Retrospective Comparison of Outcomes and Cost of Virtual Versus Center‐Based Cardiac Rehabilitation Programs Authors: Neil D Shah, Conor W Banta, Andrea L Berger, Andrea Hattenberger, Alex Zimmerman, Bryan E Martin, Edward Wu, Usnish Majumdar, H Lester Kirchner, Martin E Matsumura Published: 2024 ABSTRACT AbstractBackgroundDespite robust evidence supporting an association with improved outcomes in eligible patients, cardiac rehabilitation (CR) remains underused, with a minority of eligible patients participating. Virtual cardiac rehabilitation (VCR) has been proposed as an alternative to traditional center‐based CR (CBCR) to improve usage rates. However, data supporting the efficacy and cost‐effectiveness of VCR are limited. In the present study, we compared the outcomes and cost of a VCR versus traditional CBCR program.Methods and ResultsIn a retrospective cohort study comparing VCR versus CBCR, CBCR data were collected from a period of January 2018 to September 2023. VCR data were collected from program initiation in July 2021 to September 2023. Primary health outcomes measured were 1‐year mortality rates, recurrent myocardial infarction, all‐cause hospital readmission, and emergency department visits. Primary cost outcomes were analyzed as cost ratios related to VCR versus CBCR assessing total medical costs allowed, pharmacy costs, and total costs of care over the 12 months post‐CR enrollment. VCR was associated with a significant reduction in 1‐year all‐cause hospital readmission (incident rate ratio [IRR], 0.616 [95% CI, 0.489–0.777], P<0.001) and ED admission (IRR, 0.557 [95% CI, 0.452–0.687], P<0.001) at 1 year. The IRR of myocardial infarction and all‐cause mortality did not significantly differ between VCR and CBCR. In addition, VCR was associated with significant reductions in medically related (cost ratio, 0.814 [95% CI, 0.690–0.960], P=0.0144) and total costs allowed (cost ratio, 0.838 [95% CI, 0.725–0.970], P=0.0176).ConclusionsVCR is a viable alternative to CBCR with at least comparable efficacy and cost, and as such, represents a key mechanism for improving access to and participation in CR for eligible patients.Keywords: cardiac rehabilitation, readmission, virtual careSubject Categories: Cost-Effectiveness, Digital Health, Secondary Prevention Nonstandard Abbreviations and Acronyms center‐based cardiac rehabilitation cardiac rehabilitation virtual cardiac rehabilitation Clinical Perspective What Is New? Compared with traditional center‐based cardiac rehabilitation, a virtual cardiac rehabilitation program was associated with improvements in rates of hospital admission and emergency department visits in the year following initiation of either program. Compared with traditional center‐based cardiac rehabilitation, a virtual cardiac rehabilitation program was associated with improvements in rates of hospital admission and emergency department visits in the year following initiation of either program. These improvements were associated with a decrease in total cost of care for virtual cardiac rehabilitation patients versus those in center‐based cardiac rehabilitation. These improvements were associated with a decrease in total cost of care for virtual cardiac rehabilitation patients versus those in center‐based cardiac rehabilitation. What Are the Clinical Implications? Virtual cardiac rehabilitation appears to be an acceptable alternative to traditional center‐based programs; this may have a positive effect on access to cardiac rehabilitation across geographic and socioeconomic patient populations. Virtual cardiac rehabilitation appears to be an acceptable alternative to traditional center‐based programs; this may have a positive effect on access to cardiac rehabilitation across geographic and socioeconomic patient populations. Cardiac rehabilitation (CR) is a structured exercise and counseling program focused on the secondary prevention of cardiovascular disease. CR has been demonstrated to drive robust improvements in rates of all‐cause mortality and rehospitalization via improvements in physical fitness, heart function, and mental health. 1 , 2 Accordingly, CR is a Class 1 recommendation by the American Heart Association and American College of Cardiology for patients with acute coronary syndromes, patients who undergo coronary revascularization (including in nonacute coronary syndrome settings and after coronary bypass grafting), as well as a Class 2a recommendation for those with symptomatic heart failure. Despite broad applicability, CR remains underused. A recent national study of Medicare enrollees found that only 28.6% of eligible patients started a CR program, and only 7.9% completed a full 36‐session CR program. 3 The Million Hearts Initiative, a program of the Centers for Disease Control and Prevention, targets 70% CR participation, a figure that cannot be achieved without significant changes to the health care delivery ecosystem. 4 Although behavioral and educational interventions have been developed to improve the rate of referral, 5 , 6 there remain significant structural challenges to overcome: availability of CR facilities, distance to CR facility, out of‐pocket costs, lack of transportation, and language barriers. 7 A recent modeling study estimated that if all center‐based CR facilities operating before the COVID‐19 pandemic operated at maximal capacity, only 47% of eligible US patients could be accommodated. 8 Furthermore, a recent geographic survey revealed that approximately 14% of US adults older than age 65 live in a CR facility “desert,” with untenable ratios of eligible patients to existing facilities. 9 These access and capacity barriers were worsened by the COVID‐19 pandemic, which led to the closures of approximately 8% of all US CR facilities. 10 Postpandemic, there remains a sustained 14% decrease in CR use as compared with before. 11 These declines in use were most profound among patients residing in rural areas and in vulnerable communities, which further compounded existing disparities in CR use along lines of social disadvantage. 12 , 13 , 14 Several alternative CR delivery models have emerged as potential solutions to the problem of limited access to traditional center‐based CR (CBCR). The majority of these models use digital technologies to augment or replace components of CBCR in the patient's home. Although virtual CR (VCR) programs benefit from flexibility in delivering traditional CR exercise sessions, this flexibility presents an opportunity to offer a more comprehensive approach to secondary cardiovascular disease prevention, with many models including counseling on nutrition, medication adherence, comorbid mental health conditions, and tobacco cessation. 15 Previously described VCR programs at the Veterans Affairs Health System and Kaiser Permanente Southern California, for example, involve self‐guided exercise sessions completed in the home, risk factor‐oriented counseling delivered via a series of telephonic or audiovisual interactions with clinical staff and basic biometric monitoring via mobile app or smartwatch. 16 , 17 Elsewhere, “hybrid” models have emerged, combining exercise sessions performed at a CR facility with sessions performed at home. 18 VCR involves direct clinical supervision of exercise sessions performed at home via synchronous audio‐visual technology. 15 In the United States, the Veterans Affairs Healthy Heart Home‐Based CR program has demonstrated reductions in all‐cause mortality among those participating as compared with those not participating in the program. 16 The magnitude of this reduction may be comparable to prior clinical trials of CBCR. A cohort study of patients in Kaiser Permanente's Southern California VCR program revealed a 21% lower readmission rate versus those who underwent CBCR. 17 A similar cohort study at the University of California, San Francisco found that functional outcomes were comparable between VCR, CBCR, and a hybrid CR model. 19 Although these non‐CBCR models present opportunities to innovate and extend capacity, there remain limited data directly comparing the efficacy of home‐based CR to CBCR, especially when leveraging VCR. In the present study, we evaluated a novel VCR program across the Geisinger Health System, a large regional health system serving both urban and rural patients throughout Pennsylvania. Geisinger's CBCR program was temporarily closed during the COVID‐19 pandemic, during which a VCR program was launched. Subsequently, Geisinger's CBCR program was relaunched in 2022 and operated alongside the VCR program. In this pragmatic community context, we aim to compare the clinical efficacy and cost of VCR and CBCR. METHODS The data that support the findings of this study are available from the corresponding author upon reasonable request. The study was approved by the Institutional Review Board of Geisinger Health System. A waiver of informed consent was obtained due to the retrospective nature of the study design. Strobe cohort reporting guidelines were employed for the study. 20 Study Design and Data Source This retrospective cohort study across Geisinger Health System compared VCR and CBCR. Geisinger Health System serves a large portion of central and northeastern Pennsylvania with urban and rural settings, diverse socioeconomic backgrounds, and variable access to hospitals where CBCR programs are located. During the COVID‐19 pandemic, CBCR programs were closed, and VCR became available in July 2021. The study investigated the impact of VCR on common cardiovascular disease risk factors and clinical outcomes as compared with CBCR. CBCR data were collected from a period of January 2018 to September 2023. VCR data were collected from its initiation in May 2021 to September 2023. CBCR data, including cohorts before and after VCR implementation, were pooled due to the similarity in baseline characteristics. Prospective data were collected on patients enrolled in Geisinger CR via electronic medical records, individualized treatment plans, and VCR session data. Study Population Patients were included in the study if they were ≥18 years of age and had a qualifying diagnosis for enrollment into CR, including acute myocardial infarction within the preceding 12 months, coronary artery bypass grafting, current stable angina pectoris, heart valve repair or replacement, percutaneous transluminal coronary angioplasty or coronary stenting, heart or heart‐lung transplant, stable congestive heart failure with left ventricular ejection fraction ≤35%, or other cardiac conditions as specified through Medicare National Coverage Determinations. Exclusion criteria included severe arthritis, dementia, Parkinson's, and hypertrophic cardiomyopathy. Patients were excluded if they did not complete at least 1 session or had interruptions in their CR series that resulted in CR sessions spanning beyond 1 year following the enrollment date. For patients who were enrolled more than once into CR during the study period, only the initial CR series was included in the analysis. Outpatient Cardiac Rehabilitation Programs Patients were referred to CR through a provider‐placed referral order or via chart review identifying patients with eligible diagnoses. From January 2018 to March 2020, patients referred for CR were offered CBCR. From March 2020 to June 2022, CBCR locations were closed due to the COVID‐19 pandemic, and VCR was offered as the sole CR modality from July 2021 to June 2022. Starting in June 2022 until the end of the data collection period in September 2023, both options were offered, and providers were permitted to specify a request for VCR, CBCR, or no preference. For the latter period, if a provider preference was not stated on the referral order, patients were presented with both options. All referrals were reviewed for CR exclusion criteria and adequate insurance coverage for their selected CR modality. The Geisinger CBCR program operates across multiple Geisinger sites. The program is structured to include 3 sessions per week for a total of 12 weeks using a traditional CBCR structure based on guidelines of the American Association of Cardiovascular and Pulmonary Rehabilitation. This program consists of 1‐hour sessions, 3 times per week. The personnel present during CR sessions are an exercise physiologist and a registered nurse who are present for the duration of the session. There is a medical director who oversees the administration of the program. The program duration target is 36 sessions. The Geisinger VCR program is a partnership with Recora (New York, NY). Recora's VCR platform is a comprehensive virtual program that includes both technology and services. After establishing a VCR consult, patients were mailed a kit including an internet‐connected tablet with the Recora app preloaded, blood pressure and heart rate monitor, and exercise bands. Patients also had the choice to use their own smart device and download the app via the Apple App Store or Google Play Store. Patients were then onboarded via a real‐time video visit followed by up to 36 real‐time video sessions, including monitored group exercise sessions, nutrition counseling, risk factor modification, and psychosocial/well‐being discussions. Sessions occurred 2 to 3 times per week for 12 to 18 weeks. Nonphysician staff included exercise physiologists, dieticians, and ancillary staff, all provided by Recora. A standardized individualized treatment plans was developed at the first visit and updated every 30 days until program completion. A Geisinger cardiologist was “available and accessible for medical consultations and emergencies” as part of virtual supervision responsibilities. The program concluded with a final set of exercise tests and questionnaires to document progress. Given that exercises in a CBCR program typically use treadmills or cycle machines, the VCR exercises were modified for the home environment without the need for special equipment to minimize patient cost or logistics barriers. VCR exercise programs focused on exercise goals similar to those in CBCR: heart rate, strength, function, and endurance. Sample exercises in the home include stepping in place, sit‐to‐stands, leg raises, and resistance band exercises. Geisinger's costs associated with the VCR program are a bundled fee per session delivered by Recora. These fees include the technology platform, recovery kit, technical support, and clinical staffing assistance. No other Geisinger staffing or infrastructure costs were required to enable the VCR program. Outcome Measures Primary health outcomes measured beginning from the first CR session were 1‐year mortality, recurrent myocardial infarction, all‐cause hospital readmission, and emergency department visits. Primary cost outcomes were total medical costs allowed, pharmacy costs allowed, and total allowed costs of care over the 12 months post‐CR enrollment. The cost analysis was restricted to patients covered by the Geisinger Health Plan to allow collection of claims data. The allowed amount is the amount this plan and any other insurer paid plus any member payments (eg, copay and claim balance) on the claim. Statistical Analysis Descriptive statistics including mean±SD for continuous variables and frequency percentages for categorical variables are reported. Demographic and clinical characteristics of the sample were presented and stratified by program type. Baseline characteristics between programs were compared using Pearson's chi‐square and Wilcoxon rank‐sum tests. Because this is a nonrandomized study, we expected some baseline variables to vary between programs. To account for the imbalance (ie, confounding) we estimated the propensity score for being in VCR using logistic regression. The propensity score was then used in the analysis to weight observations to create a population balanced on all baseline variables. Balance was measured using standardized mean differences. Standardized mean difference <0.10 is considered balanced. See Table 1 for the list of covariates used in the propensity score model. Because the CBCR period spanned the prepandemic and pandemic time periods, we compared outcomes and number of completed sessions. The CBCR time periods were found to not be significantly different; therefore, we chose to combine into a single CBCR cohort. Table 1Demographics and Comorbidities of Study GroupsPreweightingPostweightingCBCR n=2303VCR n=703 P value|SMD|* CBCR n=2303VCR n=703|SMD|Age, y, mean±SD66.9±11.3570±10.83<0.00010.3067.4±27.9268.4±44.500.03Female sex, n (%)749 (32.4)269 (38.3)0.00310.1231.9%34.5%0.05Race, n (%)0.00160.090.13White2251 (98.2)673 (96.8)98.2%96.1%Other race† 41 (1.8)22 (3.2)1.8%3.9%Ethnicity, n (%)<0.00010.150.16Non‐Hispanic2278 (99.4)675 (97.5)99.5%97.4%Latino/Hispanic14 (0.6)17 (2.5)0.5%2.6%History of hypertension, n (%)1427 (61.7)546 (77.7)<0.00010.3563.6%66.8%0.07History of hyperlipidemia, n (%)1311 (56.7)490 (69.7)<0.00010.2756.2%58.8%0.05History of diabetes, n (%)664 (28.7)268 (38.1)<0.00010.2028.8%30.0%0.03History of myocardial infarction, n (%)1044 (45.1)288 (41.0)0.14410.0945.6%41.5%0.08History of coronary artery disease, n (%)1538 (66.5)602 (85.6)<0.00010.4572.0%76.6%0.11History of congestive heart failure, n (%)445 (19.2)204 (29.0)<0.00010.2322.2%24.0%0.04History of chronic kidney disease, n (%)751 (32.5)335 (47.7)<0.00010.3135.6%37.3%0.03Alcohol abuse, n (%)93 (4.0)40 (5.7)0.02950.084.3%4.4%0.00Anemia/nutritional disorder, n (%)392 (17.0)188 (26.7)<0.00010.2418.2%20.5%0.06Iron deficiency, n (%)60 (2.6)23 (3.3)0.55710.042.6%2.5%0.01Leukemia, n (%)18 (0.8)5 (0.7)0.91980.010.8%0.5%0.04Lymphoma, n (%)23 (1.0)15 (2.1)0.06670.091.0%1.4%0.03Metastatic cancer, n (%)18 (0.8)12 (1.7)0.10200.080.7%1.2%0.05Solid tumor in situ, n (%)56 (2.4)23 (3.3)0.55710.052.3%2.0%0.01Solid tumor, malignant, n (%)232 (10.1)109 (15.5)<0.00010.169.9%10.8%0.03Cerebrovascular disease, n (%)373 (16.2)127 (18.1)0.49400.0516.5%15.1%0.04Coagulopathy, n (%)217 (9.4)96 (13.7)0.00230.1310.6%10.3%0.01Dementia, n (%)61 (2.6)32 (4.6)0.03360.103.2%3.1%0.00Depression, n (%)469 (20.4)164 (23.3)0.16680.0720.2%19.6%0.01Diabetes w/ chronic comp, n (%)590 (25.6)251 (35.7)<0.00010.2226.2%27.6%0.03Diabetes w/o chronic comp, n (%)807 (35.0)302 (43.0)<0.00010.1635.0%35.9%0.02Drug abuse, n (%)41 (1.8)19 (2.7)0.00100.062.2%1.9%0.02Hypertension, uncomplicated, n (%)1756 (76.2)599 (85.2)<0.00010.2377.3%79.1%0.04Hypertension, complicated, n (%)807 (35.0)341 (48.5)<0.00010.2838.0%38.2%0.00Heart failure, n (%)693 (30.1)282 (40.1)<0.00010.2134.4%35.6%0.02Liver disease, mild, n (%)202 (8.8)96 (13.7)<0.00010.168.6%9.2%0.02Liver disease, moderate to severe, n (%)16 (0.7)7 (1.0)0.06230.030.7%0.5%0.02Chronic obstructive pulmonary disease, n (%)530 (23)224 (31.9)<0.00010.2024.9%27.7%0.06Neurologic/movement disease, n (%)77 (3.3)30 (4.3)0.11260.053.7%3.3%0.02Seizures/epilepsy, n (%)63 (2.7)23 (3.3)0.55650.042.9%2.2%0.04Paralysis, n (%)69 (3.0)19 (2.7)0.58190.023.0%1.9%0.07Obesity, n (%)972 (42.2)341 (48.5)<0.00010.1338.6%37.8%0.02Peripheral vascular disease, n (%)659 (28.6)297 (42.2)<0.00010.2932.2%33.0%0.02Pulmonary circulation disease, n (%)182 (7.9)59 (8.4)0.05050.029.2%6.6%0.09Renal failure, moderate, n (%)512 (22.2)233 (33.1)<0.00010.2424.0%25.1%0.03Renal failure, severe, n (%)79 (3.4)43 (6.1)<0.00010.134.5%4.8%0.01Hypothyroidism, n (%)410 (17.8)157 (22.3)0.00130.1117.8%17.9%0.00Other thyroid disorders, n (%)128 (5.6)59 (8.4)<0.00010.115.0%4.9%0.00Peptic ulcer with bleeding, n (%)62 (2.7)24 (3.4)0.55490.042.9%2.7%0.01Valvular heart disease, n (%)657 (28.5)269 (38.3)<0.00010.2134.2%34.6%0.01Weight loss, n (%)132 (5.7)53 (7.5)0.06230.076.4%6.2%0.01CBCR indicates center‐based cardiac rehabilitation; and VCR, virtual cardiac rehabilitation.*|SMD|=absolute standardized mean difference.†Black, Native American, Asian/Pacific Islander, and other. Demographics and Comorbidities of Study Groups CBCR indicates center‐based cardiac rehabilitation; and VCR, virtual cardiac rehabilitation. |SMD|=absolute standardized mean difference. Black, Native American, Asian/Pacific Islander, and other. After balancing the baseline covariates, the health outcomes were assessed using a weighted Poisson regression models with follow‐up time used as an offset to estimate the incident rate (#events/10 000 patient days of follow‐up). Weighted Gamma regression models that included an offset for the number of follow‐up months were used for the cost outcomes. Results are reported as annual incident rate and incident rate ratio with 95% CIs for the health outcomes and the ratio of mean per member per month costs expressed as VCR:CBCR and 95% CIs for costs. The comparison of costs between groups was accomplished by testing the ratio of the averages due to the type of CR program. Statistical analyses were performed using the twang package 21 in R v4.3.3 22 and SAS v9.4 (SAS Institute Inc., Cary, NC, USA). For all comparisons P<0.05 was considered significant. RESULTS There were 3006 patients who participated in CR with mean±SD age, 67.7 [11.3] years, 1017 [33.8%] women and 2924 [97.9%] were White. A total of 2303 patients were enrolled in CBCR (76.6%) and 703 in VCR (23.4%). In addition to being significantly older, the VCR group had significantly higher rates of hypertension, hyperlipidemia, diabetes, coronary disease, heart failure, valvular disease, peripheral vascular disease, kidney disease, and chronic pulmonary disease (Table 1). After weighting of the observations using the propensity score, all but 3 covariates were balanced (all | standardized mean differences | ≤0.10). The baseline covariates that remained above | standardized mean difference |>0.10 were race, ethnicity, and coronary disease (Table 1). Patients enrolled in VCR completed a significantly higher median number of CR sessions versus CBCR (34.4 versus 21.4, P<0.001, data not shown). Propensity‐weighted health outcomes of CBCR and VCR are presented as 1‐year incidence rate and incident rate ratio (IRR). VCR was associated with a statistically significant reduction in 1‐year all‐cause hospital readmission (14.9 versus 24.2 events, IRR, 0.616 [95% CI, 0.489–0.777]), P<0.001) and ED admission (26.8 versus 48.0 events, IRR, 0.557 [95% CI, 0.452–0.687], P<0.001). There were no significant differences in myocardial infarction and overall mortality between CBCR and VCR consistent with prior studies 17 , 23 (Table 2). To confirm that the impact of the VCR program on these outcomes reflected a reduction in CR‐related diagnoses, we analyzed the percentage of ED visits and readmissions that were for CR‐related International Classification of Diseases, Tenth Revision (ICD‐10) codes. There were significantly fewer CR‐related ED visits (12.7% versus 59.3%, P<0.001) and admissions (48.7% versus 84.5%, P<0.001) at 1 year for VCR versus CBCR, suggesting the VCR program's impact was related to these diagnoses rather than a general effect related to more frequent program “touches.” (Table 3). Table 2Weighted Analysis of Health Outcomes at 1 Year Post EnrollmentCBCR (n=2303)VCR (n=703)Incident rate ratio P valueInpatient readmission at 1 y (n)4441180.616 (0.489–0.777)<0.001IRR (25th–75th%tile)24.2 (21.9–26.7)14.9 (12.1–18.4)Emergency department admission at 1 y (n)8061500.557 (0.452–0.687)<0.001IRR (25th–75th %tile)48.0 (44.5–51.8)26.8 (22.0–32.5)Myocardial infarction at 1 y (n)30101.250 (0.535–2.914)0.605IRR (25th–75th %tile)1.4 (0.9–2.1)1.8 (0.8–3.7)Mortality at 1 y (n)57160.618 (0.345–1.104)0.104IRR (25th–75th %tile)3.3 (2.5–4.3)2.0 (1.2–3.4)CBCR indicates center‐based cardiac rehabilitation; IRR, incident rate ratio; and VCR, virtual cardiac rehabilitation. Weighted Analysis of Health Outcomes at 1 Year Post Enrollment CBCR indicates center‐based cardiac rehabilitation; IRR, incident rate ratio; and VCR, virtual cardiac rehabilitation. Table 3Emergency Department Visits and Admissions at 1 Year for ICD‐10 Codes Reflecting CR DiagnosesCBCR (n=806)VCR (n=150) P valueEmergency department visits with CR‐related ICD‐10 code (%)478 (59.3)19 (12.7)<0.001Admissions for CR‐related ICD‐10 code (%)375 (84.5)58 (48.7)<0.001CBCR indicates center‐based cardiac rehabilitation; CR, cardiac rehabilitation; ICD‐10, International Classification of Diseases, Tenth Revision; and VCR, virtual cardiac rehabilitation. Emergency Department Visits and Admissions at 1 Year for ICD‐10 Codes Reflecting CR Diagnoses CBCR indicates center‐based cardiac rehabilitation; CR, cardiac rehabilitation; ICD‐10, International Classification of Diseases, Tenth Revision; and VCR, virtual cardiac rehabilitation. Adherence to CR was measured as mean encounters completed by participant cohorts. Patients in VCR completed a mean of 34.4 CR encounters, whereas those in CBCR had a mean of 21.4 CR encounters (P<0.001) (data not shown). There was a total of 220 VCR and 375 CBCR patients that had Geisinger Health Plan insurance when they were enrolled in CR. Table 4 reports the weighted costs comparisons. Total medical allowed and total costs allowed were found to be significantly reduced in the VCR group. More specifically, total medical allowed in the VCR group was observed to have an 18.6% reduction in the average per member per month (95% CI, 0.690–0.960, P=0.0144). Similarly, the total costs allowed resulted in 16.2% lower average per member per month (95% CI, 0.725–0.970, P=0.0176). Table 4Weighted Analysis of Cost Outcomes at 1 Year Post Enrollment (VCR/CBCR, Ratio of Monthly Cost, Virtual Cardiac Rehabilitation/Center Based Cardiac Rehabilitation)Cost ratio VCR/CBCR P valueTotal medical costs at 1 y (25th–75th %tile)0.814 (0.690, 0.960)0.0144Total prescription costs at 1 y (25th–75th %tile)0.928 (0.753, 1.143)0.4834Total cost of care at 1 y (25th–75th %tile)0.838 (0.725, 0.970)0.0176CBCR indicates center based cardiac rehabilitation; VCR, virtual cardiac rehabilitation. Weighted Analysis of Cost Outcomes at 1 Year Post Enrollment (VCR/CBCR, Ratio of Monthly Cost, Virtual Cardiac Rehabilitation/Center Based Cardiac Rehabilitation) CBCR indicates center based cardiac rehabilitation; VCR, virtual cardiac rehabilitation. DISCUSSION In the present study, we retrospectively compared outcomes of a novel VCR program and a traditional CBCR program using a propensity matching strategy to control for bias introduced by a retrospective observational study design. Our results suggested that compared with CBCR, VCR was associated with a significantly lower rate of all‐cause rehospitalization and emergency department use at 1 year following initiation of CR sessions. These improved outcomes were associated with lower overall total cost of care for VCR participants. Our study adds to the existing body of data supporting the concept that VCR is a cost‐effective strategy to deliver CR with patient outcomes at least comparable to traditional CBCR. Prior work has examined the efficacy and cost‐effectiveness of VCR versus CBCR. Nkonde‐Price et al. demonstrated a significant reduction in 1‐year all‐cause hospitalization in a retrospective cohort study of VCR versus CBCR in the Kaiser Permanente system. 17 Oehler et al. noted a nonsignificant trend toward cost savings associated with VCR compared with CBCR; however, hospitalization and emergency department usage were not different between CR modalities. 23 To the best of our knowledge, our study is the first to demonstrate a statistically significant improvement in outcomes associated with VCR versus CBCR coupled with a favorable trend in overall health care costs among patients participating in VCR. Placing the present study in context with available studies assessing the efficacy of VCR versus CBCR is challenging due to the heterogeneous nature of VCR programs. Meta‐analyses comparing CBCR and remote CR programs are difficult to generalize as they combine a multitude of delivery models and include scant data from the United States, where different barriers are likely to exist with regard to access to digital technologies and participation in exercise sessions. 24 Existing data on the comparative efficacy of VCR, which delivers a core CR experience closest to CBCR, are limited to low‐ and moderate‐risk patients, with little coverage of patients with multiple comorbidities, who are >75 years of age, women, or who belong to minority racial and socioeconomic groups. 25 Differing from the existing data and a novel aspect of this study, was the higher medical complexity and number of comorbidities in the patients enrolled in VCR as compared with CBCR. As our study used a unique VCR program developed and delivered by a third‐party entity, no assumption can be made that similar improvements in outcomes would be seen using alternative and less comprehensive programs. However, it is important to note that our study adds to the data that a VCR program using smart devices and synchronous video visits is feasible in an older patient population that tends to have limited technology literacy. This was in agreement with Harzand et al. who showed the feasibility of a smartphone approach to CR in an older veteran population. 26 In tandem with clinical efficacy, the cost‐effectiveness of VCR models is another important factor in their feasibility. A recent meta‐analysis of 9 different VCR programs found that 7 were likely cost effective, though again these findings are difficult to generalize due to the heterogeneity in CR delivery models and country‐level differences between programs. 27 A recent study found that asynchronous CR may be more cost effective to deliver than traditional CBCR, when driving comparable health outcomes. 22 However, asynchronous CR may face limited uptake given its lack of supervision and clinician feedback during sessions. The present study adds to the available evidence that VCR can be associated with favorable overall patient cost while delivering an efficacious CR program. The driver of improved outcomes of VCR in our study is unknown but several possibilities can be hypothesized. First, the intensity and comprehensive nature of a synchronous VCR program may play a role. The Recora program allows a focus on lifestyle modifications beyond the limited interactions during CR sessions. The application allows patients to log vitals, meals, weight, daily steps, physical activity, and mood. Tracking these parameters on a frequent basis rather than only at CBCR sessions could have led to improved outcomes through better patient awareness. Specifically, the availability of home blood pressure and wearable physical activity monitors may have been beneficial as shown in a meta‐analysis by Hannan et al. that concluded wearable physical activity monitors were superior to no device in improving cardiorespiratory fitness in the maintenance phase of CR. 28 Although these data may be logged by patients more frequently, any review of the data would occur on VCR sessions. It is possible that the higher number of potential interaction points in a VCR program would lead to higher patient engagement, and thus improved outcomes. In general, Recora would discuss logs only if patients brought this as a discussion topic at the next session. The logs are a feature for patients to track their own blood pressure and were not part of a remote monitoring program. Additionally, it is possible that this higher patient engagement may have directed patients more readily to their primary care physician or cardiologist, avoiding a delay in care that would have resulted in an emergency department visit or hospitalization. A second possibility is that the benefit may be related to the dose of CR afforded by VCR versus CBCR. Patients using VCR in our study completed 60% more sessions than those participating in CBCR. Prior studies support the concept that CR, like many interventions in cardiology, demonstrates a dose–response relationship, lending evidence to support this theory. 29 , 30 As the Geisinger Health System serves patients in a very large catchment area, it is possible that the dose of CR afforded by VCR was much higher than CBCR due to the ease of access and reduced barriers to attendance. Finally, the possibility of an undefined confounder related to the nonrandomized nature of the study cannot be excluded. Virtual and telehealth services have proliferated in recent years, and these modalities have improved access across a wide spectrum of clinical services and specialties. However, for continued broad adoption, telehealth will need to demonstrate both cost‐effectiveness and clinical efficacy. Of note, the effects of telehealth on quality and Medicare spending have been recently studied, revealing a slight increase in spending and small improvements in quality. 31 The present study supports VCR as a clinical service that may be particularly well suited to telehealth delivery, given gains in access, clinical outcomes, and cost of care reductions. It is hoped that the broad adoption of VCR will yield an improvement in access for eligible patients, particularly those of traditionally underserved demographics. The Geisinger Health System serves counties that are range from 75% to 95% White. 32 Regardless, it was noted that compared with CBCR, VCR enrolled a small but significantly greater percentage of non‐White participants. Whether VCR can be leveraged to reach this and other underserved populations should be the subject of further study. The study has several limitations. The ideal method of comparison of outcomes of VCR versus CBCR would be a randomized trial design. The present study was a retrospective, nonrandomized observational study and as such bias cannot be completely eliminated. It is possible that the discretion left to the ordering provider to select a mode of rehabilitation favored healthier patients enrolling in VCR rather than CBCR. Our robust efforts to risk match groups likely minimized but cannot eliminate this issue. It is notable that the outcomes of CBCR were very similar when comparing the study period before the introduction of VCR and during the period when patients were offered both VCR and CBCR. This consistency adds some degree of strength to the retrospective cohort design and results regarding the favorable outcomes of VCR versus CBCR. Additionally, this study was conducted at a large integrated health care system that may not make this study generalizable to other patient populations. However, the benefit of VCR may actually be blunted in this study given the several CBCR options available. Conclusions In conclusion, the present study found that VCR was associated with reduced rates of hospitalization and ED usage with reduced overall costs as compared with traditional CBCR. These findings lend further support for VCR as an alternative to CBCR especially in areas of limited access to traditional center‐based programs. Sources of Funding The study was supported by an unrestricted research grant from Recora Inc. and the Geisinger Research Institute. Disclosures Edward Wu is employed by Recora Inc, New York NY. The remaining authors have no disclosures to report. This article was sent to Tazeen H. Jafar, MD MPH, Associate Editor, for review by expert referees, editorial decision, and final disposition. Preprint posted on MedRxiv May 07, 2024. doi: https://doi.org/10.1101/2024.05.05.24306717. For Sources of Funding and Disclosures, see page 8. References