The Cardiac Rehabilitation Program (CRP) at KPNC is a home-based virtual telehealth program, in place since 1990. Each of 21 medical centers in Northern California has an independent local CRP, coordinated by a central Regional Director as a part of population care management. Patients experiencing a qualifying cardiac event (MI, coronary artery bypass graft [CABG], or percutaneous coronary intervention [PCI]) become eligible for the program and are accordingly referred to the CRP before hospital discharge. Within 2 weeks after the cardiac event, the nurse care manager and Program Director assess patients and initiate enrollment as appropriate.

A program care manager, typically a nurse, contacts the patient to do an intake visit, including the patient’s medical history, medications, labs, dietary and exercise habits, and then follows up with the patient monthly to deliver the CRP over 6 months. During scheduled telephone-based encounters, the care manager will assess beta- blocker adherence, dual antiplatelet therapy, statin adherence, BP and diabetes control, low-density lipoprotein (LDL) goal to <70, and psychological health. They guide the patients on an appropriate diet and exercise regimen. Patients are given personalized exercise prescriptions based on the Rate of Perceived Exertion (RPE) scale . Some local programs also perform treadmill testing for functional capacity and exercise recommendations. Care managers also counsel patients about smoking cessation, alcohol, and substance use. All study patients, regardless of CR participation, continue to be closely followed and clinically managed by primary care and cardiologist physicians and clinical department standard outreach.

In this multicenter retrospective cohort study, all CR-eligible adults (18 to 85 years) experiencing a qualifying cardiac event as defined above, from January 01, 2016, to December 31, 2022, at any of the 21 KPNC medical centers were included in the study population. Patients with gaps in insurance coverage ≥60 days, diagnosed as frail, experiencing previous CR, hospice, or palliative care within 12 months preceding the qualifying event, or missing key data, were excluded from further study (Supplementary Figure 1). Study data were extracted from the KPNC Electronic Health Record and from program administrative data.

To identify a homogenous cohort and to reflect program criteria designed to align with CR quality measures established by the National Committee for Quality Assurance, the study population included only those who suffered MI, or underwent PCI or CABG. Notably, other indications for CR, such as heart and heart/lung transplantation, heart valve repair, or replacement, were not included in the study.

Of note, all patients included in the study are followed at Kaiser and managed by standard care practices from primary care teams and cardiologists.

We defined exposure to CR as three or more CR encounters in the 6-month exposure period after the initial cardiac event. We did not include the initial recruitment and intake visit in this count of intervention exposure.

The study outcomes included all-cause mortality, inpatient hospitalization or emergency department (ED) visit, revascularization procedures (PCI or CABG), and major adverse cardiovascular event (MACE), defined as MI, cardiogenic shock, or cardiac arrest. All outcomes were assessed over a 12-month duration, which started after the initial 6-month CR intervention exposure period.

To account for imbalanced covariates with standardized mean differences (SMDs) >0.2, our study implemented inverse probability of treatment weighting (IPTW). Weights were created using a propensity score model consisting of the following variables: sex (male or female), race (White, Asian, Hispanic, Black, and Other), age categories (25 to 64, 65 to 74, 75 to 84, 85+ years of age), LDL (above or below 70), diagnosis of depression, active portal user status, qualifying cardiac event type, smoking status, comorbidity point score (COPS) risk score quartiles, degree of hypertension (stage 2, stage 1, elevated, normal, missing), neighborhood deprivation index, year of index event, and medical center. The propensity score from this model was converted to a weight using the average treatment effect method to estimate the effect of CR treatment on the population of eligible patients. To determine if IPTW was successful, we compared the SMDs before and after weighing with a threshold of 0.2.

Multivariable modified Poisson was used to estimate risk ratios for binary outcomes: hospitalization, ED visits, MACE, and revascularization. To investigate the associations between CR exposure and mortality, both modified Poisson (risk ratios) and Cox proportional hazards (hazard ratios) modeling were used.

In addition to covariate adjustment through IPTW, we also added covariates to the model to address residual confounding. The covariates used for adjustment in multivariable models were the same variables used in the propensity score model: sex (M/F), race (White, Asian, Hispanic, Black, Other), age categories (25 to 64, 65 to 74, 75 to 84, 85+), LDL (above or below 70), history of depression (yes/no), active patient portal user (yes/no), qualifying event type (inpatient MI, inpatient CABG, inpatient PCI, or outpatient PCI, smoking status (current, former, never, missing), COPS comorbidity risk score (quartiles), blood pressure group (hypertension stage 2, hypertension stage 1, elevated, normal, missing), neighborhood deprivation index, year of index event (2016 to 2022), and medical center (21 facilities).

We conducted several sensitivity analyses. We examined alternate definitions of cardiac rehab exposure based on one or more completed visits and based on six or more completed visits. We further explored a subgroup analysis excluding study patients with extended hospitalizations ≥30 days, entered palliative/hospice care, had repeat MACE events, or died during the exposure period.

All analyses were performed using SAS 9.4 (Cary, North Carolina). The study was approved by the Institutional Review Board of the Kaiser Foundation Research Institute. The study protocol follows the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines for cohort studies.