Cost-effectiveness of Cardiac Telerehabilitation With Relapse Prevention for the Treatment of Patients With Coronary Artery Disease in the Netherlands

Key Points Question Is cardiac telerehabilitation with relapse prevention cost-effective compared with center-based cardiac rehabilitation for the treatment of patients with coronary artery disease? Findings In this economic evaluation of data from 300 participants with coronary artery disease enrolled in the SmartCare-CAD randomized clinical trial, patients who received cardiac telerehabilitation with relapse prevention vs traditional center-based cardiac rehabilitation experienced comparable quality of life and nonsignificantly lower cardiac-associated health care costs and non–health care costs. Meaning This study found that cardiac telerehabilitation with relapse prevention was likely to be cost-effective compared with center-based cardiac rehabilitation and may be used as an alternative to center-based cardiac rehabilitation among patients with coronary artery disease.


Introduction
Cardiac rehabilitation (CR) is recommended in European Society of Cardiology clinical guidelines (class IA) 1 for the treatment of patients with coronary artery disease (CAD) because it has been associated with reductions in cardiovascular mortality and morbidity and increases in quality of life (QOL) [2][3][4] as well as substantial decreases in health care costs. 5,6 Moreover, CR intervention has been found to be cost-effective compared with no CR intervention, even in the modern era in which CAD is being successfully treated with optimized medical therapy and coronary revascularization. 7 However, participation rates in current center-based CR programs remain low 8,9 despite efforts to increase their use. 10 Cardiac telerehabilitation (CTR) has been reported to be a safe and beneficial alternative to traditional center-based CR 11,12 ; 3 small-scale studies [13][14][15] and 2 recent systematic reviews 7,16 found that CTR may be cost-effective compared with center-based CR. However, larger-scale studies, including comprehensive cost-effectiveness analyses, are needed. 7,16,17 The SmartCare-CAD (Effects of Cardiac Telerehabilitation in Patients With Coronary Artery Disease Using a Personalized Patient-Centred Information and Communications Technology Platform) randomized clinical trial 18 evaluated the effectiveness of a novel extended CTR intervention, which included video consultation and relapse prevention through on-demand coaching, compared with traditional center-based CR. The SmartCare-CAD study, which was the largest clinical trial of CTR to date, demonstrated that CTR with relapse prevention and center-based CR were equally effective, and both approaches led to sustained increases in physical activity levels, exercise capacity, and QOL after 1 year. 19 The current economic evaluation performed a comprehensive cost-utility analysis of data from the SmartCare-CAD clinical trial. We hypothesized

Study Design
The SmartCare-CAD study was a prospective randomized clinical trial of patients who received CR at the Máxima Medical Center, Eindhoven/Veldhoven, which serves 2 general hospitals in the Netherlands. The trial protocol was reviewed and approved by the institutional review board of the Máxima Medical Center, and the clinical trial was registered in the Netherlands Trial Register (registration number: NL5001). All participants provided written informed consent before study entry, and the SmartCare-CAD clinical trial was conducted according to the Declaration of Helsinki. 20 The SmartCare-CAD protocol has been described in detail elsewhere. 18 The present economic evaluation comprised a cost-utility analysis of data from the SmartCare-CAD study, comparing the cost-effectiveness of 3 months of CTR followed by 9 months of relapse prevention with the costeffectiveness of traditional center-based CR. This study followed the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) reporting guideline for economic evaluations of health care interventions. 21

Study Population, Randomization, and Blinding
The study included 300 patients who received care at the Máxima Medical Center between May 23, 2016, and July 26, 2018. All patients were entering phase 2 of outpatient CR for the treatment of CAD (ie, patients with stable CAD, patients with an acute coronary syndrome, and/or patients who had received coronary revascularization). Patients were eligible if they were referred for the exercise training module of the CR program. 22 After baseline measurements were obtained, the investigator (R. B.) randomly allocated patients on a 1:1 ratio to receive CTR (intervention group) or center-based CR (control group) using computerized block randomization. The investigator, supervising health care professionals (one of whom was H. K.), and patients were not blinded to group allocation because of the nature of the intervention. Patients were followed up for 1 year (until August 14, 2019).

Exercise Training and Other Core Components
The exercise training module in both groups was composed according to evidence-based algorithms. 23 Module content was determined individually and consisted of aerobic, functional, and resistance training. Other CR core components (eg, psychological counseling) occurred at the outpatient clinic for both groups.

Control Group
Patients in the center-based CR control group participated in group-based training sessions under the supervision of physical therapists and exercise specialists (Ն1 supervisor per 5 patients). Aerobic training was prescribed with a frequency aimed at two 60-minute sessions per week. Because exercise training programs were individually tailored, 23 the number and intensity of supervised sessions varied among patients.

Intervention Group
After allocation to the CTR intervention group, patients received access to a web-based application and were provided with a wrist-worn heart rate monitor (Mio Alpha; Physical Enterprises, Inc) and a hip-worn triaxial accelerometer (ActiGraph wGT3x-BT; Actigraph, LLC). The exercise training module began with 6 supervised group-based sessions, which were similar to those provided to the control group. Exercise training was then continued at home unless the patient preferred otherwise. Weekly video consultations with the physical therapist through the web-based application were scheduled until individual goals were achieved or the program was completed and evaluated. 23

Telemonitoring and Relapse Prevention
Patients in the intervention group uploaded sensor data to the web-based application at least once per week and reviewed these data during video consultations. Weekly telemonitoring guidance was concluded after 3 months; patients were then instructed to continue using their sensors and uploading data until study completion after 12 months. For this novel extension comprising on-demand coaching, the investigator evaluated sensor data every 4 weeks until study completion and notified the patient's physical therapist when relapses occurred, which were defined as (1) nonadherence to the intervention (no data uploaded for Ն4 weeks), (2) reduced exercise training (a decrease of Ն50% in mean weekly training minutes), or (3) reduced physical activity (a decrease of Ն50% in mean time spent per day in moderate to vigorous activity). In cases of relapse, a video consultation was scheduled, in which exercise and physical activity targets and reasons for relapse were evaluated.

Utilities and Quality-Adjusted Life-Years
Quality of life was assessed using 3 measures. The Dutch versions of the EuroQol 5-Dimension 5-Level survey (EQ-5D-5L) and the EuroQol Visual Analogue Scale (EQ-VAS) 24 were administered at 3, 6, 9, and 12 months after baseline, and the Dutch version of the MacNew Heart Disease Health-Related Quality of Life Questionnaire (KVL-H) 25 was administered at baseline and at 3 and 12 months after baseline.
Utilities were then calculated to measure patient QOL on a scale of 0 to 1, with 0 indicating worst health and 1 indicating best health. We calculated EQ-5D-5L utilities for the Netherlands using Dutch tariffs 26 and transformed the values from the EQ-VAS to a utility scale using the following power transformation: (1) the EQ-VAS value was divided by 100, (2) the quotient was subtracted from 1, (3) the difference was then subtracted from 1, and (4) the final difference was raised to the power of 1.61. 27 Because no EQ-5D-5L measurements were available at baseline, we converted the KVL-H scores at baseline to EQ-5D-5L utilities using mapping 28 (eFigure 1 in the Supplement). The KVL-H utility at baseline and the utilities from the EQ-5D-5L and EQ-VAS at the other available time points were used to calculate the mean utility for each quarter of the year and for the whole year after randomization to generate the area under the utility curves, equaling quality-adjusted life-years (QALYs).

Costs
We estimated the societal costs for a follow-up period of 1 year; therefore, costs were not discounted.
We converted costs to 2020 price levels (in euros) using the Dutch consumer price index (to convert to US dollars, euro values were multiplied by 1.142, which was the mean exchange rate in 2020). 29 We differentiated between cardiac health care costs, noncardiac health care costs, and non-health care costs. We also distinguished between costs from CAD and CAD-associated comorbidity (classified as cardiac health care costs) and costs from other health conditions (classified as noncardiac health care costs) to isolate the impact of CTR and center-based CR for cardiac health care costs. Patients reported their medical consumption, including general health care visits, hospital visits and treatments, hospital admissions, medications received, and home care, using an adjusted version of the Medical Consumption Questionnaire from the Institute for Medical Technology Assessment. 30 We used Dutch standard prices for health care use, 31 supplied with mean prices for health care use from the Dutch Health Care Authority, 32 to calculate costs. The numbers of centerbased and home-based exercise training sessions were registered in case report forms. An overview and description of CTR and center-based CR intervention costs are available in eTable 1 in the Supplement.
Patients reported presenteeism and absenteeism at work using the Productivity Cost Questionnaire from the Institute for Medical Technology Assessment, 33  Technology Assessment, 34 which together constituted non-health care costs. We applied a limit of 18 h/d spent on informal care to account for 6 h/d that caregivers may have spent addressing their own personal needs. 35 The hours of absenteeism from work during the entire 1-year follow-up period were calculated using the friction cost method, which calculates productivity costs by multiplying the standard mean hourly wage by the period it takes for an organization to restore production levels to normal after an employee is absent from work because of illness. The friction period consisted of 85 days at a standard mean hourly wage of €38.09 ($43.50) per hour. 31

Statistical Analysis
The intention-to-treat principle was applied to all analyses. Approximately 20% of items were missing for the EQ-5D-5L, 15% were missing for the KVL-H, 15% were missing for the EQ-VAS, and model also included cardiac health care costs and QOL measured by the EQ-5D-5L), and (4) all societal costs (model included cardiac health care costs, noncardiac health care costs, non-health care costs measured using the friction cost method, and QOL measured by the EQ-5D-5L).
To account for the statistical uncertainty of differences between costs and QALYs, we first applied nonparametric bootstrapping with 1000 replications. Next, we plotted cost-effectiveness acceptability curves, which represent the probability that a strategy is cost-effective as a function of WTP. At a probability of 50%, no preference for either strategy (eg, CTR vs center-based CR) exists.
We then performed a value of information analysis to estimate the net gain from additional research compared with the cost of making the wrong reimbursement decision based on the current study results. Following the Wilson methods, 36 we estimated the expected value of perfect information and the optimal sample size for a future study. We estimated fixed study costs (ie, costs of performing a future study, including personnel costs and maintenance of the ICT platform) at €350 000 ($400 000) and variable study costs (ie, mean costs of intervention and control training) at €190 per participant ($217 per participant) based on the current costs of the SmartCare-CAD clinical trial. We assumed a population of 60 000 patients in the Netherlands per year, 50% of whom were eligible for telerehabilitation, and we applied 3% discounting for a 5-year period.  Figure 1.

Exercise Training and Relapse Prevention
In

Utilities and QALYs
The QOL for patients receiving CTR vs center-based CR was similar throughout the study period, as measured by both the EQ-5D-5L (eg, 0.859 for CTR vs 0.850 for center-based CR at 3 months) and the EQ-VAS (eg, 0.885 for CTR vs 0.878 for center-based CR at 3 months) (Figure 2). The difference in QALYs between the CTR and center-based CR groups was not significant in any of the 4 quarters for both the EQ-5D-5L (  CTR measured by EQ-5D-5L CR measured by EQ-5D-5L CTR measured by EQ-VAS CR measured by EQ-VAS Utilities according to the Dutch versions of the EuroQol 5-Dimension 5-Level survey (EQ-5D-5L) and the EuroQol Visual Analogue Scale (EQ-VAS). CR indicates cardiac rehabilitation; CTR, cardiac telerehabilitation.

Cost-Utility Analysis
In the base-case cost-utility analysis (including cardiac health care costs and non-health care costs using the friction cost method), the incremental QALYs for CTR compared with center-based CR were spread around 0, meaning the cost-effectiveness of CTR vs center-based CR was comparable. The incremental societal costs were often lower for CTR because most data points had an incremental cost of €0 or lower ( Figure 3A). Therefore, the probability that CTR was cost-effective compared with center-based CR ranged from 69% at a WTP of €0 ($0) to 86% at a WTP of €215 000 ($245 530), and the probability was 84% at a disease severity-adjusted WTP of €20 000 ($22 840) in the Netherlands. 37 In addition, in the sensitivity analyses, the probabilities that CTR was costeffective compared with center-based CR all remained higher than 50%, irrespective of WTP ( Figure 3B). In other words, the difference in costs was mostly in favor of CTR, despite being statistically nonsignificant; therefore, CTR was likely to be cost-effective compared with centerbased CR, although uncertainty remained.
The value of information analysis revealed that the optimal sample for a future randomized clinical trial would be 1600 participants per arm (eFigure 3 in the Supplement), and the total expected value of perfect information per patient would be approximately €558 ($637). This finding means that after eliminating uncertainty in future research, the expected improvement in the net monetary benefit would be €558 ($637) per patient. A, The cost-effectiveness plane for the base-case analysis included cardiac-associated health care costs, non-health care costs according to the friction cost method (FCM), and quality of life (QOL) measures from the EuroQol 5-Dimension 5-Level survey (EQ-5D-5L). To convert to US dollars, multiply by 1.142, the mean exchange rate in 2020 (eg, €5000 is equivalent to $5710, €10 000 is equivalent to $11 420, €15 000 is equivalent to $17 130, and €20 000 is equivalent to $22 840). B, The base-case analysis included cardiacassociated health care costs, non-health care costs according to the FCM, and QOL measures from the EQ-5D-5L. Sensitivity analysis 1 included cardiacassociated health care costs and QOL measures from the EQ-5D-5L. Sensitivity analysis 2 included cardiacassociated health care, non-health care costs according to the FCM, and utility measures from the EuroQol Visual Analogue Scale. Sensitivity analysis 3 included cardiac-associated health care costs, productivity costs according to the human capital method, and QOL measures from the EQ-5D-5L. Sensitivity analysis 4 included total health care costs, non-health care costs according to the FCM, and QOL measures from the EQ-5D-5L. The dashed horizontal line at 50% probability represents the point at which no preference for either strategy (center-based cardiac rehabilitation [CR] or cardiac telerehabilitation [CTR]) exists. To convert to US dollars, multiply by 1.142, the mean exchange rate in 2020 (eg, €50 000 is equivalent to $57 100, €100 000 is equivalent to $114 200, €150 000 is equivalent to $171 300, and €200 000 is equivalent to $228 400).

Discussion
This economic evaluation of data from the SmartCare-CAD clinical trial found that a novel CTR intervention that incorporated relapse prevention was likely to be cost-effective compared with center-based CR at a disease severity-adjusted willingness to pay of €20 000 ($22 840)  However, besides intervention costs, only hospital service use and medication costs were estimated, the latter of which revealed a significant difference in favor of CTR. 14 Compared with the Telerehab III clinical trial, 15 the present analysis of data from the SmartCare-CAD study revealed similar cardiac health care cost savings in favor of the intervention group but no between-group differences in QALYs. In the Telerehab III study, in which CTR was provided in addition to center-based CR, QALYs increased and health care costs were lower during 2 years of follow-up (cost savings of €878 [$1003] in favor of the intervention group) compared with center-based CR alone. 15 The main reason for these cost savings appeared to be the difference in cardiovascular rehospitalizations (32 in the intervention group vs 60 in the control group), which may also have had consequences for the QOL of participants. This difference in rehospitalizations could be associated with the fact that participants in the Telerehab III study had higher residual cardiovascular risk; thus, patients in the intervention group may have experienced greater benefit from an extended CR program. The results of the Telerehab III clinical trial suggest that prolonged CR interventions may have added value among higher-risk populations, both in terms of benefits and costs.
Our finding that CTR was likely to be cost-effective compared with center-based CR was consistent with 2 recently published systematic reviews. 7, 16 Shields et al 7 reported that losses because of absenteeism and presenteeism were less relevant among participants receiving CR owing to their older ages (ie, approaching retirement). However, given that 60.0% of participants in the SmartCare-CAD clinical trial were still employed (and their ages were comparable with those of other CTR study populations) and that unpaid labor has substantial implications for societal costs (regardless of retirement age), the application of a societal perspective in our cost-utility analysis served as a strength, and this approach may be useful for future analyses.
Furthermore, we distinguished between cardiac and noncardiac health care costs. Although it may be difficult for patients to differentiate between health care use associated with cardiac disease vs noncardiac conditions, the high variability in noncardiac health care costs highlights the importance of separating these cost items in a comprehensive cost-effectiveness analysis.
Our findings add to the increasing body of evidence of the cost-effectiveness of CTR, suggesting that CTR is a suitable alternative to center-based CR. Yet, larger-scale analyses are still needed. 7 The value of information analysis performed based on our study results suggested that the optimal JAMA Network Open | Cardiology sample for a future study would be 1600 participants per arm. Future research may support more rapid implementation of CTR interventions as part of regular care as well as the reimbursement of CTR by health care insurers and the inclusion of CTR in national guidelines. 38 Home-based CR or CTR interventions are currently used by a small number of patients, and these interventions may be successful in increasing CR participation rates in hospitals or CR centers with limited facilities.
Therefore, implementation studies are needed to assess the ways in which implementation of CTR could be facilitated in various health care settings for the treatment of different patient populations. 17 Although CTR interventions in their current form were cost-effective compared with center-based CR, existing CTR interventions may be adapted to appeal to patients with higher residual cardiovascular risk and/or patients who are underrepresented in both center-based CR and CTR programs. These adaptations may not only enhance the implementation of CTR in regular care, but enable researchers to evaluate whether CTR is associated with increased CR participation rates.

Limitations
This study has several limitations. First, health care consumption questionnaires are subject to recall bias. To account for potential recall bias, we calculated the intervention costs using case report forms, and we cross-checked ambiguous questionnaire responses with participants' case report forms.
Second, the current economic evaluation had a horizon of 1 year. However, because the CTR intervention occurred in the first 3 months of the study and QOL remained stable among both groups during the 9 months thereafter, we did not expect to observe between-group differences in the QOL groups after the 1-year study period. Third, the study was conducted at a single CR center serving 2 general hospitals in the Netherlands, which means that caution is warranted in the generalization of this study's findings to other settings, especially those outside the Netherlands.

Conclusions
The comprehensive cost-utility analysis performed in this economic evaluation found that a CTR intervention with relapse prevention was likely to be a cost-effective alternative to conventional center-based CR among patients with CAD. Together with the findings of previous economic analyses, this study's results may be used to further support the implementation of CTR in regular practice.