There is increasing emphasis on optimizing evidence-based medication (EBM) persistence as a means to improve longitudinal patient outcomes after acute myocardial infarction (MI); yet it is unknown whether differences in medication persistence exist between patients discharged from academic versus nonacademic hospitals. We linked Medicare pharmacy claims data with 3,184 patients with non–ST-segment elevation MI >65 years of age who were treated in 2006 at 253 hospitals participating in the Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes with Early Implementation of the American College of Cardiology and American Heart Association guidelines registry. Using multivariate regression, we compared persistent filling of β blockers, angiotensin-converting enzyme inhibitors and/or angiotensin receptor blockers, clopidogrel, and statins at 90 days and 1 year postdischarge between patients discharged from academic and nonacademic hospitals. Patients treated at academic hospitals were more frequently nonwhite (19% vs 8%, p <0.001) and had a greater co-morbidity burden (Charlson score ≥4 in 36% vs 30%, p = 0.001) than patients treated at nonacademic hospitals. Composite persistence to all EBMs prescribed at discharge was low and not significantly different between academic and nonacademic hospitals at 90 days (46% vs 45%, adjusted incidence rate ratio = 0.99, 95% confidence interval 0.95 to 1.04) and at 1 year (39% vs 39%, adjusted incidence rate ratio = 1.02, 95% confidence interval 0.98 to 1.07). Rates of persistence to EBMs were similar between patients with MI >65 years old treated at academic versus nonacademic hospitals; however, persistence rates are low both early and late postdischarge, highlighting a continued need for quality improvement efforts to optimize post-MI management.
Several previous studies have observed that patients with myocardial infarction (MI) treated at academic hospitals are more likely to receive evidence-based medications (EBMs) in-hospital and at discharge compared with those treated at nonacademic hospitals ; yet whether a similar relation is seen for postdischarge persistence of EBMs between patients with MI cared for at academic and nonacademic hospitals remains unknown. In this study, we propose to compare the rates of EBM persistence between patients with MI treated at academic and nonacademic hospitals. We hypothesize that there will be a significant difference in persistence early after MI discharge at 90 days, but this difference will no longer be significant 1 year after the initial hospitalization. In secondary analyses, we will examine differences in length of initial hospitalization, in-hospital and predischarge treatments, and time to first postdischarge follow-up visit that may potentially explain persistence differences between patients treated at academic versus nonacademic hospitals.
Methods
The Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes with Early Implementation of the American College of Cardiology and American Heart Association guidelines (CRUSADE) registry was a voluntary quality improvement initiative designed to track guideline adherence, provide performance feedback, and develop tools to improve adherence to the American College of Cardiology and American Heart Association guidelines for patients with non–ST-segment elevation acute coronary syndrome. Inclusion and exclusion criteria and data collection processes have been described previously. Briefly, patients were included if they presented within 24 hours of anginal symptom onset lasting >10 minutes and had an electrocardiogram showing >1 mV of ST-segment depression or transient ST-segment elevation for <30 minutes, or elevated serum cardiac biomarkers. The institutional review board of each hospital approved participation in CRUSADE. All data were abstracted retrospectively and anonymously; therefore, informed consent was not required.
In 2006, Medicare implemented the Part D prescription drug benefit program. By linking the CRUSADE registry with Medicare Part D pharmacy data, we had the opportunity to study prescription medication filling patterns after hospital discharge for patients with non–ST-elevation MI (NSTEMI) >65 years of age. As data in CRUSADE were collected anonymously without direct patient identifiers, we performed a probabilistic linkage of patients included in CRUSADE with unique Medicare records using a combination of indirect identifiers (hospital, admission date, discharge date, age, and gender), as previously described. This probabilistic linkage resulted in the availability of linked medication data on 5,312 patients with NSTEMI >65 years who were admitted to CRUSADE hospitals from January 1, 2006 to December 31, 2006, and were enrolled in Part D within 1 year of discharge. We excluded patients who died during the index hospitalization (n = 259), patients who were transferred to another acute care hospital for whom we do not have information on their discharge medications (n = 1,597), and patients who were discharged on none of the indicated evidence-based therapies (n = 272). After exclusions, our final study population consisted of 3,184 patients with NSTEMI treated at 253 hospitals in the United States.
We identified academic hospitals by their membership in the Council of Teaching Hospital of the Association of American Medical Colleges as listed in the American Hospital Association Annual Survey database.
We examined medication persistence, defined as the proportion of patients still taking a medication prescribed at discharge, as well as at 90 days and 1 year postdischarge from the index MI hospitalization. Using Medicare Part D data, we determined whether the preceding date and quantity of prescription filling covered the time point of interest for each of the following EBMs: β blockers, clopidogrel, statins, and angiotensin-converting enzyme inhibitors and/or angiotensin receptor blockers. We did not examine aspirin use, as this was often purchased over the counter and not captured in pharmacy data. We summarized composite persistence as the number of EBM classes that the patient remains on at either the 90-day or 1-year time point divided by the total number of EBM classes that the patient was discharged on during the index MI hospitalization. For instance, if a patient filled a prescription for 2 medication classes, but was initially discharged on 3, then the overall persistence rate would be 66%. Substitutions of medications within a class did not affect persistence.
Baseline, in-hospital, and discharge characteristics were compared between patients treated at academic versus nonacademic hospitals. Categorical variables are presented as percentages and differences were assessed using the chi-square test when the sample size was sufficient; otherwise an exact test was used. Continuous variables are presented as medians (twenty-fifth and seventy-fifth percentiles) and were compared using the Wilcoxon rank-sum test.
Poisson regression with generalized estimating equations to account for within-hospital correlation was used to assess the association between persistence rates and academic versus nonacademic hospital status. The results were expressed as incident rate ratios with 95% confidence intervals (CIs). Variables for adjustment were adapted from the validated CRUSADE long-term mortality risk score model and included age, initial serum creatinine, initial systolic blood pressure, signs of heart failure at presentation, initial heart rate, weight, previous heart failure, initial hematocrit, initial troponin value, previous stroke, diabetes, gender, previous peripheral artery disease, and Charlson co-morbidity index >3. Categorical variables were imputed to the most frequent group. Systolic blood pressure, hematocrit, and heart rate were imputed to the median of the nonmissing values. Weight, creatinine, and troponin were imputed to the gender-specific median of the nonmissing values. Finally, we compared the risk of 1-year major adverse cardiovascular events (MACEs) between patients treated at academic versus nonacademic hospitals. MACE was defined as the composite risk of death or readmission for MI, stroke, heart failure, percutaneous coronary intervention, or coronary artery bypass graft surgery. For these outcomes, we performed Cox regression adjusting for the variables described previously.
To graphically display hospital variation in overall persistence rates, we estimated the distribution of hospital persistence rates separately for academic and nonacademic hospitals using a hierarchical model for persistence with hospital as a random effect. Including hospital as a random effect allowed us to remove the effect of random sampling variation due to small sample sizes per hospital. Persistence to medications out of total discharge medications was modeled using an unadjusted grouped logistic regression model. The log-odds for a random hospital is typically assumed to be normally distributed with mean equal to the intercept and variance equal to the random effect variance. We estimated these parameters and transformed from the log-odds scale to the probability scale.
Statistical significance was defined as p <0.05 with no correction for multiple comparisons. All statistical tests were 2-sided. SAS statistical software, version 9.2 (SAS Institute, Cary, North Carolina) was used for all calculations.
Results
The study population comprised of 253 CRUSADE hospitals in the United States; of these, 65 (26%) were academic hospitals and 185 (74%) were nonacademic hospitals. Compared with nonacademic hospitals, academic ones were larger (median number of beds 594 vs 325, p <0.001), not-for-profit (89.8% vs 87.2%, p = 0.03), have cardiac surgery capability (97.6% vs 83.2%, p <0.001), and have a cardiologist (as opposed to a noncardiologist like a hospitalist and intern) who primarily cares for patients with MI (71.2% vs 59.7%, p <0.001).
Of the total 3,184 patients with NSTEMI in the CRUSADE registry from January to December 2006, 1,107 patients (34.8%) were admitted to academic hospitals and 2,077 patients (65.2%) were admitted to nonacademic ones. The baseline characteristics of patients treated at academic versus nonacademic hospitals are listed in Table 1 .
| Variable | Overall (n = 3,184) | Academic (n = 1,107) | Nonacademic (n = 2,077) | p-Value |
|---|---|---|---|---|
| Age (years), median (IQR) | 76 (70, 82) | 76 (70, 82) | 76 (70, 82) | 0.90 |
| Women | 53.4 (%) | 53.2 (%) | 53.5 (%) | 0.88 |
| <0.0001 | ||||
| European American | 83.7 (%) | 77.4% | 87.0% | |
| African-American | 9.2% | 15.5% | 5.8% | |
| Asian | 1.1% | 1.3% | 1.0% | |
| Hispanic | 3.3% | 3.0% | 3.5% | |
| Other | 1.7% | 2.2% | 1.4% | |
| Prior myocardial infarction | 29.8% | 33.0% | 28.2% | 0.01 |
| Prior percutaneous coronary intervention | 25.0% | 27.0% | 24.0% | 0.14 |
| Prior coronary artery bypass graft surgery | 23.4% | 23.9% | 23.1% | 0.87 |
| Prior heart failure | 18.3% | 19.0% | 18.0% | 0.64 |
| Hypertension | 78.7% | 81.3% | 77.4% | 0.06 |
| Diabetes mellitus | 34.6% | 35.7% | 33.9% | 0.49 |
| Current/recent smoker | 15.4% | 15.5% | 15.3% | 0.96 |
| Dyslipidemia | 58.9% | 59.2% | 58.8% | 0.70 |
| Chronic obstructive pulmonary disease | 27.5% | 27.8% | 27.3% | 0.75 |
| Cancer | 4.2% | 3.9% | 4.3% | 0.59 |
| Dementia | 2.6% | 3.2% | 2.4% | 0.18 |
| Charlson index ≥4 | 31.9% | 35.7% | 29.9% | 0.008 |
| Presenting signs | ||||
| Signs of congestive heart failure | 24.4% | 24.9% | 24.2% | 0.83 |
| Systolic blood pressure (mm Hg), median (IQR) | 147 (127, 168) | 145 (127, 167) | 148 (128, 168) | 0.13 |
| Weight (kg), median (IQR) | 76.4 (65.0, 88.8) | 75.8 (65.0, 88.7) | 77.0 (64.9, 88.8) | 0.50 |
| Troponin (xULN), median (IQR) | 2.0 (0.5, 8.0) | 1.9 (0.4, 7.3) | 2.0 (0.5, 8.3) | 0.11 |
| Hematocrit (%), median (IQR) | 39.4 (35.5, 42.8) | 39.0 (35.1, 42.4) | 39.7 (35.9, 42.9) | 0.004 |
Academic and nonacademic hospitals had similar rates of diagnostic catheterization, revascularization, and occurrence of in-hospital complications, including cardiogenic shock, heart failure, stroke, and recurrent MI. The median length of stay for both academic and nonacademic hospitals was 4 days ( Table 2 ). There were no significant differences in weekend discharges between academic and nonacademic hospitals (19.7% vs 20.4%, p = 0.63). Rates of discharge EBM prescriptions were similar, with the exception of clopidogrel and statin prescription. Academic hospitals were significantly more likely to prescribe a statin on discharge, but less likely to prescribe clopidogrel. Nonacademic hospitals were more likely to offer cardiac rehabilitation referral and smoking cessation counseling before discharge (67.4% vs 55.6%, p <0.001 and 88.6% vs 84.5%, p <0.001, respectively).
| Variable | Overall (n = 3,184) | Academic (n = 1,107) | Nonacademic (n = 2,077) | p-Value |
|---|---|---|---|---|
| Median LOS (days), median (IQR) | 4 (3, 6) | 4 (3, 6) | 4 (3, 6) | 0.51 |
| Discharge prescriptions | ||||
| ACEI/ARBS | 70.4% | 70.4% | 70.4% | 0.74 |
| B-blockers | 94.9% | 95.1% | 94.8% | 0.81 |
| Statin | 85.1% | 87.2% | 84.0% | 0.02 |
| Clopidogrel | 75.9% | 72.2% | 77.9% | 0.0002 |
| All four agents | 52.0% | 51.9% | 52.1% | 0.92 |
| Discharge services | ||||
| Cardiac rehabilitation referral | 63.3% | 55.6% | 67.4% | <0.0001 |
| Smoking cessation counseling | 86.9% | 84.2% | 88.4% | 0.06 |
| Time to follow-up (days), median (IQR) | ||||
| Days to first follow-up visit with any specialty ∗ | 13 (7, 23) | 12 (7, 22) | 13 (7, 23) | 0.18 |
| Days to first follow-up visit with cardiologist | 28 (16, 55) | 28 (15, 53.5) | 28 (16, 56) | 0.25 |
| Number of cardiac follow-up visits within 90 days post-discharge | 1 (0, 2) | 1 (0, 2) | 1 (0, 1) | 0.19 |
| Number of any specialty follow-up visits within 90 days post-discharge | 3 (2, 5) | 3 (2, 5) | 3 (2, 5) | 0.20 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
