increase in physical activity in HF patients is relatively limited.⁵

IntellIoT [Intelligent, distributed, human-centred, and trustworthy Internet of Things (IoT) environments] is a Pan-European project focusing on developing integrated, distributed, human-centred, and trustworthy IoT frameworks, with federated learning, trustworthiness, and human in the loop as key pillars. The aim of the healthcare use case of the project was to develop an IoT environment that could facilitate home-based remote care in patients with HF, thus helping them safely re-immersing in their normal environment.

Methods

Study population

We included HF patients who were enrolled in a CR programme. Key eligibility criteria included stable clinical status—defined as no hospitalization or emergency room presentation for a primarily cardiac cause—for at least 3 months before enrolment, being on optimal medical or, if indicated, device HF therapy, as well as the absence of contraindications for exercise training or of any condition that could prove prohibitive for efficient use of a mobile application. The trial consisted of a 12-month run-in phase, whereby the patient’s follow-up was conducted without the use of the system, and a 12-month remote follow-up phase. Study procedures at enrolment included quality of life assessment by the completion of the validated Minnesota Living with Heart Failure Questionnaire (MLHFQ), as well as cardiopulmonary exercise testing.

After completion of the standard-of-care period, all patients were provided with an IoT-based e-health system constituted by (i) a set of medical devices with wireless connectivity (pulse oximeter: Beurer PO60; weight scales: Beurer BF600; thermometer: Beurer FT95; blood pressure monitor: Omron M7 IntelliIoT), (ii) a state-of-the-art smartwatch (Samsung Galaxy Watch4), and (iii) a latest-generation smartphone (Samsung Galaxy S21FE 5G), to which a dedicated mobile patient application was installed. The latter was coupled to a web-based platform designed for physician use, thus forming the patient–physician interface. The effect of system use on patient physical activity levels, as well as the on-time dedication by the physicians to enrolled patients’ care, was selected as key system performance indices.

The study was conducted by the Declaration of Helsinki, the protocol was approved by the Hospital Ethics Committee, and patients gave written informed consent.

Study outcome measures

To fulfil our study’s objectives, the following endpoints were defined: the degree of change in physical activity from baseline, throughout the duration of the remote follow-up study phase, total time dedicated by physicians to enrolled patients’ care with and without the aid of the e-health system, and patient adherence rates to e-health system use during the first study month and by the end of follow-up and their associations to key clinical parameters.

Device-derived data flow and processing

Medical device-derived data—i.e. blood pressure and heart rate values, body weight, oxygen saturation, and body temperature—are wirelessly transmitted to the mobile phone application, which—in parallel to acting as the phone-based data repository—also includes federated learning worker and notification-generating components. For retrieval of smartwatch-derived physical activity level, the system’s mobile application was linked to the watch manufacturer’s application indirectly through Google Fit, whereas the upload of one-lead electrocardiogram tracings into the system was performed actively by the patients.

The system automatically analyses data on the phone and can generate two distinct types of notifications:

1. Standard-of-care recommendations or alerts to patients, which are triggered by biological parameter value thresholds that are pre-specified in an individualized manner; this notification type is delivered to the patient’s phone in real time, while the physician can view and validate it by logging into the web-based platform (Figure 1).

2. Artificial intelligence (AI)-generated predictions of mean resting heart rate in the next 7 days; importantly, notifications of this category were only accessible to the physician for validation for developmental purposes but were not transmitted to patients.

Figure 1.

Illustration of device-derived data flow and patient–physician interaction through the e-health system. Blue-framed panel: Device-derived data flow and processing. Black arrows: Flow of information that was used in patient management. Red arrows: Flow of data used only for framework testing and developmental purposes. Green arrows: Physician–patient and physician–system interactions. Yellow arrows: Transmission of notifications. Green-framed panel: An example of exercise prescription via the web platform according to the frequency–intensity–time–type principle (time is determined in seconds). Yellow-framed panels: Examples of system-generated notifications as can be viewed by the patient on the mobile application (upper panel) and by the physician on the web-based platform (lower panel). Black-framed panels: Physician’s view of device-derived data through the web-based platform (left panel: blood pressure, heart rate, and body weight data; right panel: graphical display of physical activity data). Local AI, local artificial intelligence component; global AI, global artificial intelligence component; FL, federated learning.

All data were transmitted to the hospital-based patient data repository over secure connections and are available to the physician on demand via the password-protected web platform. The physician acts as human in the loop, by evaluating device-derived data according to each patient’s clinical condition, validating the appropriateness of notifications generated, and, ultimately, providing advice to patients or amending their management plan as per clinical judgment. Study participants were instructed to perform daily vital signs and weight assessments and transmit data via the mobile application.

Framework facilitating artificial intelligence-generated predictions

At the initial stages of framework development, a healthcare AI model was implemented, which provides its output to the mobile application, which functions as the local AI. The local AI uses rule-based algorithms, while the healthcare AI model uses neural networks, for the detections and predictions. The local AI sends training results to the global AI, where the results are aggregated to update the base model (Figure 1).

Study outcome evaluation

Mean daily step count during the first study month was defined as the baseline physical activity level. Subsequently, after activation of exercise notifications, the mean daily step count was assessed for each patient on a monthly basis until the completion of follow-up.

Time dedicated by study physicians to patients’ cardiovascular care was estimated by accounting for 30 min for every outpatient department visit for comprehensive clinical cardiovascular evaluation, 10 min for each visit for implantable defibrillator interrogation only, and 120 min for each emergency department visit. The time approximations outlined above for each type of scheduled visit are based on the typical duration of clinic appointments. In contrast, the estimated time required for completing an emergency department visit includes the average time needed for emergency laboratory workups. Regarding time dedicated to evaluation of device-derived data and system-generated alerts, 13 min were accounted for each week after initiation of device-facilitated follow-up; this value was determined by directly measuring the time a physician requires for a comprehensive evaluation of all device-derived values and notifications of a 1-week period.

Subjects’ adherence to the use of the system was defined as the percentage of days with at least one successful transmission of data via the mobile application—the minimum degree of interaction of a patient with the e-health system—on the total number of days. Adherence rates were assessed every month throughout the 12-month-long remote follow-up study phase and subsequently evaluated for any correlations to subjects’ baseline clinical characteristics, as well as to their score on MLHFQ.

Statistical analysis

Continuous and numerical patient variables at baseline and at the end of follow-up were compared with paired t-tests or non-parametric tests, as appropriate. Associations between continuous variables were assessed with Pearson’s correlation coefficient.

Results

Nineteen patients were included in our pivotal study. Key clinical characteristics of our study population are summarized in Tables 1 and 2.

Table 1.

Baseline subjects’ baseline clinical characteristics

Clinical parameters	Patients (n = 19)
Age (years)	49.0 (24–78)
Men (%)	73.7%
Arterial hypertension (%)	36.8%
Coronary artery disease (%)	26.3%
Atrial fibrillation (%)	10.5%
Diabetes mellitus type 2 (%)	10.5%
Implantable cardioverter defibrillator (%)	36.8%
Chronic kidney disease	21.1%

Table 2.

Key clinical parameters at baseline and at end of follow-up

	Baseline	End of follow-up	P-value
Mean LVEF (%)	43.0 ± 5.1	44.1 ± 5.8	NS
Mean SBP (mmHg)	120.1 ± 14.4	119.2 ± 13.8	NS
Mean DBP (mmHg)	75.6 ± 8.2	76.1 ± 8.4	NS
Mean resting heart rate (b.p.m.)	69.1 ± 6.3	69.9 ± 6.5	NS
Mean body mass index (kg/m2)	32.91 ± 4.9	32.2 ± 5	NS
MLHFQ score	23.27 (8–47)	22.7 (8–45)	NS
Peak VO2 (mL/kg/min)	22.77 (14.8–36.1)	23.02 (14.0–39.1)	NS

LVEF, left ventricular ejection fraction; SBP, systolic blood pressure; DBP, diastolic blood pressure; MLHFQ, Minnesota Living with Heart Failure Questionnaire; peak VO₂, maximal oxygen consumption as assessed by ergospirometry; ns: non-significant.

None of the subjects did a contraindication to regular aerobic physical activity arise after enrolment. The mean increase in daily step count was calculated at 23.34% (−20.33 to 103.63%). Only three patients failed to achieve any increase in daily step count average.

The total actual time dedicated by physicians to enrolled patients’ cardiovascular care during the run-in control period was calculated at 1770 min, whereas the total time spent for patients’ cardiovascular care plus system-related activities during the telerehabilitation phase was reduced at 1601 min (patients’ care: 850 min; time for data/alerts evaluation: 611 min; time for patient support on system use: 140 min). The estimated total time that would have been dedicated to patients’ care during the device-facilitated follow-up period without the system was calculated at 1750 min. In four patients, it is judged that unscheduled visits to the outpatients’ clinic and/or the emergency department were obviated.

Mean adherence to system use was estimated at 63.12 ± 30.45% during the first study month, whereas a decline to 53.97 ± 38.98% was evident by the end of the follow-up. Age was not shown to be negatively associated with adherence to the system; in fact, a slight trend towards positive correlation was observed at 1-month post-enrolment (r = 0.3312, P = 0.23). More importantly, a significant negative correlation was observed between age and absolute drop-in compliance rate from the first month of participation to the lowest value (r = −0.5722, P = 0.02). No statistically significant correlation was observed between adherence rates and physical activity levels or any other clinical parameters.

Discussion

This study is the first to attempt to develop an integrated, distributed, human-centred, and trustworthy IoT framework for monitoring HF patients. Despite the study's clear limitations, including non-randomized design and small sample size, our results suggest that mobile application-based telerehabilitation constitutes a promising approach in HF patients’ care, as it could aid them achieve a clinically relevant increase in physical activity. This is in agreement with existing literature.⁶ However, our findings underline that these effects may be transient, thus highlighting the importance of continued patient motivation and physician’s oversight.

Moreover, our work indicates that advanced age might not per se be prohibitive for endorsing digital solutions targeted to optimization of HF patient care; on the contrary, the positive correlation of age to long-term consistency in e-health application use—irrespective of functional status—implies that the elderly could constitute the subpopulation of HF patients who are most likely to benefit from telerehabilitation.

In addition, state-of-the-art components were implemented to ensure a secure exchange of medical information, which, to our knowledge, has not been a point of focus in similar studies. Lastly, the IntellIoT project has developed a sophisticated framework that can facilitate training AI models within federated learning solutions from a technological standpoint. Despite the small number of patients enrolled in our study, the system's architecture may be able to be leveraged in the future for prediction model training.

Conclusions

Mobile application-based healthcare constitutes a promising approach in HF patients’ rehabilitation and may reduce the time dedicated by physicians to patients’ care. However, it should be emphasized that larger, randomized studies are required to confirm this hypothesis.

Lead author biography

Maria E. Marketou is an associate professor of cardiology, Faculty of Medicine, University of Crete, Greece, the head of the Laboratory of Molecular Cardiology, School of Medicine, University of Crete, the head of the Hypertension and Cardiovascular Prevention outpatient clinic, and the head of the Cardio-oncology Outpatient Clinic. She has authored more than 130 peer-reviewed articles in PubMed (h-index Scopus 29, citations: 12 523). She is the past president of the Working Group of Essential Hypertension of the Hellenic Society of Cardiology. She is also a member of the European Society of Cardiology Committee for the European Examination of General Cardiology (EEGC) and a member of the Digital Health Committee of the European Society of Cardiology.

Funding

IntellioT (Intelligent, distributed, human-centred, and trustworthy IoT environments) project was an EU-funded project (grant agreement ID: 957218).

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Consent

All subjects provided written informed consent before initiation of any study-related procedures, and the study was conducted by the Declaration of Helsinki.