Aspirin in the Treatment of Acute Myocardial Infarction in Elderly Medicare Beneficiaries
Patterns of Use and Outcomes
Background Although aspirin is an effective, inexpensive, and safe treatment of acute myocardial infarction, the frequency of use of aspirin in actual medical practice is not known. Elderly patients, a group with low rates of utilization of effective therapies such as thrombolytic therapy, also may be at risk of not receiving aspirin for acute myocardial infarction. To address this issue, we sought to determine the current pattern of aspirin use and to assess its effectiveness in a large, population-based sample of elderly patients hospitalized with acute myocardial infarction.
Methods and Results As part of the Cooperative Cardiovascular Project Pilot, a Health Care Financing Administration initiative to improve quality of care for Medicare beneficiaries, we abstracted hospital medical records of Medicare beneficiaries who were hospitalized in Alabama, Connecticut, Iowa, or Wisconsin from June 1992 through February 1993. Among the 10 018 patients ≥65 years old who had no absolute contraindications to aspirin, 6140 patients (61%) received aspirin within the first 2 days of hospitalization. Patients who were older, had more comorbidity, presented without chest pain, and had high-risk characteristics such as heart failure and shock were less likely to receive aspirin. The use of aspirin was significantly associated with a lower mortality (OR, 0.78; 95% CI, 0.70 to 0.89) after adjustment for potential confounders.
Conclusions About one third of elderly patients with acute myocardial infarction who had no contraindications to aspirin therapy did not receive it within the first 2 days of hospitalization. The elderly patients with the highest risk of death were the least likely to receive aspirin. After adjustment for differences between the treatment groups, the use of aspirin was associated with 22% lower odds of 30-day mortality. The increased use of aspirin for patients with acute myocardial infarction is an excellent opportunity to improve the delivery of care to elderly patients.
Aspirin is an effective, inexpensive, and safe treatment of AMI. In 1988, ISIS-2 reported that aspirin therapy reduced short-term mortality in patients with suspected AMI by 23%.1 An informal cost-effectiveness analysis estimated that the use of aspirin in this setting cost $13 per life saved.2 In 1990, a joint task force of the American College of Cardiology and the American Heart Association recommended that aspirin therapy should be initiated immediately for all patients with AMI and continued for at least 1 month.3
Despite the importance of aspirin therapy, there is evidence that not all patients with AMI receive aspirin.4 5 6 The Survival and Ventricular Enlargement (SAVE) study investigators reported that the use of aspirin significantly increased among the patients entered in their trial after the publication of the ISIS-2 trial.6 They also noted that in 1989, the last year of the SAVE trial, more than a quarter of their patients did not receive aspirin for AMI. Other more recently published surveys have also suggested an underutilization of aspirin.5 7 Elderly patients may be the group that needs aspirin the most and is least likely to receive it. AMI is a leading cause of morbidity and mortality in the growing elderly population, with 30-day mortality rates >20%.8 Since a high proportion of elderly patients with an AMI do not have ECG criteria for thrombolytic therapy,9 aspirin therapy may provide the best opportunity to improve outcomes in these patients. Investigators have consistently reported that elderly patients are less likely than younger patients to receive a variety of cardiac medications, tests, and procedures.10 11 12 13 At least one report suggested that older patients with AMI may be less likely to be treated with aspirin than younger patients.14 There is a need to evaluate the use of aspirin in a large group of elderly patients to determine current medical practice and to characterize patients who are least likely to receive aspirin.
There is also a need to augment current available information about the effectiveness of aspirin for the treatment of AMI in actual clinical practice in the elderly population. Even though ISIS-2 did not have an age restriction, only 3411 of its 17 187 subjects (20%) were >70 years old. Furthermore, although the overall effectiveness of aspirin for AMI is widely accepted by experts,3 fewer than one half of the generalists and only three quarters of the cardiologists believe that it definitely improves survival even among patients <75 years old.15
We undertook this study to provide an in-depth analysis of the current pattern of aspirin use and to assess its effectiveness in a large, population-based sample of elderly patients. This study is part of the Cooperative Cardiovascular Project Pilot, a Health Care Financing Administration initiative to improve the quality of care for Medicare beneficiaries with AMI.7 16 17 The Cooperative Cardiovascular Project Pilot cohort includes more than 16 000 Medicare patients discharged from hospitals in Alabama, Connecticut, Iowa, and Wisconsin from June 1, 1992, through February 28, 1993, with a principal discharge diagnosis of AMI. Our objectives were (1) to determine the current rates of aspirin therapy in treating AMI in elderly patients without contraindications for aspirin therapy, (2) to identify the characteristics of patients not treated with aspirin, and (3) to evaluate the effectiveness of the short-term use of aspirin on 30-day mortality in this group by using multivariable methods to adjust for baseline differences between the patients who did and did not receive aspirin.
This cohort was drawn from patients identified from Medicare Provider Analysis Record (MedPAR) files with a principal discharge diagnosis of AMI (ICD-9-CM 410)18 in Alabama, Connecticut, Iowa, or Wisconsin between June 1, 1992, and February 28, 1993. Admissions that were not related to the care of an AMI (the fifth digit of the ICD-9-CM code, 2) and admissions that resulted from a transfer from another institution were not included. Hospital charts were requested for the admissions. The study sample included patients ≥65 years old who did not have absolute contraindications to aspirin therapy (active bleeding or allergy to aspirin). Patients were excluded if they did not have an AMI as documented by having a peak creatine kinase–MB fraction >5% or LDH2 >1.5 times the normal value and LDH1 > LDH2 or two of the three following criteria: chest pain, a twofold elevation of the creatine kinase, or new Q waves on the ECG. Patients were also excluded if they died on the first hospital day, since they might not have had the opportunity to receive aspirin.
Collection of Data
Trained nurses and medical record technicians abstracted hospital medical records. Data were directly entered into a computerized database management system using the Uniform Clinical Data Set System.19 On-line data definitions and range checks were used to decrease errors and variability. A 5% random sample of charts was reabstracted by a different individual to evaluate interobserver variability.
Variables in this study included age, sex, race, history of hypertension, diabetes, smoking, history of MI, history of congestive heart failure, prior coronary revascularization, history of stroke, history of chronic obstructive pulmonary disease, medications on admission, presence of chest pain on admission, duration of symptoms before admission, presence of congestive heart failure on admission, vital signs, admission laboratory tests (blood urea nitrogen, creatinine, hematocrit, and prothrombin time), treatment with thrombolytic therapy on the first day, β-adrenergic blocking agents within the first 2 days and/or heparin within the first 2 days, shock, intubation, and Killip class. Systolic blood pressure, pulse, and respiratory rate were taken as the highest value recorded within 24 hours after admission. Renal dysfunction was defined as blood urea nitrogen >40 mg/dL or creatinine >2.0 mg/dL. Patients were considered to have a terminal illness if their records documented that they were terminally ill or had a life expectancy of <6 months, if the admission orders indicated that the patient should be given palliative care only, or if a “do not resuscitate” order was written at the time of admission. To summarize risk for use in figures, the cohort was stratified by Killip class.
In the first set of analyses, the principal outcome was the use of aspirin within the first 2 hospital days. The time period was chosen to allow for the possibility that some patients might receive aspirin in the Emergency Department or just before presentation, and it might not have been charted in the medical record. Patients who were given any medications that contained aspirin were considered to have been treated with aspirin. We reabstracted 594 charts to test the reliability of the data collection. For the variable of aspirin during the hospitalization, the κ value was 0.966. The principal end point for the outcome analysis was 30-day mortality ascertained from the Medicare Enrollment Database. In addition, we tabulated potential complications that might have resulted from the use of aspirin, including in-hospital hemorrhage, transfusions, dialysis, and stroke.
First, we evaluated the bivariate association between the use of aspirin and a family of variables that included demographic (age, sex, and race) and clinical (coronary artery disease risk factors, past medical history, admission cardiac medications, cardiac history, acuity of presentation, and ECG presentation) characteristics. Variables that were associated with the use of aspirin with a value of P<.10 in the bivariate analysis were considered to be candidate variables in the development of a stepwise multivariate logistic regression model with use of aspirin as the dependent variable. This model was constructed with an entry significance level of P=.01 and exit significance level of P=.05. Partial residual plots were used to identify possible problems with the model.
Second, we compared the rates of potentially aspirin-related complications between the two treatment groups. χ2 analysis determined whether the proportion of patients with complications differed significantly between the patients who did and did not receive aspirin.
Third, we determined whether the use of aspirin was associated with better 30-day survival after adjustment for potential confounders. We developed a series of logistic regression models to predict 30-day mortality. The model began with aspirin only, and demographic variables, clinical variables, and cointerventions were added in sequential steps. For each model, we calculated an OR and 95% CI for those who received aspirin compared with those who did not receive aspirin. In addition, we report a χ2, probability value, and area under the receiver operating characteristic curve for each model. A survival analysis was performed after patients were stratified by Killip class and by whether they received aspirin. A log-rank test was performed to compare the aspirin and no-aspirin groups for each of the Killip subsets.
The analysis was repeated with a cohort of ideal candidates for aspirin. This smaller cohort of 7917 patients was developed by excluding patients with any possible contraindication to aspirin therapy, including history of a hemorrhagic stroke, active bleeding, history of peptic ulcer disease, history of gastrointestinal bleeding, suspected aortic dissection, allergy to aspirin, history of a bleeding disorder, platelet count <100 000, hematocrit <30%, prothrombin time >14 seconds, creatinine >3 mg/dL, or terminal illness with death expected within 6 months. All calculations were performed with the software system stata 3.0 (STATA Corp).
There are 16 189 patients in the Cooperative Cardiovascular Project Pilot database (Table 1⇓). After patients <65 years old (1074 patients), patients transferred from another institution or who were found not to have a principal diagnosis of ICD-9-CM 410 (2543 patients), those who died during the first day (237 patients), and those who did not have an AMI confirmed (1798 patients) were excluded, the remaining sample was 10 537. Among these patients, 387 were excluded because of an allergy to aspirin, and an additional 132 were excluded because of active bleeding. The remaining 10 018 were the study sample for the subsequent analyses.
Use of Aspirin
Among the 10 018 hospital admissions in the study sample, 6140 patients (61%) received aspirin within the first 2 days of presentation. The use of aspirin was also evaluated in a more restricted cohort of patients who were ideal candidates for the therapy by virtue of their complete lack of relative or absolute contraindications. In this group of 7917 patients, 5103 (64%) received aspirin within the first 2 days of presentation.
Table 2⇓ presents data from the overall study sample and compares the use of aspirin in patients with various characteristics. There were differences by demographic and clinical characteristics. Patients who were older, female, or nonwhite; had more comorbidity; and presented without chest pain and did not receive thrombolytic therapy, β-adrenergic blocking agents, and heparin on admission were significantly less likely to receive aspirin. Patients with higher-risk characteristics were much less likely to receive aspirin. This relation is illustrated by the association of Killip class and the use of aspirin (Fig 1⇓).
In the multivariate model, 19 variables were strongly associated with the use of aspirin (Table 3⇓). The variables that were associated with the use of aspirin included better overall health status (lower age, no history of stroke, no history of peptic ulcer disease, no evidence of terminal illness, and hematocrit >30%), presence of chest pain, and lower-risk characteristics at presentation (absence of congestive heart failure, absence of intubation on admission, absence of renal dysfunction, pulse rate <100 beats per minute, respiratory rate <30 breaths per minute, and highest systolic blood pressure >125 mm Hg).
Aspirin and Complications
The use of aspirin was not associated with an increased risk of complications. The rate of hemorrhage during the hospitalization was slightly lower among patients who received aspirin (139 of 6140, 2.3%) compared with those who did not receive aspirin (122 of 3878, 3.2%). The rate of transfusions was also lower in patients who received aspirin (471 of 6140, 7.7%, versus 446 of 3878, 11.5%). The incidence of severe renal failure in the cohort was very low; only 3 of 6140 patients who received aspirin and 5 of the 3878 patients who did not receive aspirin required dialysis. The number of strokes in the two groups was similar (103 of 6140 who received aspirin, 1.7%, versus 67 of 3878 who did not receive aspirin, 1.7%, within the first 2 hospital days).
Aspirin and Mortality
The 30-day mortality rate in the study sample was 18.0% (1893 of 10 018). Patients given aspirin within the first 2 days after admission had a 30-day mortality rate of 14.0% (860 of 6140) compared with 24.3% (943 of 3878) for patients not given aspirin (P<.0001). The length of hospital stay was 9.1±6.5 days for patients who received aspirin versus 10.2±7.7 days for patients who did not receive aspirin.
In each Killip class, patients who were given aspirin had a significantly lower 30-day mortality than patients who did not receive aspirin (Fig 2⇓). A series of multivariate logistic regression models was developed to test the association of use of aspirin and survival (Table 3⇑). In a model that adjusted for demographic and clinical variables, aspirin remained strongly associated with a lower 30-day mortality (OR 0.74; 95% CI, 0.66 to 0.83). After variables indicating the use of thrombolytic therapy, β-adrenergic blocking agents, and heparin were added, aspirin remained strongly and significantly associated with lower mortality (OR, 0.78; 95% CI, 0.70 to 0.89).
The analyses were repeated after patients who died on the first hospital day were included, and there was no substantial change in the results. In the final multivariable model predicting 30-day survival that adjusted for demographic, clinical, and therapeutic differences between the groups, the OR for aspirin was 0.73 (95% CI, 0.65 to 0.82). The analyses were also repeated with a restricted cohort of 7917 ideal candidates for aspirin therapy. In the final multivariable model, the results were virtually identical, with an OR for aspirin of 0.77 (95% CI, 0.67 to 0.89).
In this study, we demonstrate that about one third of elderly patients with AMI who are candidates for aspirin therapy do not receive it within the first 2 days of hospitalization. We also report that better outcomes were associated with the use of aspirin. The use of aspirin was associated with 22% lower odds of 30-day mortality, even after possible confounders were considered. These findings strongly reinforce the value of aspirin in treating AMI in the elderly.
Our data are consistent with the reports from selected patients suggesting that results from clinical trials do not always translate into clinical practice.4 5 6 One possible explanation is that, despite the findings of ISIS-2, some physicians are still uncertain about the efficacy of aspirin. A recently reported survey of physicians in New York and Texas found that only about half of internists and family practitioners and three quarters of cardiologists believed that aspirin definitely improves survival after MI in patients <75 years old.15 These numbers might have been much smaller if the physicians had been asked about patients ≥75 years old.
To understand better why physicians do not always prescribe aspirin for the treatment of AMI, we sought to characterize the patients who were the least likely to receive aspirin. Our analyses suggest that patients who did not present with classic signs and symptoms of AMI were less likely to be given aspirin. For example, elderly patients who presented to the hospital without chest pain received aspirin significantly less often than patients with chest pain. This observation is particularly important for older patients, since atypical presentations of AMI occur commonly in this group.20 21 For many of these patients, appropriate treatment with aspirin was not started in the hospital, possibly because of uncertainty about the diagnosis. Since atypical presentations of AMI are common in the elderly, a substantial proportion are at risk of not being treated with aspirin.
Another important group that did not receive aspirin were the most critically ill patients. Patients who presented with characteristics that were associated with a higher risk of mortality from AMI were much less likely to receive aspirin, and these factors were independent of whether the patient presented with chest pain. While the explanation for this observation is not obvious, it may be that physicians focus on high-technology interventions in high-risk patients and inadvertently neglect the simple use of aspirin. Also, patients who are intubated may not have aspirin ordered because nasogastric administration is not considered.
This information provides an excellent opportunity for improvement in our care of patients. Since aspirin tablets are inexpensive, efficacious for the treatment of AMI, and relatively safe, physicians should have a low threshold for administering aspirin to patients whose signs and symptoms are compatible with MI. Elderly patients with atypical presentations or with high-risk characteristics should be identified soon after their arrival to the hospital as candidates for aspirin therapy.
In addition to clinical characteristics, we also found some demographic characteristics associated with the use of aspirin. Older patients, female patients, and nonwhite patients were significantly less likely to receive aspirin, even after adjustment for clinical factors including the presence of chest pain and the acuteness of the presentation. Other studies have found similar results with respect to invasive cardiovascular procedures.4 12 22 23 24 Although it seems implausible that aspirin is being withheld from certain patient groups on the basis of their age, sex, and race, further investigation is required to determine the reasons for these findings.
The use of aspirin in our sample was strongly associated with other pharmacological interventions. Patients who received aspirin were significantly more likely to be treated with thrombolytic therapy, heparin, nitrates, and β-adrenergic blocking agents. These associations suggest that it is important to consider cointerventions in multivariable analyses of outcomes, since the use of aspirin may be a surrogate for an aggressive pharmacological management style or an earlier recognition of an MI.
Aspirin was not associated with complications in our sample. We compared the rate of potential complications in the two treatment groups and found no excess risk among those who received aspirin. On the basis of the results of ISIS-2,1 we did not expect an increased rate of bleeding or transfusions. Our observation that there were slightly fewer bleeding episodes in the aspirin group is consistent with our observation that patients with bleeding risks were less likely to receive aspirin. For this reason, we repeated our survival analysis after excluding patients with any risk of bleeding. The rate of stroke was almost identical in the two groups, as would be expected on the basis of the ISIS-2 results.
Since the value of aspirin for the treatment of AMI was demonstrated by ISIS-2, a large randomized trial, we expected that patients who received aspirin in our cohort would have better outcomes. To test this assumption, we undertook to evaluate the therapeutic efficacy of aspirin in our study sample. One important limitation of this analysis is the nonexperimental selection of treatment strategy. Since our analysis lacks the unique strength of randomization in allocating potential confounders, we adopted a methodology to minimize the problems inherent to drawing inferences from observational data.25 We defined a restricted cohort of patients who were free of contraindications to aspirin. We designated the first day of admission as the reference time at which baseline clinical status was characterized and from which the subsequent follow-up was counted. We excluded patients who died on the first hospital day to remove the possibility that some patients died before they had the opportunity to receive aspirin. Also, we collected the detailed clinical data available on admission so that we could adjust for differences between the treatment groups in their initial susceptibility to 30-day mortality.
In a series of multiple logistic regression models developed to adjust for these baseline differences in the groups, we showed that differences in baseline characteristics between the groups explained some, but not all, of the benefit associated with the use of aspirin. Aspirin was strongly associated with a favorable outcome even after demographic, clinical, and treatment characteristics were added to the multivariate models. Although it is possible that there were important unmeasured differences between the groups and despite important differences from ISIS-2 in the definition of the study sample, the 22% odds reduction associated with the use of aspirin in our study bears an encouraging resemblance to the results of the ISIS-2 trial.
Patients in every Killip class who received aspirin had a significantly better chance of survival than the patients who did not receive aspirin. Although highest-risk patients were the least likely to receive aspirin, they were just as likely to benefit from its use. In fact, the highest-risk patients had the greatest absolute difference in mortality between the aspirin and no-aspirin groups. Similarly, aspirin would be expected to provide more absolute benefit for older patients than younger patients.
This study suggests that thousands of lives a year could be saved by improvements in the process of care for patients with an AMI. More than 200 000 elderly Medicare beneficiaries are hospitalized annually with an AMI.8 Our data imply that as many as 70 000 are not receiving aspirin on presentation. According to our findings and the results of ISIS-2, aspirin for the treatment of 70 000 patients with an estimated mortality without aspirin of ≈20% could result in more than 3000 lives saved annually.
This investigation was a result of a Health Care Financing Administration–based initiative to improve health care for Medicare patients by providing data in an attempt to improve health care delivery. These results identify an excellent opportunity for improvement by increasing use of a simple, inexpensive, and yet highly effective intervention, ie, administration of aspirin to all elderly patients with an AMI. The data also show an association between good process of care (the prescribing of aspirin) with better outcomes (lower mortality). Our findings suggest that efforts to develop programs that will ensure that all eligible patients who are suspected of having an AMI receive aspirin on a timely basis may have the potential to save a large number of lives.
Selected Abbreviations and Acronyms
|AMI||=||acute myocardial infarction|
|ICD-9-CM 410||=||International Classification of Diseases, Ninth Revision, Clinical Modification|
|ISIS-2||=||Second International Study of Infarct Survival|
This study was supported in part by the Patrick and Catherine Weldon Donaghue Medical Research Foundation. We are indebted to the members of the Peer Review Organizations from Alabama, Connecticut, Iowa, and Wisconsin and to all the other individuals, hospitals, and organizations who contributed to the development and implementation of the Cooperative Cardiovascular Project. We are also indebted to Drs Ralph Horwitz, Viola Vaccarino, Carlos Mendes de Leon, Barry Zaret, John Finkle, and Richard Krumholz for their constructive comments regarding the manuscript.
Reprint requests to Harlan M. Krumholz, MD, Cardiovascular Section, Yale School of Medicine, 333 Cedar St, PO Box 208017, New Haven, CT 06520-8017.
1 This article does not necessarily represent the official position of the Health Care Financing Administration.
- Received April 26, 1995.
- Revision received June 23, 1995.
- Accepted June 25, 1995.
- Copyright © 1995 by American Heart Association
Rogers WJ, Bowlby LJ, Chandra NC, French WJ, Gore JM, Lambrew CT, Robinson RM, Tiefenbrunn AJ, Weaver WD, for the Participants in the National Registry of Myocardial Infarction. Treatment of myocardial infarction in the United States (1990 to 1993): observations from the National Registry of Myocardial Infarction. Circulation. 1994;90:2103-2114.
Gurwitz JH, Osganian V, Goldberg RJ, Chen Z, Gore JM, Alpert JS. Diagnostic testing in acute myocardial infarction: does patient age influence utilization patterns? The Worcester Heart Attack Study. Am J Epidemiol. 1991;134:948-957.
Krumholz HM, Douglas PS, Lauer MS, Pasternak RC. Selection of patients for coronary angiography and coronary revascularization early after myocardial infarction: is there evidence for a gender bias? Ann Intern Med. 1992;116:785-790.
International Classification of Diseases, Ninth Revision, Clinical Modification. Washington, DC: Public Health Service, US Department of Health and Human Services; 1988.
Steingart RM, Packer M, Hamm P, Coglianese ME, Gersh B, Geltman EM, Sollano J, Katz S, Moyé L, Basta LL, Lewis SJ, Gottlieb SS, Bernstein V, McEwan P, Jacobson K, Brown EJ, Kukin ML, Kantrowitz NE, Pfeffer MA. Sex differences in the management of coronary artery disease. N Engl J Med. 1991;325:226-230.