In the Netherlands, it is mandatory to have health insurance. For the present study, health insurance claims data were used from Dutch patients with CVD with a multidisciplinary outpatient CR program indication and who were insured at Coöperatie Volksgezondheidszorg (VGZ), one of the largest health insurance companies in the Netherlands. The VGZ cohort consists of 4.1 million beneficiaries, representing 24% of the Dutch population,¹⁵ with nationwide coverage and a representative sample of individuals from all age categories, SES, and both rural and urban areas (eFigure 1 in the Supplement). Individuals eligible for the present study were patients with a new diagnosis of CVD with an indication for CR according to Dutch guidelines,¹⁶ including an acute coronary event (myocardial infarction [MI] or unstable angina pectoris [AP]), stable AP, chronic heart failure, or cardiothoracic surgery (coronary artery bypass grafting [CABG], valve replacement, or percutaneous coronary intervention [PCI]). Reimbursement for outpatient CR is provided on the condition that the patient is referred by a cardiologist. Health insurance claims data were used to assess the type and date of diagnosis, and patients were enrolled between July 1, 2012, and December 31, 2017, with follow-up to March 19, 2020. The dates of analysis were March to May 2020. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. The Radboud University Medical Center Institutional Review Board deemed that this observational cohort study was exempt from informed consent because it involved retrospective analysis of an anonymized data set.

The Dutch Society of Cardiology released a multidisciplinary CR guideline,¹⁶ which is consistent with international recommendations.^17,18 Group-based outpatient CR programs consist of the following 4 modules: (1) supervised exercise training, (2) mental health and stress relief, (3) social health, and (4) cardiovascular risk management. The multidisciplinary CR program typically lasts 6 to 12 weeks; on average, 85% of the patients receive exercise training, 39% receive relaxation therapy, 17% receive lifestyle modification therapy, and 75% receive education.¹⁹ Hence, the content of the contemporary Dutch CR program is largely comparable to that of CR programs in other countries. Accessibility to CR in the Netherlands is excellent because most CR facilities are located near the residence of patients (<30 km).

A total of 101 940 patients met the initial selection criteria. Consistent with previous observational studies,^20,21 patients were classified as having received CR if a claim was filed for at least one of the group-based outpatient treatments. Excluded from further analysis were patients who (1) were insured less than 365 days before or less than 180 days after their diagnosis of CVD, (2) began CR more than 90 days after their diagnosis, or (3) died within 180 days after their diagnosis. Patients with multiple CVD-related insurance claims within 30 days after the initial diagnosis were reclassified according to the most severe diagnosis using the following rankings for CVD diagnosis and cardiothoracic surgery.²⁰ For CVD diagnosis, the ranking was (1) ST-segment elevation MI (STEMI), (2) non-STEMI (NSTEMI), (3) unstable AP,(4) stable AP, and (5) chronic heart failure. For cardiothoracic surgery, the ranking was (1) CABG, (2) valve replacement, (3) acute PCI, and (4) elective PCI. Accordingly, 83 687 eligible patients with CVD were available for analysis (eFigure 2 in the Supplement).

Survival status of study participants was obtained from the Dutch Population Register, and the date of death was extracted if applicable. All-cause mortality was the primary outcome, and survival time in days was calculated from the date of the initial diagnosis of CVD. The maximum follow-up in the cohort was 7.7 years (92 months). Patients were censored if they switched to another health insurance company or if they were still alive during the last survival check.

Sex and age were extracted from the VGZ health insurance database. Income and SES were obtained from the national statistical office (Statistics Netherlands) using postal code information for the patient’s residence. Socioeconomic status was classified using the SES score from Statistics Netherlands as low (bottom 40%), moderate (middle 30%), or high (upper 30%) (eFigure 1C in the Supplement).

Comorbidity was identified using diagnosis codes from the Dutch Business Intelligence Center for Healthcare (Vektis). Insurance claims data for pharmaceutical agents were used to evaluate the association of diabetes, cancer, gout, Parkinson disease, respiratory diseases, thyroid diseases, and dementia as binary variables with CR outcomes.²² We evaluated the consequences of multimorbidity using the Charlson Comorbidity Index (CCI). As a continuous variable, the CCI was calculated as the sum of comorbid conditions²³ as follows. One point was given for MI, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, ulcer disease, mild liver disease, and diabetes. Two points were given for diabetes with end-organ damage, any tumor, leukemia, and lymphoma. Three points were given for moderate to severe liver disease. Six points were given for metastatic solid tumors and AIDS. We compared the association of CR with all-cause mortality across patients with CCI scores of 1 or less, 2, 3, 4, and 5 or greater.

Baseline characteristics were summarized as mean (SD) or median (interquartile range) for continuous variables and as number (percentage) for categorical variables. Differences in individual and disease characteristics between CR participants and nonparticipants were assessed using t test for continuous variables and χ² test for categorical variables. Kaplan-Meier curves and a log-rank test were used to assess the difference in all-cause mortality between CR participants and nonparticipants.

Information about patient characteristics, income, SES, distance to CR facility, disease characteristics, comorbidity, pharmaceutical agents in the year of diagnosis of CVD, and health care expenses was used to calculate the inverse propensity score. The crude hazard ratio (HR) (with 95% CI) was calculated by univariable Cox proportional hazards models. Subsequently, we corrected for sex, age, category of CVD diagnosis, cardiothoracic surgery, Charlson Comorbidity Index, and cardiac medication. Stabilized inverse propensity score weighting was applied to the fully adjusted multivariable Cox proportional hazards models to account for differences between CR participants and nonparticipants.^24,25,26 We evaluated the robustness of the main outcomes via sensitivity analyses. For this purpose, the period between the initial event date and the minimum follow-up period was changed from 180 days to 1 to 6 years.

The presence of effect modification was tested for using an interaction term to assess the implications of sex (reference group, men), age (reference group, <50 years), SES (reference group, low SES), CVD diagnosis (reference group, chronic heart failure), cardiothoracic surgery (reference group, elective PCI), and comorbidity (reference group, CCI≤1) for the association between CR participation and all-cause mortality risk. Two-sided P < .05 was considered statistically significant. All statistical analyses were performed in SAS, version 9.4 (SAS Institute Inc).

In this cohort of 83 687 patients with CVD (mean [SD] age, 67 [12] years; 60.4% [n = 50 512] men), only 31.3% (n = 26 171) participated in CR at 89 CR facilities in the Netherlands. The CR participation rate increased across calendar years from 25.7% in 2012 to 38.6% in 2017 (eTable 1 in the Supplement). Distance to the CR facility was greater for CR participants than for nonparticipants (mean [SD], 15.4 [19.3] vs 14.7 [20.2] km; P < .001). Compared with nonparticipants, CR participants were younger, included more male patients, had a lower CCI score, and had a higher income (Table 1). Cardiac medication use was lower in the year before CVD diagnosis among CR participants compared with nonparticipants, but this association largely reversed in the year after CVD diagnosis. Cardiac rehabilitation participation rates were higher among men (37.2%), patients younger than 75 years (33.5% to 39.8%), those with a diagnosis of STEMI (66.6%) or NSTEMI (49.2%), those with a CCI score of 3 or less (32.0% to 41.7%), and those undergoing CABG (73.0%), valve replacement (54.8%), or acute (66.0%) or elective (41.8%) PCI. In contrast, CR was underused in women (22.2%), patients 75 years or older (6.7% to 21.5%), those with comorbidity (12.7% to 26.5%), and those having a CVD diagnosis of unstable (25.9%) or stable (15.9%) AP or chronic heart failure (5.1%) (Figure 1).

During a mean (SD) of 4.7 (1.8) years of follow-up (56 [22] months), 1966 CR participants (7.5%) and 13 443 nonparticipants (23.4%) died (P < .001) (Table 2). Kaplan-Meier analysis revealed that CR participation was associated with better event-free survival (Figure 2). Cox proportional hazards models showed a 68% lower risk of all-cause mortality (crude HR, 0.32; 95% CI, 0.30-0.33) in CR participants compared with nonparticipants (Table 2). Mortality risk remained lower in CR participants after stepwise correction for confounders (adjusted HR, 0.68; 95% CI, 0.65-0.71) (eTable 2 in the Supplement). After multivariable adjustment, CR participation was associated with a 32% lower risk of all-cause mortality (adjusted HR, 0.68; 95% CI, 0.65-0.71) compared with nonparticipation. Mortality rates during follow-up were variable across subgroups (Table 2), ranging from 1.3% to 63.9%. Cardiac rehabilitation participation was associated with improved event-free survival and a statistically significantly lower risk of all-cause mortality among all subgroups, except for patients with dementia (Figure 3). Sex, age, SES, and comorbidity did not alter risk reduction after CR participation, with no effect modification observed. However, a statistically significant interaction association was found across categories of CVD diagnosis and cardiothoracic surgery. Larger reductions in risk estimates for all-cause mortality were found after CR participation for STEMI (adjusted HR, 0.59; 95% CI, 0.52-0.68 vs 0.72; 95% CI, 0.65-0.79; P < .001), NSTEMI (adjusted HR, 0.64; 95% CI, 0.58-0.70 vs 0.72; 95% CI, 0.65-0.79; P < .001), and stable AP (adjusted HR, 0.69; 95% CI, 0.63-0.76 vs 0.72; 95% CI, 0.65-0.79; P < .001) compared with patients with chronic heart failure, whereas unstable AP had a smaller risk reduction (adjusted HR, 0.75; 95% CI, 0.67-0.85 vs 0.72; 95% CI, 0.65-0.79; P < .001). Effect modification was also present for cardiothoracic surgery, with acute PCI having a larger reduction in the risk estimate for all-cause mortality after CR participation (adjusted HR, 0.64; 95% CI, 0.57-0.72 vs 0.82; 95% CI, 0.74-0.91; P < .001) compared with elective PCI, whereas no difference in risk reduction was found for CABG (adjusted HR, 0.75; 95% CI, 0.63-0.90 vs 0.82; 95% CI, 0.74-0.91; P = .53) or valve replacement (adjusted HR, 0.76; 95% CI, 0.62-0.93 vs 0.82; 95% CI, 0.74-0.91; P = .53).

To evaluate the robustness of the main outcomes, the adjusted HRs of CR participation on all-cause mortality were recalculated. For this purpose, mortality risk reduction for different durations during follow-up was calculated. These sensitivity analyses showed risk reduction for all-cause mortality similar to the initial findings (adjusted HR, 0.68-0.86) (eTable 3 in the Supplement).

The crude risk reduction in all-cause mortality after CR participation was 68%, which was reduced to 32% after adjustment for confounding factors. The quantification of mortality risk reduction in the present study is substantially larger compared with outcomes of previously published meta-analyses (−1% to 13%).^2,5,27,28 A potential explanation for the discrepant outcomes may relate to the study design. We used an observational population-based cohort to study the health benefit of CR in a real-world setting, yielding high external validity. This design is in contrast to meta-analyses, which use data from randomized clinical trials (RCTs). Although RCTs may reduce the risk of confounding because of randomization, vulnerable patients (eg, older individuals or those with comorbidity or multimorbidity) are typically excluded. Hence, RCTs have high internal validity but typically have lower external validity.²⁹ The health benefit of CR may be larger in patients who are underrepresented in or excluded from RCTs. Alternatively, the content of the CR program may alter the outcomes.¹⁷ Our patients were enrolled in a multidisciplinary program, including supervised exercise training, mental health and stress relief, social health, and cardiovascular risk management,¹⁶ whereas most RCTs compared exercise-based CR with a control condition.⁵ Hence, mortality risk reduction may have been larger in our cohort because of the multifaceted treatment that patients received.³⁰ Indeed, our findings align with other large observational population studies^20,31,32 reporting a 33% to 35% risk reduction in CR participants. Independent of the risk reduction of the present study vs data from RCTs, these joint outcomes further reinforce the Class IA recommendation for CR participation in current and future clinical guidelines of professional associations.^7,8,9,10

A unique aspect of the present study is the assessment of mortality risk across different subgroups of patients eligible for CR. Better survival was found for all subgroups participating in CR. The magnitude of risk reduction after CR differed statistically significantly across categories of CVD diagnosis and type of cardiothoracic surgery, but risk estimates and their 95% CIs were less than 1 for every subgroup (Figure 3), indicating that CR participation provides a better outcome than nonparticipation. This finding is notable because the level of evidence of CR health benefits is weak for some subgroups. Few studies on CR report sex-specific outcomes,³³ but risk reduction after CR differs between men and women.²⁰ Although CR use is lower in women than in men, we observed a similar mortality risk reduction with CR participation in men (32.1%) and women (31.3%) (Figure 3), suggesting that both sexes benefit equally from CR.

Older patients are often underrepresented in CR programs, despite their higher disease prevalence.³⁴ Hence, recommendations for treatment of older patients (≥75 years) are needed that are typical of those individuals encountered in routine clinical practice.³⁵ Age did not alter the association between CR participation and better mortality outcomes. In fact, CR was associated with large reductions in all-cause mortality in patients aged 75 to 84 years (27.8%) and 85 years or older (32.0%) (Figure 3). These findings suggest that CR improves survival even in older age groups, which have mortality rates of up to 63.9%.

The implications of CR for mortality in patients with stable AP remain unclear. In a recent systematic review and meta-analysis²⁸ (including 7 studies and 581 patients), risk estimates could not be calculated because of insufficient power. Among 24 549 patients with stable AP in our study, CR participation was associated with a 31% mortality risk reduction (adjusted HR, 0.69; 95% CI, 0.63-0.77) (Figure 3). Similarly, among 13 813 patients with unstable AP, CR participation was associated with a 25% mortality risk reduction (adjusted HR, 0.75; 95% CI, 0.67-0.85). Our observations underscore the mortality benefit of CR for patients with stable AP and patients with unstable AP.

Multiple comorbid conditions are increasingly prevalent among patients with CVD.³⁶ Patients with multimorbidity have a worse prognosis³⁷ and a lower CR participation rate,³⁸ and current clinical practice is mainly targeted toward care of patients with a single disease. We found that CR participants with diabetes, cancer, gout, Parkinson disease, respiratory diseases, and thyroid diseases had better survival rates compared with nonparticipants. Patients with CVD with dementia did not demonstrate a statistically significant risk reduction after CR, but the risk estimate indicated a 10.8% reduction in all-cause mortality (adjusted HR, 0.89; 95% CI, 0.55-1.46) (Figure 3). The lack of statistical significance is likely because of the small sample size (n = 393), which contributed to a wide 95% CI (−45% to 46%). Cardiac rehabilitation participation among patients with CVD with the highest CCI scores was associated with notable risk reduction (≥30%). These findings emphasize that CR has substantial health value for patients with CVD with comorbidity and should lead clinicians to encourage their patients to participate in CR.

Findings from the present study not only reinforce the health benefit of CR but also demonstrate the underuse of this powerful intervention. Only 31.3% of eligible Dutch patients participated in CR, which is high compared with US data^11,12 but is similar to values reported in European studies.^39,40 However, more notable is the increase in CR use from 25.7% in 2012 to 38.6% in 2017. Future studies are warranted to elucidate the reasons for this substantial improvement as well as to understand temporal changes in predictors of participation. Outcomes from such analyses can be used to develop strategies to further increase CR use across the globe.

Large variability in CR participation rates across different subgroups was also observed. Participation rates were high for medical conditions with clear guidelines and strong recommendations, such as STEMI⁹ (66.6%), CABG⁷ (73.0%), valve replacement⁷ (54.8%), and acute (66.0%) or elective (41.8%) PCI.⁴¹ Participation rates were low in subgroups with chronic conditions, such as older patients and those with stable AP, chronic heart failure, comorbidities (ie, diabetes, cancer, gout, etc), and multimorbidity (CCI≥4) (Figure 1). The lower participation rates likely relate to the frailty of these patients, a lack of physician referral, or to practical barriers, such as a lack of transportation.⁴² Alternative strategies to deliver CR, such as home-based CR,^43,44 telerehabilitation,⁴⁵ or eHealth interventions,⁴⁶ may enable patients with CVD to profit from CR.

This study has some limitations. The main limitation is its observational design. Analyses based on CR components were not possible to perform because of a change in the medical claims registry system during the study period and an insufficient level of detail regarding the number of sessions and the combination of components. The use of health insurance claims data did not allow us to control for all potential confounding factors, such as cardiovascular risk factors, disease severity, lifestyle (eg, physical activity, smoking, and diet), and patient motivation. Nevertheless, we were able to extract data regarding patient characteristics, CVD diagnosis, type of cardiothoracic surgery, hospital stay, pharmaceutical agent use, comorbidity, and health care costs to correct for differences between CR participants and nonparticipants. Stabilized inverse propensity score weighting was used to create a pseudopopulation and subsequently adjust for the covariates in the Cox proportional hazards models to minimize potential group differences. Matched-pairs analysis (n = 32 780) confirmed the findings of our primary analysis (data not shown). Race was not available, but most of the Dutch population is of white race/ethnicity (approximately 87.7%).⁴⁷ Residual confounding may have occurred because unmeasured variables could explain some of the observed mortality differences between CR participants and nonparticipants. However, results from the present study demonstrated a consistent mortality risk reduction among all subgroups of CR participants, and the health benefit of CR was further reinforced in the sensitivity analyses. Finally, the inclusion of a large and heterogeneous patient population in combination with the high accuracy of health insurance data⁴⁸ supports the value of the present study beyond previously published RCTs and meta-analyses.