Effect of Initiating Cardiac Rehabilitation after Myocardial Infarction on Subsequent Hospitalization in Older Adults

Abstract

Purpose:

Outpatient cardiac rehabilitation (CR) participation after myocardial infarction (MI) reduces all-cause mortality; however, less is known about effects of CR on post-MI hospitalization. The study objective was to investigate effects of CR on hospitalization following acute MI among older adults.

Methods:

Medicare beneficiaries between 65 and 88 yr hospitalized in 2008 with acute MI, who survived at least 60-d post-discharge, had a revascularization procedure during index hospitalization, and did not have an MI in previous year were eligible for this study. CR initiation was assessed in the 60-d post-discharge. Competing risk survival analysis was used to estimate the proportion of discharged beneficiaries hospitalized between the end of 60-d exposure window and December 31, 2009 treating death as a competing event.

Results:

The mean age of 32,851 Medicare beneficiaries meeting study criteria was 75 ± 6 yr, approximately half were male (52%), and majority were white (88%). In this study, 21% of beneficiaries initiated CR within the exposure window. At 1-yr post discharge, CR initiators had a lower risk of recurrent MI (4.2% 95%CI, 3.5 - 5.1%), cardiovascular (15.7% 95%CI, 14.3 - 17.2%), and all-cause (30.4% 95%CI, 28.8 - 32.1%) hospitalization than non-initiators (5.2% 95%CI, 5.0 - 5.5%; 18.0% 95%CI, 17.6 - 18.4%; 33.2% 95%CI, 32.5 - 33.8% respectively). There was no difference in fracture risk (negative control outcome).

Conclusions:

This study provides evidence that CR can reduce the 1-yr risk of cardiovascular and all-cause hospital admissions in Medicare aged MI survivors.

Condensed Abstract

The effect of cardiac rehabilitation (CR) was investigated in a population of Medicare beneficiaries surviving an acute myocardial infarction (MI). At 1 yr post discharge, CR initiators had a lower risk of cardiovascular hospitalization than non-initiators accounting for the competing risk of death. CR participation should be encouraged even among older adult MI survivors.

Each year an estimated 580,000 U.S. adults experience a myocardial infarction (MI) and approximately 86% of these patients survive the event.¹ Patients with a history of MI have a higher risk of a recurrent MI or other cardiovascular (CV) disease than the overall population.¹ For MI survivors, American Heart Association guidelines recommend the use of evidence-based medications and outpatient cardiac rehabilitation (CR) to reduce the risk of future CV events²

CR programs have demonstrated a protective effect on mortality similar to that of pharmacotherapy when compared to usual care in meta-analyses of randomized control trials comprised of patients hospitalized with an MI or acute coronary syndrome. ^3,4 The clinical trials in these meta-analyses have focused on mortality outcomes, while few investigators reported the benefits of CR on CV readmission or all-cause hospitalization after an MI. ^3,4 Because of the heterogeneous results from studies investigating the association between CR and subsequent hospitalization after an index event, more investigation is needed to improve the level of evidence of short-term CR benefits.^5–7 Since previous clinical trials enrolled younger mostly white male participants, an opportunity exists to clarify the efficacy of CR in older and more gender diverse populations than have previously been studied.^5,6,8 Given the large US population that is affected by recurrent CV events and the limited evidence on the efficacy of CR on non-mortality outcomes in older adults, it is essential to investigate the routine and optimal use of secondary prevention methods in this population.

The goal of this study was to investigate the effect of initiating CR within 60 d post-MI on cardiac related and all-cause hospitalizations in a modern medical environment. For each type of hospital admission outcome, the 1-yr risk difference between CR initiators and non-initiators was computed among Medicare beneficiaries. While traditional hazard ratios are appropriate for estimating the effects of CR on all-cause mortality, competing risk methods are needed to investigate the association of CR on subsequent non-fatal outcomes such as hospital admission. ⁹ Since CR is underutilized in practice, we also performed stratified analyses by age, comorbidity score, and cardiac medication use to explore subpopulations that can be targeted for future efforts to improve CR participation. ^10–12 Our study provides important knowledge regarding understudied hospitalization outcomes after CR using advanced analysis methods.

METHODS

STUDY POPULATION

Medicare beneficiaries between 65 and 88 yr who were hospitalized between January 1, 2008 and December 31, 2008 for acute myocardial infarction (AMI), had a revascularization procedure during the index hospitalization period, and were continuously enrolled in Medicare Part A, B, and D between January 2007 and death or December 2009 were eligible for this study. We defined AMI by International Classification of Diseases, Clinical Modification 9^th revision (ICD-9) diagnostic code 410.xx (excluding 410.x2) in the first or second discharge code position from Medicare Provider Analysis and Review (MedPAR) files. Revascularization was identified from ICD-9 procedure codes, Healthcare Common Procedure Coding System (HCPCS) codes, and Current Procedural Terminology (CPT) codes (SDC 1). Beneficiaries were excluded from this study if they had an AMI in the year prior to the index AMI; had history of limited mobility (claims for paralysis, bed sores, Parkinson’s disease, home hospital bed, or wheelchair)¹³; died within 60 d of index discharge; or were not discharged home after index hospitalization period.

STUDY DESIGN

This is a retrospective study of healthcare utilization after AMI hospitalization. The timeline used to identify study measures is presented in Figure 1. The index hospital period began with AMI admission and concluded at discharge after any contiguous transfers to other facilities, if applicable. The 12 mo before index hospitalization was used to define baseline covariates. CR initiation was assessed in the 60-d exposure window following the index hospitalization period. CR initiation was defined by any occurrence of CPT codes 93797 or 93798 in Medicare data within the 60-d exposure period. Outcomes were assessed from the end of the 60-d exposure window until December 31, 2009.

Figure 1.

Study Design

OUTCOME MEASURES

The primary outcome for this study was AMI admission. Secondary outcomes were hospitalization for any reason or a major CV event (MACE) defined by ICD-9 codes for AMI, angina, heart failure, or stroke: 410.xx, 411.1, 428.xx, 430, 431, 432, 433.x1, 434.x1, 435, and 436.¹⁴ To evaluate the presence of unmeasured confounding, we defined bone fractures at any location (SDC 2) as a negative control outcome. ¹⁵ We did not expect to see an association between CR and fractures.¹⁵ Beneficiaries were administratively censored on December 31, 2009 (end of follow-up). Mortality was defined by the date of death in Medicare enrollment data.

Covariates

We identified confounders from beneficiary demographic, clinical, and prescription refill characteristics. Demographic characteristics (age, gender, and race) were obtained from Medicare enrollment files. US Census-block median household income was identified by linking beneficiary zip codes to 2010 American Community Survey data.¹⁶ Comorbid conditions (Gagne¹⁴, Elixhauser¹⁷, SDC 3) during baseline and index hospitalization period were identified from ICD-9 and HCPCS codes in MedPAR, outpatient, and Carrier Medicare files. Faurot et al ¹³ developed a statistical model to estimate the probability of diminished daily activities from administrative claims data. Applying these model coefficients, we estimated the probability of diminished daily activities as a proxy for frailty in this study. Total hospitalization days and ≥ 1 d in an intensive care unit and/or coronary care unit were identified from MedPAR files. Prescription claims from Medicare Part D files linking National Drug Codes (NDCs) and/or generic names of medications to Anatomical Therapeutic Chemical (ATC) codes were used to identify guideline-directed medical therapy (GMT) and other CV medications during baseline and the exposure period. Post-MI GMT consists of antiplatelet therapy, beta-blockade therapy, angiotensin converting enzyme inhibitors, angiotensin receptor blockers, and statins.^18,19 Although aspirin has multiple indications, we assumed that prescription claims in this population were for prevention of CV events.

STATISTICAL ANALYSIS

Descriptive statistics were used to describe characteristics of the study population. The number of person years of follow-up, number of hospital events, and the proportion of deaths were also summarized for each exposure group. Inverse probability of treatment (IPT) weights were computed to adjust for confounding between CR initiation and subsequent CV hospitalization after index AMI. ²⁰ Age and length of hospital stay were included in the models as restricted cubic splines with 3 and 4 knots, respectively. ²¹ We examined the distribution of the predicted probabilities from the final logistic regression model for overlap between exposure groups. In a sensitivity analysis, conditions defined by Elixhauser et al were used to model comorbidity instead of the Gagne score in computing IPT weights. ¹⁷ Standardized differences were calculated to detect differences between CR initiators and non-initiators in unweighted and IPT weighted populations.

Cumulative incidence of each outcome was estimated using non-parametric cumulative incidence estimators with death treated as a competing event.²² The 1-yr risk difference and risk ratio between CR initiators and non-initiators were computed from cumulative incidence estimates with and without IPT weight adjustment separately for each outcome. Confidence intervals were computed using a non-parametric bootstrap procedure.²³ The 1-yr risk difference calculated from IPT weighted cumulative incidence estimators represents the population average treatment effect. ²⁴ We conducted stratified analyses by age group (65 - 74 and 75 - 84 yr), exposure period GMT count (0, 1-2, and 3-4) and Gagne comorbidity score (<5 or ≥5). Gagne et al. defined high risk of mortality as ≥ 17% chance of death within 1 yr, corresponding to a comorbidity score ≥5.¹⁴ All analyses were completed using SAS 9.4.

RESULTS

We identified 32, 851 beneficiaries who met study criteria. The majority of the study population were male (52%), white (88%), and the mean age at index AMI was 75 ± 6 yr. CR was initiated differently across several key baseline characteristics such as race, type of revascularization procedure, and mean comorbidity score as indicated by the unadjusted standardized differences ≥10% for these population characteristics (Table 1). By applying stabilized inverse probability of treatment (IPT) weights, we were able to eliminate the large measured imbalances between exposure groups. The standardized differences between treatment groups were < 5% in the IPT weighted population for all but a few population characteristics.

Table 1.

Characteristics of CR Initiators (n = 6,916) and CR Non-Initiators (n = 25,935).

	Unadjusted			IPT Weighted
	CR Non-Initiators	CR Initiators	Standardized Difference	Standardized Difference
Demographics
Age, yr	75.1 ± 6.0	74.2 ± 5.6	−0.157	0.001
Female	12,788 (49)	2,933 (42)	−0.139	−0.013
White	22,462 (87)	6,570 (95)	0.294	−0.037
Index Hospitalizations
Revascularization
Angioplasty	21,358 (82)	5,101 (74)	−0.209	−0.007
CABG	4,741 (18)	2,004 (29)	0.254	0.015
Stent	19,539 (75)	4,690 (68)	−0.167	−0.011
Any ICU Stay	15,477 (60)	4,265 (62)	0.041	0.009
Any CCU Stay	12,311 (48)	3,374 (49)	0.026	0.007
Hospital Transfer Count
0	19,605 (76)	5,725 (83)	0.178	0.003
1	5,080 (20)	1,026 (15)	−0.126	0.001
>2	1,250 (5)	165 (2)	−0.131	−0.008
Baseline Comorbidities
Gagne Score	1.7 ± 2.2	1.0 ± 1.8	−0.322	0.015
Baseline Conditions
Hypertension	21,664 (84)	5,486 (79)	−0.108	−0.002
Hyperlipidemia	18,509 (71)	5,275 (76)	0.112	−0.007
Uncomplicated Diabetes	10,727 (41)	2,271 (33)	−0.177	0.013
Congestive Heart Failure	8,947 (34)	1,629 (24)	−0.243	−0.007
Cardiac Arrhythmia	7,992 (31)	2,007 (29)	−0.039	0.035
Chronic Pulmonary Disease	7,844 (30)	1,481 (21)	−0.203	0.018
Osteoporosis or Osteoarthritis	6,319 (24)	1,656 (24)	−0.010	0.005
Complicated Diabetes	3,147 (12)	502 (7)	−0.165	0.013
Cancer	2,745 (11)	753 (11)	0.010	0.007
Cerebrovascular Disease	2,588 (10)	471 (7)	−0.114	−0.011
Rheumatic Disease	1,242 (5)	326 (5)	−0.004	0.006
Day 60 GMT Group
0	1,651 (6)	235 (3)	−0.138	−0.046
1-2	7,695 (30)	1,896 (27)	−0.050	0.017
3-4	16,589 (64)	4,785 (69)	0.111	0.005

Data presented as mean ± SD or n (%).

Abbreviations: CABG, coronary artery bypass grafting; CCU, cardiac care unit; CR, outpatient cardiac rehabilitation; GMT, Guideline-directed medical therapy; ICU, intensive care unit; IPT, inverse probability of treatment weighted.

Only 21% of the study population initiated CR during the exposure period. Initiators participated in an average of 10 ± 6 sessions by 60 d post-discharge. In the weighted population, CR initiators contributed 9,758 person-yr of follow-up time for subsequent AMI outcome and had an observed mortality rate of 2.1 beneficiaries per 100 person-yr while CR non-initiators contributed 35,672 person-yr of follow-up time and had an observed mortality rate of 3.5 beneficiaries per 100 person-yr. The observed mortality rate was lower in the CR initiator group than in the non-initiator group for all outcomes studied (Table 2).

Table 2.

Summary of Person-Time on Study, Risk, Risk Difference, and Relative Risk of Study Outcomes at 1yr Post-AMI for Outpatient Cardiac Rehabilitation Initiators verses Non-Initiators

Outcome	Cardiac Rehab	Person years	Events	Deaths	Mortality Rate (/100 person years)	Unadjusted % Risk (95% CI)	Unadjusted Risk Difference	Adjusted % Risk (95% CI)	Adjusted Risk Difference	Adjusted Risk Ratio
AMI	Non-Initiators	35,354	1,900	1,362	3.9	5.6 (5.3 - 5.9)		5.2 (5.0 - 5.5)
	Initiators	9,960	250	127	1.3	2.7 (2.3 - 3.1)	−2.9 (−3.0 to −2.8)	4.2 (3.5 - 5.1)	−1.0 (−1.5 to −.4)	.81 (.70 - .92)
MACE	Non-Initiators	28,638	5,591	958	3.3	19.2 (18.7 - 19.8)		18.0 (17.6 - 18.4)
	Initiators	8,808	960	109	1.2	11.7 (10.9 - 12.5)	−7.5 (−7.8 to −7.3)	15.7 (14.3 - 17.2)	−2.4 (−3.3 to −1.2)	.87 (.81 - .93)
All-Cause Admission	Non-Initiators	21,466	8,462	725	3.4	34.6 (33.9 - 35.4)		33.2 (32.5 - 33.8)
	Initiators	7,020	1,947	92	1.3	26.0 (24.9 - 27.2)	−8.6 (−9.0 to −8.1)	30.4 (28.8 - 32.1)	−2.8 (−3.7 to −1.7)	.92 (.89 - .95)
Fracture	Non-Initiators	36,915	804	1,466	4.0	2.1 (2.0 - 2.3)		2.0 (1.8 - 2.1)
	Initiators	10,169	142	136	1.3	1.4 (1.1 - 1.7)	−0.8 (−0.9 to −0.6)	1.9 (1.4, - 2.4)	−.1 (−.4 to .3)	.94 (.78 - 1.15)

Abbreviations: MACE, major adverse cardiovascular event (AMI, angina, heart failure, or stroke); AMI, acute myocardial infarction.

Plots of the cumulative incidence during follow-up (Figure 2) reveal that CR initiators had a lower risk of hospital admission (solid lines) and competing death event (dashed lines) than non-initiators. Confounding adjustment by IPT weighting attenuated risk differences between exposure groups for each outcomes (Table 2). We did not observe an association between CR and fractures (Table 2) so no additional steps were taken to control for unmeasured confounding. Side by side depiction of adjusted cumulative incidence curves for AMI, MACE, and all-cause hospital admission illustrate only a small difference between CR initiators and non-initiators for each of these outcomes (Figure 2). This small difference persisted for all 3 outcomes during follow-up. At 1-yr post discharge there was a < 3% absolute risk reduction in AMI (1.0%), MACE (2.4%), and all-cause (2.8%) hospitalizations among CR initiators when compared to non-initiators (Table 2) after adjusting for confounding and the competing risk of death. The small absolute differences had corresponding risk ratios (95% CI) of 0.81 (0.70, 0.92), 0.87 (0.81, 0.93), and 0.92 (0.89, 0.95) for AMI, MACE, and all-cause hospitalization respectively.

Figure 2.

Inverse Probability of Treatment Weighted Cumulative Incidence of All Study Outcomes and Competing Risk of Death in Outpatient Cardiac Rehabilitation Initiators compared to Non-Initiators

We also investigated if there was a difference between exposure groups in the risk of each outcome by age group, comorbidity score, and exposure period GMT use (Table 3). The absolute 1-yr risk difference for recurrent AMI between CR initiators and non-initiators was similar for both age groups. However, beneficiaries in the older age group had a larger 1-yr risk reduction than younger age group for all-cause admissions. Beneficiaries with a Gagne comorbidity score ≥5 had approximately 2.5 times the 1-yr risk of experiencing a CV outcome than beneficiaries with a score < 5. A negative risk difference between CR initiators and non-initiators was observed for all study outcomes among beneficiaries with a Gagne comorbidity score < 5 but not in the higher comorbidity score group. Beneficiaries who had 1-2 GMTs at day 60 had a negligible decrease in the 1-yr risk of AMI between CR initiators and non-initiators but at least a 2% decrease in the 1-yr risk of MACE or all-cause admission outcomes. Beneficiaries who had 3-4 GMTs at day 60 had the highest risk reduction of hospital admission in the GMT stratified analyses regardless of outcome. The sensitivity analysis using the Elixhauser definitions of comorbid conditions produced estimates that were closer to the null than the main analysis but did not change our conclusions (data not shown).

Table 3.

Stratified Analysis of Risk, Risk Difference, and Relative Risk of Study Outcomes at 1-Year Post-AMI for Outpatient Cardiac Rehabilitation Initiators verses Non-Initiators

	CR Non-Initiators % Risk (95% CI)	CR Initiators % Risk (95% CI)	Risk Difference (95% CI)	Risk Ratio (95% CI)
AMI
Age Group
65-74, yr	4.6 (4.2 - 5.0)	3.7 (2.8 - 4.9)	−.8 (−1.4 to −.0)	.82 (.66 - 1.00)
75-84, yr	5.8 (5.4 - 6.2)	4.7 (3.4 - 6.2)	−1.1 (−2.1 to −.0)	.81 (.62 - .99)
Morbidity Risk
Gagne Score < 5	4.7 (4.4 - 5.0)	3.5 (2.8 - 4.1)	−1.2 (−1.6 to −.8)	.74 (.64 - .83)
Gagne Score >=5	10.6 (9.5 - 11.9)	11.2 (6.2 - 17.1)	.6 (−3.3 to 5.3)	1.05 (.65 - 1.44)
Day 60 GMT Group
0	5.2 (4.1 - 6.3)	7.0 (1.3 - 15.3)	1.8 (−2.8 to 9.0)	1.34 (.33 - 2.43)
1-2	5.7 (5.2 - 6.2)	4.8 (3.0 - 6.5)	−.9 (−2.1 to .3)	.84 (.59, - 1.05)
3-4	5.0 (4.7 - 5.4)	3.8 (2.9 - 4.5)	−1.3 (−1.8 to −.9)	.75 (.62 - .84)
MACE
Age Group
65-74 yr	15.4 (14.8 - 16.0)	13.6 (12.0 - 15.4)	−1.8 (−2.8 to −.6)	.88 (.81 - .96)
75-84 yr	20.0 (19.2 - 20.9)	18.0 (15.6 - 21.3)	−2.0 (−3.6 to .4)	.90 (.81 - 1.02)
Morbidity Risk
Gagne Score < 5	16.3 (15.8 - 16.8)	13.5 (12.6 - 14.7)	−2.7 (−3.2 to −2.1)	.83 (.80 - .88)
Gagne Score >=5	37.9 (35.9 - 39.9)	36.1 (26.4 - 45.7)	−1.7 (−9.5 to 5.8)	.95 (.74 - 1.15)
Day 60 GMT Group
0	20.0 (18.0 - 22.0)	21.5 (12.8 - 30.7)	1.5 (−5.1 to 8.7)	1.08 (.71 - 1.40)
1-2	19.1 (18.1 - 19.9)	17.1 (14.7 - 19.6)	−2.0 (−3.4 to −.3)	.90 (.81 - .98)
3-4	17.4 (16.8, 18.0)	14.6 (12.7 - 16.5)	−2.8 (−4.1 to −1.5)	.84 (.76 - .92)
All-Cause Admission
Age Group
65-74 yr	29.9 (29.0 - 30.9)	28.7 (26.7 - 31.3)	−1.2 (−2.3 to .4)	.96 (.92 - 1.01)
75-84 yr	35.7 (34.6 - 36.6)	32.6 (28.9 - 35.4)	−3.0 (−5.7 to −1.2)	.92 (.84 - .97)
Morbidity Risk
Gagne Score < 5	31.3 (30.7 - 31.9)	27.7 (26.0 - 28.9)	−3.6 (−4.7 to −3.0)	.89 (.85 - .91)
Gagne Score >=5	56.6 (54.0 - 58.8)	57.0 (46.9 - 65.8)	.4 (−7.1 to 7.0)	1.01 (.87 - 1.12)
Day 60 GMT Group
0	33.0 (30.1 - 36.0)	39.7 (28.5 - 51.4)	6.7 (−1.6 to 15.4)	1.20 ( .95 - 1.43)
1-2	34.1 (32.5 - 35.4)	31.8 (28.4 - 35.0)	−2.3 (−4.1 to −.4)	.93 (.87 - .99)
3-4	32.8 (32.0 - 33.6)	29.1 (26.8 - 31.1)	−3.7 (−5.2 to −2.5)	.89 (.84 - .93)

Abbreviations: AMI – acute myocardial infarction; GMT, guideline-directed medical therapy; MACE, major cardiovascular event (AMI, angina, heart failure, or stroke);

DISCUSSION

The low (4 - 5%) 1-yr risk of recurrent AMI reported in this study is consistent with a recent randomized clinical trial of the effects of CR after MI on outcomes by West et al.²⁵. The observed low CR initiation was similar to what has been reported in earlier studies of Medicare beneficiaries with an index AMI occurring in 1997 (24%) and AMI patients referred for CR included in a national registry between 2007 and 2010 (23%).^10,26 In our study, a small difference in absolute estimates of the 1-yr risk of recurrent AMI between CR initiators and non-initiators translated into a large relative difference due to the low risk of recurrent AMI in each group. The observed relative risk of recurrent AMI within 1 yr of discharge of .81 (95% CI, .70 - .92) was similar to results reported in a meta-analysis of CR clinical trials by Clark et al. (.83 95% CI, .74 - .94).³ While Clark reported a statistical difference in recurrent MI between rehabilitation users and controls³, West concluded there was no difference in risk of recurrent MI between CR users and controls.²⁵ Based upon prior studies and our results, we conclude that the small effect that CR initiation has on the 1-yr risk of recurrent AMI is not the most important result of this study when compared to the effect of CR on the other hospitalization outcomes.

The 1-yr risk of MACE and all-cause hospitalization was 4 - 7 times > the 1-yr risk of AMI in this study. While the magnitude of the 1-yr risk difference between CR initiator and non-initiators was similar for both MACE and all-cause hospitalization outcomes, the relative measure of CR initiation was dissimilar between outcomes due to the difference in the risk of each outcome for the non-initiators. In a study of CR participation among Olmsted County Minnesota residents with an average follow-up time of 7.6 yr, Dunlay et al reported a 20% and 25% decrease in the long-term relative risk of CV admission and all-cause hospitalization respectively between initiators and non-initiators.⁷ With a 1-yr hazard ratio of .85 (95% CI, .79 - .91) for MACE and .89 (95%CI, .84 - .94) for all-cause hospitalization, our study supports a protective effect of CR participation on these hospitalization outcomes.

Stratified analysis showed that older beneficiaries (>75 yr) in this study had a greater reduction in all-cause hospitalization risk than MACE risk This is an important observation given a recent American Heart Association scientific statement that encourages the use of secondary prevention methods in elderly patients and notes the lack of information on the effect of exercise therapy on non-mortality outcomes.²⁷ In addition, the risk of each study outcome for CR initiators with 1-2 medications at day 60 is similar to the risk of that same outcome in CR non-initiators with 3-4 medications at day 60. More research is needed to investigate the interaction CR participation and GMT use especially among MI survivors who cannot tolerate or are non-compliant with all GMTs.

Conducting research using administrative claims data has both limitations and benefits. First, administrative claims data were created for financial purposes rather than research. While there is some overlap in these objectives, research relevant details are absent from claims data where these objectives begin to diverge. For example, administrative data lack clinical details such as laboratory results, smoking status, and body mass index. In this study, we attempted to overcome this potential for unmeasured confounding using propensity score techniques but realize that residual confounding may still exist in our results. Because there was an insignificant difference between CR initiators and non-initiators for the fracture negative control outcome, we surmised that residual confounding did not significantly bias the results for the other study outcomes.

Our study population was restricted to beneficiaries who received a revascularization procedure during their index hospitalizations for several reasons. First, previous literature has reported that revascularization is highly associated with referral for CR;²⁸ therefore, restriction may impose a level of homogeneity of treatment groups that should reduce indication bias. Since we cannot measure which of the non-initiators were not referred for CR, this restriction should eliminate from our risk set a pool of patients who would never initiate CR because they were never referred. However, restriction by revascularization is not a perfect proxy for referral since we are also potentially removing patients from our population that were referred for CR.

We were limited to claims for beneficiaries who were continuously enrolled in Medicare Parts A, B, and D from 1 year before their index hospitalization until December 31, 2009 or death. By restricting to the continuously covered population, we may be limiting the generalizability of our study findings beyond the stably insured. In addition these results may have limited generalizability to late adopters who may appropriately start CR after 60 d. However, the size and nature of the population presented is still of public health importance. There was the potential for misclassification of exposure to GMT due to use of over the counter aspirin, free medication samples,²⁹ and purchase of prescription drugs without filing an insurance claim.³⁰ Besides over-the-counter aspirin use, generic options of the medication classes included in this study were offered at prices low enough to be regularly purchased using cash without filing a claim. These cash transactions would not be captured in our study data. However, we expect the use of these medications to be non-differential with respect to the use of CR services.

There are several strengths to this study. Although the study investigates beneficiaries experiencing an AMI in 2008, our results are important because we are adding to the limited knowledge about the effects of CR on subsequent hospitalization events.⁵ Furthermore, most of the clinical trials of CR were small (<300 enrollees) enrolling younger (< 60 yr) predominately male populations, conducted before improvements in secondary prevention medications, and conducted in countries outside the US where access to care is different.⁵ The larger sample size in our study compared to clinical trials, the inclusion of large percentage of women, and inclusion of older adults are advantages to this study. In addition, our claims-based analysis is not subject to recall bias since we have documentation on the receipt and timing of health services studied. While filling a prescription does not guarantee that the medication has been taken, it is likely a better measure of medication exposure than definitions by prescriptions written at discharge.

In conclusion, our results suggest that outpatient CR may reduce CV and all-cause hospital admissions 1-yr post discharge in elderly AMI survivors. However, public health interventions are needed to improve the observed low rates of CR initiation for these benefits to be fully realized in AMI survivors.