A Detailed Analysis of Cardiac Rehabilitation on 180-Day All-Cause Hospital Readmission and Mortality

This study investigated the efficacy of cardiac rehabilitation (CR) after controlling for potential confounders in coronary artery disease. After controlling for demographics, controllable risk factors/comorbidities, and cardiovascular discharge criteria, 180-d all-cause readmission or death was decreased in patients who attended CR compared with those who did not.

Abstract

Purpose:

Cardiac rehabilitation (CR) is endorsed for coronary artery disease (CAD), but studies report inconsistent findings regarding efficacy. The objective of this study was to determine whether confounding factors, potentially contributing to these heterogeneous findings, impact the effect of CR on all-cause readmission and mortality.

Methods:

Patients (n = 2641) with CAD, CR eligible, and physically able were identified. Electronic medical records were inspected individually for each patient to extract demographic, clinical characteristic, readmission, and mortality information. Patients (n = 214) attended ≥1 CR session (CR group). Survival was considered free from: all-cause readmission; or composite outcome of all-cause readmission or death. Cox proportional hazards models, adjusting for demographics, comorbidities, and discharge criteria, were used to determine HR with 95% CI and to compare 180-d survival rates between the CR and no-CR groups.

Results:

During 180 d of follow-up, 12.1% and 18.7% of the CR and non-CR patients were readmitted to the hospital. There was one death (0.5%) in the CR group, while 98 deaths (4.0%) occurred in the non-CR group. After adjustment for age, sex, race, depression, anxiety, dyslipidemia, hypertension, obesity, smoking, type 2 diabetes, and discharge criteria, the final model revealed a significant 42.7% reduction in readmission or mortality risk for patients who attended CR (HR = 0.57: 95% CI, 0.33-0.98; P = .043).

Conclusions:

Regardless of demographic characteristics, comorbidities, and cardiovascular discharge criteria, the risk of 180-d all-cause readmission or death was markedly decreased in patients who attended CR compared with those who did not.

KEY PERSPECTIVES

What is novel?

• Many studies have accounted for demographics and clinical characteristics in their modeling of clinical outcomes in patients with heart disease, but have failed to account for the importance of cardiac rehabilitation (CR) participation.

• We only included patients with minimal barriers to participation in on-site CR and addressed several confounding factors, including demographics, discharge criteria, underlying risk factors, and medication.

What are the clinical/and/or research implications?

• When controlling for demographic characteristics, controllable risk factors/comorbidities, and cardiovascular referral diagnosis, CR participation resulted in a significant 43% risk reduction for readmission or death compared with those who did not participate in CR.

• Future studies evaluating the efficacy of CR should continue to identify and control for as many key individual-level variables as possible to adequately assess the true magnitude of beneficial CR effects on important clinical outcomes.

Affecting 6.7% (15 million) of the US adult population, coronary artery disease (CAD) is a major health burden.1 The incidence of acute myocardial infarction (AMI) in the United States is 805 000/yr; 954 000 undergo percutaneous coronary interventions (PCIs); and 397 000 have coronary artery bypass graft surgery (CABG).1 Advances in modern-era (post-2000) revascularization strategies and pharmacologic treatments have decreased mortality rates for CAD; however, despite these efforts, heart disease remains the leading cause of death in both men and women in the United States and worldwide.1–3 Furthermore, the high incidence of delayed mortality from CAD has led to a growing number of survivors in need of risk stratification and effective secondary prevention strategies.

Many studies have accounted for multiple demographics and clinical variables in their modeling of readmission or death risk in patients with heart disease, but have failed to account for the importance of cardiac rehabilitation (CR) participation.4–11 Cardiac rehabilitation is the cornerstone of CAD secondary prevention; a class 1 indication for CAD; and an internationally endorsed strategy12–14 supported by numerous cohort studies and meta-analyses for readmission15–19 and mortality.19–25 Meta-analyses and individual cohort studies investigating the effects of CR on readmissions and death generally demonstrate a beneficial effect of CR; however, studies suffer from study heterogeneity and poor data quality, thus limiting their clinical inferences and generalizability.26 Many studies of CR are focused on a specific CAD subpopulation (discharge criteria for CR), such as PCI,26–28 AMI,15,20,21,27 CABG,29,30 or the elderly22,23 while others use mixed populations in an analysis.31 Few studies control for these different CAD subpopulations when evaluating clinical outcomes. To be complete, studies should adequately account for confounding variables related to the ability to participate with on-site CR, such as discharge disposition (eg, home self-care, hospice, and skilled nursing) or geographic distance from the on-site CR center. Women should be adequately represented; and if demographics such as age and sex are reported, they should be controlled for in the analysis.17,32,33 There are other potential sources of bias associated with trials of clinical outcomes possibly resulting in null findings or confounding an accurate estimate of CR effects.34 Controlling for these factors to the extent possible often requires accessing individual health records and may result in inconsistent findings. Although most studies suggest a benefit from CR, inconsistent results suggest further investigations are still needed.17,20,35–38

To address some of the limitations of previous reports, this single-site cohort study assessed CR effects on all-cause readmission and mortality to 180 d from the index referring cardiovascular event. We only included patients with minimal barriers to participation in on-site CR and addressed several confounding factors noted previously, including demographics, discharge criteria, underlying risk factors, and medication—all confirmed by chart review.

METHODS

DATA SOURCE: PATIENT ELIGIBILITY

We used the Duke Hospital financial database to identify patients discharged with a cardiovascular disease during fiscal year 2011 and 2012 (July 1, 2010, to June 30, 2012, inclusive). These years include modern-era statin therapy and modern revascularization protocols. Patients were identified based on CR eligibility at discharge using diagnostic-related group, primary diagnostic and primary procedure codes defined by International Classification of Diseases (ICD), Ninth Revision, Clinical Modification (ICD-9-CM) codes.

This list was further narrowed to only those patients with known occlusive cardiovascular disease diagnosis making them eligible for CR: AMI, PCI, or CABG surgery. As the purpose of this study was to evaluate the effects of CR on secondary prevention of CAD, patients with valve repair/replacement, heart transplant, and stable/atypical angina were excluded. We further excluded patients who had died prior to discharge; and those with a discharge disposition of extended skilled nursing home, hospice, rehabilitation facility, correctional facility, federal hospital, long-term acute care, or other type of institutions (eg, psychiatric). Finally, patients < 18 yr and those living > 50 mi from the CR facility were omitted from the study population. Figure 1 shows the consort diagram for the study.

Figure 1.

Patient flow diagram.

DATA SOURCE: PATIENT DEMOGRAPHICS AND CLINICAL CHARACTERISTICS

Demographic and clinical characteristics, including modifiable risk factors, comorbidities, and medications, were extracted by inspecting hospital discharge notes of each patient in the Duke EPIC (Maestro Care) electronic medical record. Each CR-eligible patient chart was inspected for CR participation, including time from discharge to first session and number of CR sessions attended. To document mortality status at 180 d after discharge, records were checked for date of death and last known clinic visit. Readmissions or deaths within 180 d from the qualifying hospitalization were identified and included in the analysis. The protocol was approved by the Institutional Review Board of Duke University. No subject informed consent was obtained because of exempt protocol status, nor, as this was not considered a randomized clinical trial, is there an associated clinicaltrials.gov registry number.

STATISTICAL ANALYSIS

This was a retrospective analysis to evaluate the impact of CR versus no CR on all-cause mortality and all-cause readmission in patients following their natural clinical course inclusive of 180 d after hospital discharge. The maximum time increment that the Centers for Medicare & Medicaid Services allows for an eligible patient to complete 36 CR sessions for reimbursement is 180 d. Thus, 180 d was chosen for survival analyses below. Initially at discharge, every patient was offered CR. Only patients who attended ≥ 1 CR session within 180 d of discharge were categorized into the CR group for the main analyses. Using means ± SD for continuous variables and frequencies for nominal variables, baseline characteristics were summarized by group. Differences between groups at baseline were determined by t tests for continuous variables and χ² tests for nominal variables; baseline being defined as date of hospital discharge. Survival was considered free from rehospitalization or death up to 180 d. As all-cause readmission is confounded with mortality, we performed an analysis using a composite outcome of time to first event: mortality or all-cause readmission. Therefore, time-to-event was defined in two ways: time to readmission as a first event using mortality or end-of-study as a censoring variable; and time to either mortality or readmission as a first event with end-of-study as a censoring variable. Kaplan-Meier (K-M) estimates were used to initially assess overall survival differences between the two groups (ie, those who attended CR at any point during follow-up vs those who did not attend CR). Statistical significance was determined for the product limit estimates by the log-likelihood ratio test.

We used a series of four nested Cox proportional hazards models to estimate the effect of CR and other covariates on both outcomes. Results are represented as HR and 95% CI. The log (-log [survival]) for group plot assessed the proportionality assumption. The validity of the proportionality assumption was checked by visual inspection.

• Model 1: Unadjusted comparison of the HR between groups (CR vs non-CR)

• Model 2: CR versus non-CR group HR adjusted for demographics (age, sex, and race)

• Model 3: CR versus non-CR group HR adjusted for demographics and comorbidities (Model 2 plus depression, anxiety, dyslipidemia, hypertension, obesity, smoking history, and type 2 diabetes)

• Model 4: CR versus non-CR group HR adjusted for demographics, comorbidities, and discharge criteria (Model 3 plus qualifying discharge criteria: AMI, PCI, CABG, AMI + PCI, or AMI + CABG)

All eligible patients were offered CR at discharge from the hospital. However, some time elapsed between discharge and entry into the CR program, leading to what is labeled as “immortal time.” Immortal time occurs when the patient is not at risk for the event of interest (rehospitalization or death) because the predictor of interest (CR participation) has not occurred yet and the benefit of CR cannot be accrued prior to session beginning. To account for the time to enrollment into the CR program, which differed by patient, CR was entered into the survival analyses as a time-varying variable.

In the fully adjusted model (Model 4) for both outcomes, we conducted two sensitivity analyses: to explore whether the addition of medications attenuate the relationship between CR and the outcomes, aspirin, statins, and β-blockers were included as covariates; and to explore the dose effect of CR, the number of CR sessions attended was included as a covariate. All analyses were performed with SAS version 9.4 (SAS Institute), under a .05 (two-tailed) type I error criterion for statistical significance.

RESULTS

Among the 2641 patients included in the analysis, 214 (8.1%) attended ≥ 1 CR session and were included in the CR group. Of the patients included in the CR group, 93% attended ≥ 5 CR sessions. On average, patients attended their first CR session within 41 ± 40 d of hospital discharge. Patients attended 24.8 ± 12.1 sessions over a mean period of 80 ± 48 d. The mean number of sessions/wk attended was 2.5 ± 1.1. Table 1 lists the baseline characteristics by group. The proportion of women was the same between the two groups (∼35%) and reflects previous findings: women and minorities are less likely to attend CR compared with men.24,39,40 The discharge criteria differed between groups: the non-CR group had a higher proportion of PCI only (33.0 vs 14.0%; P < .001); the CR group had higher proportions of CABG only (29.0 vs 19.3%; P < .001) and AMI plus PCI (33.2 vs 23.9%; P = .003). Type 2 diabetes (38.4 vs 30.4%; P = .019) and smoking history (59.7 vs 34.6%; P < .001) were more prevalent in the non-CR group. The CR group had a higher prevalence of hyperlipidemia compared with the non-CR group (81.3 vs 69.0%; P < .001) and was more likely to be taking the medications of a β-blocker, statin, and aspirin; the non-CR group had a higher prevalence of insulin and antianginal medications. During the 180 d of follow-up, 12.1% and 18.7% of the CR and non-CR patients were readmitted to the hospital, respectively. There was one death (0.5%) in the CR group during the follow-up period, while 98 deaths (4.0%) occurred in the non-CR group. In addition, we calculated the estimated K-M rates of all-cause readmission for 30, 60, and 90 d after discharge. At all three time points, the non-CR group had approximately double the estimated event rates compared with the CR group (30 d: 9.1 vs 4.7%; 60 d: 12.9 vs 6.1%; and 90 d: 15.4 vs 7.5%).

Table 1. Baseline Characteristics by Cardiac Rehabilitation Group^a.

Characteristic	Cardiac Rehabilitation (n = 214)	No Cardiac Rehabilitation (n = 2427)	P Value
Age, yr	64.8 ± 10.9	62.9 ± 12.6b	.032
Sex
Male	65.9	64.7	.755
Female	34.1	35.3	.755
Race
White	80.4	70.5b	.001
Black	14.9	25.3	.001
Other	4.2	3.8	N/A
Unknown	0.5	0.4	N/A
Discharge criteria
AMI only	20.6	20.4	.954
PCI only	14.0	33.0b	<.001
CABG only	29.0	19.3b	<.001
AMI + PCI	33.2	23.9b	.003
AMI + CABG	2.8	3.5	.611
Comorbidities and controllable risk factors
Type 2 diabetes	30.4	38.4b	.019
Hypertension	74.8	76.6	.531
Hyperlipidemia	81.3	69.0b	<.001
History of smoking	34.6	59.7b	<.001
Obesity	33.2	40.0	.052
Depression or anxiety	23.8	17.1	.058
Cerebrovascular disease	11.2	13.3	.368
Chronic obstructive pulmonary disease	12.6	13.6	.737
Heart failure	13.1	12.8	.824
Peripheral artery disease	6.1	9.0	.154
Renal disease	13.1	16.0	.277
Medications
β-blocker	93.5	85.9b	.002
ACE inhibitor	51.4	57.9	.055
Statin	91.6	85.2b	.012
Ca+ channel blocker	10.7	15.0	.092
Aspirin	94.9	89.1b	.008
Nitroglycerin	52.8	47.8	.167
Insulin	11.7	20.3b	.002
Other glucose control agent	12.1	15.6	.128
Antianginal	0.5	9.7b	<.001
Diuretic	43.5	39.6	.264
Anticoagulant	20.1	20.7	.827

Abbreviations: ACE, angiotensin-converting-enzyme; AMI, acute myocardial infarction; CABG, coronary artery bypass graft surgery; PCI, percutaneous coronary intervention.

^aValues represented as mean ± SD or %.

^bP ≤ .05 for difference between groups; italicized text indicates significance (P < .05).

READMISSION OUTCOME

Figure 2 shows the K-M graph illustrating the crude proportion readmission difference in those who attended CR at any point versus those who did not. In the unadjusted model, the independent association of CR, as a time-varying covariate, with all-cause readmission trended toward significance, with patients in the CR group having a 38.3% reduction in all-cause readmission risk compared with those who did not attend CR (HR = 0.62: 95% CI, 0.36-1.06; P = .078; unadjusted Model 1). As shown in Table 2, after adjusting for demographics, comorbidities, and discharge criteria, the magnitude of the relationship between CR and readmission was slightly attenuated (HR = 0.66: 95% CI, 0.38-1.13; P = .127; adjusted Model 4). The fully adjusted Model 4 showed that age (HR = 1.02/yr: 95% CI, 1.01-1.02; P = .0003), depression or anxiety (HR = 1.32: 95% CI, 1.06-1.65; P = .015), type 2 diabetes (HR = 1.49: 95% CI, 1.23-1.80; P < .0001), and discharge criteria of AMI only (HR = 1.46: 95% CI, 1.11-1.93; P = .007) relative to CABG only were independently associated with greater risk of readmission, while men had a significantly decreased risk of readmission (HR = 0.66: 95% CI, 0.55-0.79; P < .0001) relative to women (Table 3). Neither the sex-by-CR group nor the discharge criteria-by-CR group interaction terms were statistically significant, and thus were not included in any of the subsequent models. We conducted a sensitivity analysis by adding medications—aspirin, statins, and β-blockers—to Model 4; none were significant predictors nor did the inclusion of these medications attenuate the relationship between CR group and readmission risk (data not shown). In the second sensitivity analysis, adding the number of CR sessions attended expectedly attenuated the relationship between CR group and readmission risk; however, the independent association of the number of CR sessions attended trended toward significance, where each session attended resulted in a 2% risk reduction in all-cause readmission (HR = 0.98/session: 95% CI, 0.96-1.00; P = .057; adjusted Model 4).

Figure 2.

Unadjusted survival curves for all-cause readmission within 180 d of hospital discharge according to the cardiac rehabilitation (CR) group. Kaplan-Meier estimates were used to assess the crude proportion readmission difference in those who attended CR at any point versus those who did not. Those who attended CR at any point during the 180-d follow-up had a lower risk of all-cause readmission compared with the No-CR group. The blue line represents the CR group. The red line represents the no-CR group. Censoring variable defined as mortality or end-of-study. This figure is available in color online (www.jcrpjournal.com).

Table 2. Risk of All-Cause Readmission^a.

	HR (95% CI)	P Value
Model 1
Unadjusted	0.62 (0.36-1.06)	.078
Model 2
+ Demographics	0.62 (0.36-1.06)	.078
Model 3
+ Demographics
+ Comorbidities	0.65 (0.38-1.12)	.121
Model 4
+ Demographics
+ Comorbidities
+ Discharge criteria	0.66 (0.38-1.13)	.127

^aHRs indicate risk of all-cause readmission for the cardiac rehabilitation (CR) group; the no-CR group is used as a referent category. Demographics: age, sex, and race; comorbidities: depression, anxiety, dyslipidemia, hypertension, obesity, smoking history, and type 2 diabetes; qualifying discharge criteria: acute myocardial infarction, percutaneous coronary intervention, or coronary artery bypass graft surgery.

Table 3. Covariate Effects From Multivariable Analysis for Risk of All-Cause Readmission or Risk of All-Cause Readmission Plus Mortality^a.

	Risk of All-Cause Readmission		Risk of All-Cause Readmission or Mortality
	HR (95% CI)	P Value	HR (95% CI)	P Value
Cardiac rehabilitation	0.66 (0.38-1.13)	.127	0.57 (0.33-0.98)	.043
White	1.16 (0.95-1.42)	.138	1.09 (0.90-1.32)	.403
Age	1.02 (1.01-1.02)	.0003	1.02 (1.01-1.03)	<.0001
Male	0.66 (0.55-0.79)	<.0001	0.69 (0.57-0.82)	<.0001
Depression or anxiety	1.32 (1.06-1.65)	.015	1.34 (1.08-1.66)	.007
Dyslipidemia	0.91 (0.74-1.11)	.349	0.88 (0.73-1.06)	.177
Hypertension	1.08 (0.84-1.38)	.551	1.07 (0.85-1.35)	.565
Obesity	1.05 (0.86-1.28)	.638	1.06 (0.88-1.28)	.542
Smoking history	1.04 (0.86-1.25)	.709	1.08 (0.90-1.30)	.416
Type 2 diabetes	1.49 (1.23-1.80)	<.0001	1.53 (1.28-1.83)	<.0001
AMI only	1.46 (1.11-1.93)	.007	1.69 (1.29-2.20)	.0001
PCI only	0.98 (0.75-1.28)	.859	1.00 (0.77-1.30)	.992
AMI + PCI	0.84 (0.62-1.14)	.258	0.90 (0.67-1.21)	.492
AMI + CABG	1.47 (0.90-2.39)	.126	1.58 (0.99-2.53)	.056

Abbreviations: AMI, acute myocardial infarction; CABG, coronary artery bypass graft surgery; PCI, percutaneous coronary intervention.

^aReferent category for discharge criteria was CABG only; italicized text indicates significance (P < .05).

COMPOSITE OUTCOME

Figure 3 shows the K-M graph illustrating the crude survival difference in those who attended CR at any point versus those who did not for the composite outcome. Given there was only one death in the CR group, the graphic for readmission alone and the composite of death and readmission appear almost identical. Patients in the CR group had a significant 45.8% reduction in all-cause readmission or mortality risk (HR = 0.54: 95% CI, 0.32-0.93; P = .025; unadjusted Model 1). As shown in Table 4, the magnitude of the CR effect was not attenuated in the subsequent models after adjustment for demographics (Model 2), comorbidities (Model 3), and discharge criteria (Model 4). After adjustment for age, sex, race, depression, anxiety, dyslipidemia, hypertension, obesity, smoking, type 2 diabetes, and discharge criteria, the final model revealed a significant 42.7% reduction in readmission or mortality risk for patients who attended CR (HR = 0.57: 95% CI, 0.33-0.98; P = .043; Model 4; Table 3). The interaction terms for sex-by-CR group and discharge criteria-by-CR group were not statistically significant, and thus were not included. Furthermore, the inclusion of aspirin, statins, and β-blockers in Model 4 did not result in any additional significant predictors, nor did the inclusion of these medications attenuate the magnitude of the CR group effect (data not shown). Although, as expected, including the number of CR sessions attended attenuated the CR group effect, the number of CR sessions attended was significantly related to the composite outcome, where each session attended resulted in a 2% risk reduction in all-cause readmission or mortality (HR = 0.98/session: 95% CI, 0.96-1.00; P = .03; adjusted Model 4).

Figure 3.

Unadjusted survival curves for all-cause readmission or death within 180 d of hospital discharge according to cardiac rehabilitation (CR) group. Kaplan-Meier estimates were used to assess the crude survival difference in those who attended CR at any point versus those who did not for the composite outcome. Those who attended CR at any point during the 180-d follow-up had a lower risk of all-cause readmission or death compared with the no-CR group. The blue line represents the CR group. The red line represents the no-CR group. Censoring variable defined as mortality or end-of-study. This figure is available in color online (www.jcrpjournal.com).

Table 4. Risk of All-Cause Readmission or All-Cause Mortality^a.

	HR (95% CI)	P Value
Model 1
Unadjusted	0.54 (0.32-0.93)	.025
Model 2
+ Demographics	0.53 (0.31-0.91)	.022
Model 3
+ Demographics
+ Comorbidities	0.58 (0.34-0.99)	.044
Model 4
+ Demographics
+ Comorbidities
+ Discharge criteria	0.57 (0.33-0.98)	.043

^aHR indicate risk of all-cause readmission or death for the cardiac rehabilitation (CR) group; the no-CR group is used as a referent category. Demographics: age, sex, and race; comorbidities: depression, anxiety, dyslipidemia, hypertension, obesity, smoking history, and type 2 diabetes; qualifying discharge criteria: acute myocardial infarction, percutaneous coronary intervention, or coronary artery bypass graft surgery; italicized text indicates significance (P < .05).

DISCUSSION

The primary aim of this study was to investigate if and to what extent the risk of 180-d all-cause readmission or all-cause readmission plus death is decreased by CR in a single-site rigorously documented study by analyzing >2600 individual subject medical records. The most important finding of this study was that, when controlling for demographic characteristics, controllable risk factors/comorbidities, and cardiovascular referral diagnosis, the risk of 180-d all-cause readmission or death was significantly decreased in patients who attended CR compared with those who did not attend CR. Not only did CR result in a 43% risk reduction for readmission or mortality, but also the HR maintained a range of 5% between all four models (Table 4) for this composite outcome; this finding emphasizes the beneficial effects of CR in a way not previously shown. The results were further supported after accounting for use of aspirin, statins, and β-blockers. To our knowledge, the aggregate of these potential modulators of poor clinical outcomes has never been controlled in an analysis of CR efficacy, thus further strengthening the case for CR as being beneficial to patients with CAD.

Many studies have evaluated demographic and clinical factors affecting clinical outcomes, readmission, recurrent myocardial infarction, or death, <180 d after hospital discharge in the CAD population. However, statistical models incorporating CR participation are notably lacking. To allow time for both referral and enrollment in the CR program, we entered CR into our analyses as a time-varying variable. In our study, the average time from discharge to CR participation was 41 d. While only 42% of patients started CR within the recommended 30 d, the 180-d survival analysis for the composite outcome showed the CR group still had a significant 43% reduced risk for all-cause readmission or death in the fully adjusted model. In fact, the magnitude of relative risk reduction remained the greatest for CR compared with the other significant covariates in the fully adjusted model (Table 3). Moreover, our sensitivity analysis including the number of CR sessions attended further supports the benefits of CR participation. On average, CR group patients attended 25 sessions, with 93% of those patients completing ≥5 CR sessions. Importantly, attending even five sessions equated to a 10% significantly reduced risk of readmission or mortality. These findings are supported by previous CR dose-response studies examining mortality risk across a variety of CAD patient populations,41–45 including a recent study documenting a fully adjusted HR of 0.98: 95% CI, 0.97-0.99/session for major adverse cardiovascular events over a median 6 yr of follow-up, representing an apparent linear benefit of CR session attendance, with no apparent minimal nor ceiling effect.45

Review of existing literature investigating the efficacy of CR reveals significant heterogeneity of included potential confounders of clinical outcomes relative to CR, as many individual studies do not provide a listing of confounders in their analyses, or they are not included as covariates in the statistical model. Furthermore, meta-analyses are designed to assess effects across heterogeneous studies. This may be why some studies conclude CR does not beneficially affect clinical outcomes of readmission or the composite of readmission and mortality.35,36 Notably, a recent study by Bush et al46 documented the protective effects of CR on hospital readmission risk after controlling for multiple covariates. The study included older (65-88 yr), Medicare-only patients and found when CR was initiated within 60 d of discharge it was related to a 30% lower risk in all-cause hospitalization at 1 yr compared with those who did not initiate CR. Important design similarities compared with the current study are that efforts were made to exclude patients with mobility limitations or not discharged to home care and confounding factors included demographic characteristics, comorbid conditions, and medications. Differences compared with the current study are that all patients were AMI survivors only and had a revascularization procedure; distance from the CR center was not an exclusion criterion; all-cause readmission was evaluated at 1 yr; the analytic approach employed inverse probability of treatment weights to adjust for baseline confounding variables; and the outcomes included cumulative incidence and relative risk.

Interestingly, most studies reporting reduced readmission or mortality risk state the value to be approximately 20-25%.15 Based on our results, the effect size is greater, in the range of 42-47%, for the composite outcome of readmission or mortality; however, we did not observe a statistically significant reduction in risk for readmission alone. Nevertheless, our analysis when compared with others lends support to the conclusion that studies likely would show greater efficacy for CR if confounding factors were properly controlled. In addition, our methodology highlights the importance of careful chart review for obtaining accurate patient characteristics at the time of discharge and accounting for time to CR enrollment.

STRENGTHS AND LIMITATIONS

This study has several strengths, including a large single-site sample size and analyses carefully controlling for potential confounders obtained through a detailed review of individual patient records. We limited our inclusion criterion to only those residing close enough (∼50 mi) to access the CR program, discharged to home self-care, and ambulatory, thus increasing the validity of the CR versus no-CR comparisons.

Nevertheless, there are limitations of studies of this kind. Patients were not randomized, as CR participation was voluntary (it would be unethical to randomize to no CR); thus, we were unable to determine a true cause-and-effect relationship. However, after controlling for several factors identified through individual electronic medical record assessment, our findings provide strong support for CR participation being an impactful mediator for the observed reduction in readmission or mortality risk. There are unmeasured confounders affecting CR enrollment and participation, such as referral bias and self-selection bias, and our database query did not account for whether CR enrollment followed a first or second cardiovascular event. Additional potential confounders for which we could not control due to lack of access to relevant data include medication adherence, transportation availability, presence of caregiving, and socioeconomic status.

CONCLUSIONS

This single-site, highly controlled, retrospective observational study improves the understanding of the benefits of CR on clinical outcomes in patients after hospitalization for AMI, PCI, or CABG. After controlling for demographics, comorbidities, risk factors, discharge diagnosis, and medical therapy, those enrolled in CR had marked reductions in risk of all-cause readmission or mortality compared with those who did not participate in CR. As our results demonstrated a greater magnitude of risk reduction (∼42-47%) for readmission or mortality compared with previous findings, this study provides further support to clinical practice guidelines about the impactful and crucial role CR plays in mediating clinical outcomes. Overall, our study highlights the need for additional prospective studies identifying and controlling for key individual-level variables to adequately assess the true beneficial effects CR participation.