National Use and Effectiveness of β-Blockers for the Treatment of Elderly Patients After Acute Myocardial Infarction:  National Cooperative Cardiovascular Project FREE

DESPITE THE IMPORTANCE of β-blockers for secondary prevention after acute myocardial infarction (AMI),^1- 2 several studies have suggested that they are substantially underutilized,^3- 10 particularly in older patients. Although these studies agree that increasing the appropriate use of β-blockers is an important opportunity to improve secondary prevention after an AMI, they differ in study design and approach, yielding very different estimates of the use of β-blockers. None provide a national perspective on the prescribed use of β-blockers after an AMI in the treatment of elderly patients. The studies that have linked patterns of use of β-blockers with outcomes have been restricted to very selected samples⁷ or have been dependent on claims databases for clinical information⁹ but generally have found that β-blockers are associated with increased survival.

To address these issues, we sought to describe the contemporary use of β-blockers at hospital discharge among elderly AMI survivors who are ideal candidates for the therapy; to compare the use of β-blockers across the United States; to identify the most important predictors of the prescribed use of β-blockers at discharge; and to determine the independent association between the use of β-blockers and mortality in clinical practice. To perform this study, we evaluated the medical records of more than 200000 hospitalizations of Medicare beneficiaries from across the country in 1994 and 1995 with a principal discharge diagnosis of AMI, as part of the Cooperative Cardiovascular Project (CCP). The project, a Health Care Financing Administration (HCFA) collaboration with health care professionals and peer review organizations, was designed to examine patterns of care and stimulate improvements in the care and outcomes of Medicare beneficiaries with AMI.

The study sample was obtained from patients in the CCP cohort and includes Medicare patients aged 65 years or older from nongovernmental acute care hospitals in the United States and Puerto Rico with a principal discharge diagnosis of AMI, International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 410, in 1994 and 1995. We did not sample admissions that were not related to the acute care of an AMI (in which the fifth digit of the ICD-9-CM code is 2). In the CCP states, patients were identified during an approximately 8-month period (varying in each state) using hospital bills (UB-92 claims data) in the Medicare National Claims History File. The file includes all patients treated under fee-for-service plans but does not include all of the patients treated as part of Medicare managed care risk contracts. For patients admitted to hospitals in the CCP pilot states (Alabama, Connecticut, Iowa, and Wisconsin), hospitalizations were sampled from a 4-month period (August 1995 through November 1995). The hospitalizations that were abstracted from the CCP pilot states occurred after a feedback intervention based on an initial data collection from admissions in 1992 and 1993. Sampling in Minnesota was slightly delayed and some hospitals were undersampled so as not to interfere with an ongoing special study, the Minnesota Clinical Comparison and Assessment Project (MCCAP).

The study sample was restricted to patients aged 65 years or older with a confirmed AMI who were discharged alive. An AMI was defined as a discharge diagnosis of an AMI and either a creatine kinase (CK) MB level higher than 0.05 or elevation of lactate dehydrogenase (LDH) more than 1.5 times normal and LDH-1 level higher than LDH-2 level, or 2 of the following 3 criteria: chest pain, a 2-fold elevation of the CK level, or evidence of AMI on the electrocardiogram. We excluded patients who died during the index infarction. For patients who were hospitalized more than once in the sample period, only the first AMI admission was included. We excluded hospitalizations in which patients were transferred to another acute care institution because we were unable to determine their discharge medications. We excluded 3 patients in whom mortality could not be confirmed or dated. We also excluded patients who were considered to have a terminal illness, since the goals of their treatment may not have focused on a survival benefit. (Patients were considered terminally ill if their records documented that they were terminally ill or had a life expectancy of less than 6 months; patients with "do not resuscitate" status were not excluded if they did not meet the criteria for terminal illness.)

We developed restricted cohorts of patients who would be considered ideal candidates for β-blocker therapy. Patients with contraindications were excluded from the study whether or not they were taking β-blockers. However, there is no clear consensus about which patients should be considered ideal candidates for the medication. From the group of patients above, we constructed a definition for the primary study sample of ideal candidates for postdischarge β-blocker therapy (those without strong contraindications to β-blockers) based on our clinical experience and the medical literature. In developing this sample, we balanced the need to exclude patients with probable contraindications with an effort not to be overly restrictive. Consequently, we excluded patients with the following contraindications: bradycardia (heart rate <50/min while not taking β-blockers), low blood pressure (systolic blood pressure <100 mm Hg), high-grade atrioventricular block, asthma, chronic lung disease, heart failure during the hospitalization, or chart-documented intolerance to β-blockers. All contraindications were determined by chart review except for asthma, which was also based on a secondary diagnostic code (ICD-9-CM code 493.xx).

To determine if our results were dependent on our particular definition of the sample, we performed the analyses using 3 other definitions of ideal patients. We defined a sample based on the relative contraindications for β-blockers listed in the American College of Cardiology (ACC)/American Heart Association (AHA) Guidelines for the Treatment of Acute Myocardial Infarction.¹ This sample (ACC/AHA cohort) was similar to our primary sample except that we also excluded patients with heart rates less than 60/min (rather than 50/min), peripheral vascular disease, or diabetes mellitus with insulin therapy.¹ Since the ACC/AHA guidelines state that β-blockers are not as strongly indicated in low-risk patients, we also developed a subset of the ACC/AHA sample that excluded patients without any of the following high-risk characteristics, based on the guidelines: prior myocardial infarction, an anterior myocardial infarction, advanced age (we chose 75 years as the cutoff), or hemodynamic evidence of left ventricular dysfunction (defined as Killip class >1). Finally, we defined a sample based on the previously published analysis of the initial CCP pilot. For this sample, we excluded patients with the following characteristics: heart rate less than 50/min, left ventricular ejection fraction (LVEF) less than 0.35, heart failure, systolic blood pressure less than 100 mm Hg, chronic obstructive pulmonary disease, depression, diabetes with insulin therapy, dementia, high-grade atrioventricular block, no significant coronary disease by catheterizations, and low-risk characteristics (no recurrent chest pain, no arrhythmias, no prior AMI, normal stress test, and no LVEF <0.50).⁵

To obtain the information for this project from the medical records, HCFA established 2 clinical data abstraction centers to abstract records. These private organizations were responsible for the reliability and efficiency of the abstraction. Trained technicians abstracted predefined variables from copies of the hospital record and entered them directly into a computer database using interactive software available through the HCFA Internet home page at http://www.usccp.org. Data reliability was monitored by random reabstractions, with overall variable agreement averaging more than 90%. The agreement for β-blocker at discharge was 97.5% and the κ was 0.64.¹¹

The outcome variable for the first phase of the study was the frequency of β-blockers as a discharge prescription. All discharge medications were reviewed and oral β-blockers were identified. Topical β-blockers were excluded. The principal end point for the second phase of the study was mortality within 1 year of discharge. This information was ascertained from the Medicare Enrollment Database, which was derived from the Master Beneficiary Record from Social Security Administration data, a valid source of vital status.^12- 13

The independent variables in this study included age, sex, race, medical history, hospital and discharge medications, clinical status, hospital complications, hospital procedures, discharge disposition, and length of stay. Age was categorized into 3 strata: 65 through 74 years, 75 through 84 years, and 85 years and older. Comorbidities included a chart-documented history of hypertension, diabetes mellitus, renal dysfunction (defined as blood urea nitrogen level >14.3 mmol/L [40 mg/dL] or creatinine level >221 µmol/L [2.5 mg/dL]), history of myocardial infarction, albumin level less than 0.003 g/L (0.03 g/dL), anemia (hematocrit <0.30), and dementia. Hospital treatment variables on the first day of admission included aspirin, β-blockers, and thrombolytic therapy. Hospital procedure variables included cardiac catheterization, percutaneous coronary revascularization, and coronary artery bypass surgery. Variables that describe clinical events or patient characteristics at any time during the hospitalization included atrial fibrillation or flutter, stroke, recurrent chest pain, ventricular tachycardia, intubation, and CK levels more than 4 times the normal level. Left ventricular systolic dysfunction was defined as an LVEF of less than 0.35. Left ventricular ejection fraction levels were measured using 1 of the following 3 methods: radionuclide ventriculography, cardiac catheterization, or echocardiogram, prioritized in that order. Hospital length of stay was coded as more than 12 days (yes or no), the 85th percentile for length of stay. Discharge medication variables included aspirin, calcium channel blockers, and angiotensin-converting enzyme inhibitors. Discharge disposition was coded as home or to a non–acute care facility. For variables with more than 3% missing values (ie, prothrombin time, hematocrit, creatinine, and blood urea nitrogen), we created a dummy variable to indicate missing. We classified the physicians' specialties on the basis of the attending physician listed in Medicare Part A claims. Each physician's specialty was identified by linking his or her unique physician identification number with a directory of physician-reported specialties maintained by HCFA. For the geographic analysis, we classified the patients into subgroups based on major regions defined by the US Bureau of the Census: New England, Mid Atlantic, South Atlantic, East North Central, East South Central, West North Central, Mountain, and Pacific.

For the first phase of the study, we sought to determine the frequency with which elderly patients with an AMI were being appropriately discharged with a prescription for a β-blocker. In several cohorts, we determined the number of patients considered appropriate for β-blockers and their prescribed use. Using the primary study sample, we compared the use of β-blockers by state and territory. Among those patients who were not receiving β-blocker therapy prior to admission, we evaluated the bivariate associations of demographic, clinical, physician specialty, and geographic variables with the prescription of β-blockers at discharge. Then, using the variables from the bivariate analysis, we developed a multivariable logistic regression model by backward stepwise selection with the prescription of β-blockers as the dependent variable. Variables were dropped from the model at a significance level of P <.05.

For the second phase of the study, we estimated the association between the prescribed use of β-blockers at discharge and 1-year survival, in an unadjusted model and after adjusting for potential confounders using a proportional hazards model. In this model, covariates selected on the basis of clinical relevance were forced in the model. Covariates included demographic characteristics (age, sex, and race), clinical history (history of diabetes, dementia, myocardial infarction, heart failure, smoking, hypertension, renal dysfunction, or stroke), hospitalization data (LVEF <0.35), a variable indicating whether an LVEF was measured, hematocrit, albumin, heart failure, recurrent chest pain, ventricular tachycardia, length of stay more than 12 days, CK level more than 4 times normal, cardiac catheterization, percutaneous transluminal coronary angioplasty, coronary artery bypass surgery, the prescription of aspirin and angiotensin-converting enzyme inhibitors at discharge, physician specialty, and the census region of the country. We also evaluated the potential interaction between the prescribed use of β-blockers at discharge and age, sex, race, and LVEF by constructing separate fully adjusted models with interaction terms and by compared results in stratified samples. For cohorts in which patients with specific characteristics were excluded (eg, diabetes), that variable was dropped from the model.

All calculations were performed using the software program STATA 5.0 (STATA Corp, College Station, Tex).

Of the 115,015 eligible patients (ie, age ≥65 years, confirmed AMI, survived the hospitalization, no terminal illness, and not transferred), 69,707 patients (61%) had 1 or more of the following possible contraindications to β-blocker treatment: 1.1% with bradycardia, 3.8% with high-grade atrioventricular block, 20.7% with asthma or chronic obstructive pulmonary disease, 5.7% with low blood pressure, 47.0% with heart failure, and 1.0% with reported intolerance to β-blockers. Thus, the primary study sample of 45,308 elderly AMI patients were defined as ideal candidates for β-blocker therapy at discharge (Table 1).

The ideal candidates for β-blockers had a mean (SD) age of 75.3 (7.1) years; 45.7% were women. Comorbidity was common, with 61.0% having a history of hypertension, 25.9% with diabetes, and 4.4% with renal dysfunction. Prior heart disease was also common, with 24.9% with a prior myocardial infarction and 6.6% with prior heart failure. In the year after discharge, 10.1% died. Compared with the patients who met study criteria but had at least 1 strong contraindication to β-blockers, the ideal patients were significantly younger (75.3 years vs 77.2 years; P<.001) and less likely to die in the year after discharge (10.2% vs 27.7%; P<.001).

Of the 115015 eligible patients, 42822 (37.2%) had a β-blocker prescribed at discharge. Of the 45308 ideal patients in the study sample, 22665 (50.0%) had a β-blocker prescribed at discharge. Table 2, Part A, and Table 2, Part B, shows the prescribed use of β-blockers at discharge with various demographic, clinical, and geographic characteristics of the patients. Figure 1 shows the prescribed use of β-blockers at discharge across the country. There is significant variation by state and by region. The 5 states with the highest use of β-blockers were Connecticut (77.1%), Massachusetts (74.2%), Maine (68.3%), New Hampshire (68.9%), and Vermont (66.7%). The lowest use occurred in Mississippi (30.3%), Puerto Rico (32.1%), Oklahoma (33.5%), Arkansas (33.5%), and Nevada (36.4%). Compared with New England, all the other census regions had significantly lower rates.

Several other factors were notable. The prescribed use of calcium channel blockers at discharge had a strong negative association with the prescribed use of β-blockers. Physician specialty was also an important predictor, with cardiologists and internists associated with higher use of β-blocker therapy than general and family practice physicians.

Among the 8,513 patients receiving β-blocker therapy prior to admission, 6659 (78.2%) were receiving therapy at the time of discharge. Of the 36795 patients not receiving β-blocker therapy prior to admission, 16006 (43.5%) had therapy initiated on or before discharge. The initiation of β-blocker therapy in these patients was associated with several demographic and clinical characteristics, but these variables alone produced a model with only moderate discriminant ability (area under the receiver operating characteristic curve [AUC] = 0.65). Adding discharge medications, particularly calcium channel blockers, increased the AUC to 0.73. In the final model (Table 3), which included regions and physician specialty, the AUC was 0.75.

Among the ideal patients, the 1-year mortality rate was 7.7% for those who were prescribed β-blockers compared with 12.6% for those who were not (P<.001). The adjustment for baseline characteristics attenuated the strength of the association between β-blocker therapy and survival (Table 4). However, in the fully adjusted model, the benefit persisted, with β-blocker therapy at discharge being associated with a 14% lower risk of mortality (P<.001).

There was no significant interaction with mortality between the prescription of β-blockers at discharge and age, sex, or LVEF. In stratified analyses, the adjusted relative risk of mortality associated with the use of β-blockers was 0.81 (95% confidence interval [CI], 0.72-0.91) for patients aged 65 through 74 years, 0.88 (95% CI, 0.81-0.98) for patients aged 75 through 84 years, and 0.88 (95% CI, 0.78-0.99) for patients aged 85 years or more. The adjusted relative risk of mortality associated with the use of β-blockers was 0.83 (95% CI, 0.76-0.91) for men and 0.89 (95% CI, 0.81-0.97) for women. The adjusted relative risk of mortality associated with the use of β-blockers was 0.91 (95% CI, 0.76-1.08) for patients with an LVEF less than 0.35 and 0.83 (95% CI, 0.76-0.91) for patients with an LVEF of 0.35 or more.

We also developed several other samples of patients who would be considered candidates for β-blockers (Table 5). Compared with our definition, the ACC/AHA cohorts and the CCP cohorts were more restrictive. Despite the differences in the definitions of the samples, they had similar rates of β-blocker use and a similar association between β-blockers and survival.

Our principal finding is that in 1994 and 1995, across the United States, 63% of elderly survivors of an AMI were not prescribed β-blocker therapy at discharge. Many of these patients had at least 1 strong contraindication to the therapy as documented by detailed medical record review; however, among those who did not have a strong contraindication for long-term β-blocker therapy, half were not prescribed the drug at discharge. In addition, more restrictive cohorts of patients who did not have any of the relative or absolute contraindications to β-blocker therapy did not have substantially higher rates. Given that mortality after AMI is high in the elderly and that β-blockers reduce mortality in this group, our findings reveal an ample opportunity to improve the care and outcomes for such patients. The variation in the initiation of β-blockers among the ideal cohort is not well explained by clinical or demographic factors alone. However, discharge medications, particularly calcium channel blockers, did explain much more of the variation. Ideal candidates for β-blockers who received calcium channel blockers were much less likely to be prescribed β-blockers. This observation, which is disturbing given the long-term benefit of β-blockers in this population, has also been observed by others.⁹

Some of the remaining variation that is not explained by clinical and demographic factors or discharge medications is explained by the region in which the patients were hospitalized or the specialty of the admitting physician. Compared with the New England states, physicians in many other regions were significantly less likely to prescribe β-blockers to appropriate patients. Among states, the use of β-blockers ranged from 30.3% in Mississippi to 77.1% in Connecticut. The results demonstrate heterogeneity in the opportunity for improvement, although even states and regions with higher use did not excel in this area. Patients admitted by cardiologists and internists were much more likely to be discharged receiving β-blockers than those admitted by general or other types of physicians, demonstrating that opportunities for improvement are not equal among physician groups.

Prior studies have also reported underutilization of β-blockers for secondary prevention after an AMI.^4,6,8- 9,14 Some of the estimates were even lower than what we report. For example, Soumerai and colleagues⁹ merged 3 large administrative databases with data from 1987 through 1992 to examine the use of β-blockers among elderly survivors of AMI from New Jersey. Using discharge codes and medication dispensing (or claims) files to define a study sample of 3737 patients without a contraindication to β-blockers, they found that 21% of eligible patients filled a prescription for β-blocker therapy within 90 days of discharge, a much lower estimate than we report. With their approach, about 70% of the Medicare survivors of AMI in New Jersey were considered eligible for therapy, a much higher estimate than what we found using chart review data. Their less selective denominator may have led to a lower estimate. Also, the patients in their study were hospitalized 2 to 7 years before the patients in our group. Finally, the use of prescription data may have overlooked patients who were prescribed β-blockers at discharge but did not fill a prescription within 90 days. On the other hand, our study would have misclassified patients who were prescribed β-blockers at discharge but did not adhere to the treatment plan. The New Jersey study also showed a much more prominent benefit associated with the use of β-blockers, with a 43% lower mortality rate compared with the 14% that we report in this study. Which estimate is correct? Whereas the difference between the prescribed use at discharge and the actual use may have underestimated the benefit of β-blockers in our study, the use of administrative data to determine the patients who were ideal for treatment and the covariates used in the multivariable analysis may have overestimated the benefit in the Soumerai study. Our result is much closer to the estimated 20% reduction found in 16 randomized trials involving more than 18000 patients.¹⁵ The use of chart review data (rather than administrative data) to adjust for imbalances among the groups in their susceptibility to mortality may have provided our study with an advantage in avoiding the effects of confounding. Nevertheless, both studies provide reassurance that β-blockers, tested most thoroughly in younger populations, are also effective in older patients. These results reinforce the use of β-blockers after AMI as a quality indicator by the National Committee on Quality Assurance as part of HEDIS 3.0.

One challenge in this study was defining the study sample. We chose to create an ideal group of patients by excluding those with strong contraindications to β-blockers. Heart failure was one such contraindication because β-blockers should be initiated in this group of patients only when they are stable. We felt that it would be reasonable to defer the initiation of β-blockers in patients with clinical heart failure in conjunction with an AMI. To address concerns about the study sample, we also identified several patient groups. More restrictive definitions of the sample excluded larger numbers of patients. Nevertheless, all the samples revealed underutilization as well as an association between the use of β-blockers and survival.

This study does have some limitations. First, although we determined the prescribed use of β-blockers based on retrospective chart review, we could not determine patterns of use after discharge. As a consequence, we may have misclassified the long-term pattern of use in some patients. However, for the survival analysis this misclassification would have tended to diminish the association between the use of β-blockers and improved survival. Second, our ability to ascertain contraindications to β-blockers was limited by the information that is documented in the charts. Third, the study sample was hospitalized in 1994 and 1995, and it is likely that improvements in care have occurred in the interim, perhaps partly as a result of the CCP.¹⁶ Nevertheless, these data present striking and nonuniform opportunities to improve care across the country, and it is unlikely that those opportunities no longer exist. The results should stimulate hospitals and physicians to assess their current level of performance.

The study also has several important strengths. It represents the most comprehensive evaluation of the use of β-blockers in elderly survivors of AMI. As part of a national initiative to improve care for elderly patients with AMI, this study involved the chart abstraction of more than 200000 records across the country in 1994 and 1995. The charts were abstracted by trained professionals using standardized definitions with high reliability. Also, because the patients were Medicare beneficiaries, we had the opportunity to determine long-term mortality for the cohort.

In summary, we report that β-blockers are underutilized in elderly patients after AMI. Variation across the country is striking, but all areas can improve the care of such patients by increasing the appropriate use of β-blockers. The results of our study reinforce the survival benefit associated with the use of β-blockers and suggest the need for a national effort to address this issue.