International Variation in and Factors Associated With Hospital Readmission After Myocardial Infarction FREE

Acute myocardial infarction with ST-segment elevation accounts for 29% to 38% of all acute myocardial infarction.¹ In the present era of primary percutaneous coronary intervention (PCI), survival to hospital discharge has improved dramatically.^2- 6 Subsequently, patients who survive to hospital discharge are at risk for early postdischarge hospital readmission.⁷ Early readmission rates have been proposed as quality measures among hospitals in the United States. In July 2009, the Centers for Medicare & Medicaid Services began reporting 30-day readmission for 3 common medical conditions, one of which was acute myocardial infarction.⁷ In particular, variation in 30-day readmission rates across centers has been identified as an indicator of inconsistent quality of care.⁸ Therefore, to improve quality and reduce costs, policy makers are increasingly focusing on 30-day readmission rates for acute myocardial infarction as both a quality and economic metric.⁹

To improve 30-day readmission rates associated with acute myocardial infarction, investigators have sought to identify clinical predictors of readmission. Numerous studies have found several predictors of readmission, and the models have reported modest discriminatory power,¹⁰ which may be in part due to the heterogeneous pathophysiology of acute myocardial infarction related to plaque rupture, supply-demand mismatch due to other acute conditions (ie, sepsis or hemorrhage), or cardiac procedures.¹¹ The strongest predictors of readmission for the spectrum of acute myocardial infarction in the Medicare population are older age, heart failure, and previous myocardial infarction.¹⁰ Predictors of 30-day readmission for the subset of acute myocardial infarction patients with ST-segment elevation in the primary PCI era have not been well studied. Additionally, little is known with regard to the potential variation and rates of 30-day readmission around the world.

We performed a post hoc analysis of the Assessment of Pexelizumab in Acute Myocardial Infarction study, a large multinational clinical trial that enabled us to determine predictors of 30-day postdischarge all-cause and nonelective readmissions among patients with ST-segment elevation myocardial infarction (STEMI), who were intended to undergo primary PCI. We hypothesized that among other factors, the country of enrollment would be independently associated with risk of 30-day readmission.

We performed a post hoc analysis of data from the Assessment of Pexelizumab in Acute Myocardial Infarction study. The study design and results have been previously published.^12- 13 Briefly, intravenous pexelizumab was compared with placebo administered in a randomized and blinded fashion immediately before primary PCI for electrocardiographically high-risk STEMI patients within 6 hours of symptom onset. High-risk STEMI patients were defined by any one of the following electrocardiogram findings: 2-mm or greater ST-segment elevation in 2 anterior or lateral leads; 2-mm or greater ST-segment elevation in 2 inferior leads, coupled with ST-segment depression in 2 contiguous anterior leads, for a total ST-segment deviation of 8 mm or greater; or new left bundle-branch block with at least 1 mm concordant ST-segment elevation. Patients were enrolled between July 13, 2004, and May 11, 2006, at 296 sites in 17 countries, with a final enrolled trial population of 5745 patients. The overall results of the Assessment of Pexelizumab in Acute Myocardial Infarction trial were neutral. The institutional review board at each participating hospital approved the protocol, and patients were required to provide written informed consent.

Data were collected on all patients with an electronic case report form. Standard baseline characteristics of patients enrolled, such as age, cardiac risk factors, time from symptom onset to presentation, infarct location, and Killip classification on presentation, were collected. Race and ethnicity were determined by site investigators during the conduct of the clinical trial as part of the Assessment of Pexelizumab in Acute Myocardial Infarction trial protocol. Detailed procedural findings, including time to PCI, transfer status (patients were transferred for primary PCI as part of routine local practice) culprit vessel, pre–and post–Thrombolysis in Myocardial Infarction flow, and concomitant therapies were also collected. Admission electrocardiograms and 30-minute post-PCI electrocardiograms were also evaluated at a core laboratory (Canadian VIGOUR Centre, Edmonton, Canada; Duke Clinical Research Institute, Durham, North Carolina) for degree of ST-segment deviation.

Data on site location, duration of index admission, and readmission at 30 days were collected via a structured case report form. The specific clinical reason for readmission was not ascertained during the conduct of the study. However, readmission for elective revascularization in the first 30 days was captured. A central adjudication committee independently adjudicated suspected myocardial infarction, heart failure, and cause of death.

The primary endpoint of our analysis was 30-day postdischarge all-cause hospital readmission, subsequently referred to as 30-day readmission. The secondary endpoint was 30-day postdischarge nonelective readmission, in which readmissions for elective PCI or coronary artery bypass grafting were excluded from the event definition (a total of 231 events).

Baseline characteristics and in-hospital complications were described by using medians with 25th and 75th percentiles for continuous variables and frequencies and proportions for categorical variables. Descriptive data were compared between outcome groupings with χ² tests for categorical variables and 2-tailed t tests for continuous variables. When specific distributional assumptions for these tests were violated, the 2-tailed Fisher exact and Wilcoxon signed rank tests were used for the categorical and continuous variables, respectively.

Multivariable logistic regression models were fitted to describe 30-day postdischarge readmission and 30-day nonelective readmission with stepwise selection methods (α = .05). In the latter scenario, we did not count readmissions for elective PCI or coronary artery bypass grafting as events. The list of candidate variables for both models was selected according to clinical importance and included baseline characteristics, as well as in-hospital complications. Model assumptions were assessed and variables were transformed as appropriate to satisfy linearity.

To ensure adequate control for case-mix variation across countries, additional variables were included, along with the selected covariates for each model: age, preintervention Thrombolysis in Myocardial Infarction score, preintervention ST-segment deviation, postintervention ST-segment resolution, peak creatine kinase mass, Killip classification, and patient-level length of stay (LOS), which was defined as a categorical variable with 6 groups (1-2, 3, 4, 5, 6, and >6 days). For each outcome, this constituted our final patient-level model.

Additional analyses were conducted to illustrate these models and address secondary questions. Odds ratios (ORs) for each country vs the United States were derived from these models by replacing United States vs non–United States with binary indicators for every country of enrollment (US reference). In addition, readmission rates, adjusted for all covariates in the patient-level model, were calculated for each country, and standard errors were obtained by bootstrapping. To address the hypothesis that differences between countries could be associated with differences in country-level LOS, we added country-level median LOS to the patient-level models. Differences in the timing of readmission between US and non-US sites were investigated by calculating readmission rates by time (days) since discharge in each region.

Although our primary hypotheses deal with nonfatal endpoints, mortality is a competing risk in this population. As a sensitivity analysis, we repeated the stepwise variable selection for the composite endpoint of 30-day postdischarge death or readmission. To further investigate the competing risk of death (and in particular how in-hospital mortality is not counted in an analysis of postdischarge rehospitalization), we fit logistic regression models of in-hospital and 30-day postadmission death, using covariates from a previously published model¹⁴ of death within 90 days of admission.

P = .05 was considered statistically significant. All statistical testing was 2-sided and performed with SAS version 8.2.

Of the 5745 STEMI patients enrolled in the trial, there were 5571 (97.0%) patients included in the analysis who survived to hospital discharge and represented 17 countries; 631 (11.3%; 95% CI, 10.5%-12.2%) were readmitted within 30 days from hospital discharge. Of the 631 patients, 14.5% (95% CI, 12.9%-16.2%) of patients in the United States and 9.9% (95% CI, 9.0%-10.9%) outside the United States were readmitted within 30 days. Excluding readmissions for elective revascularization, 478 (8.6%; 95% CI, 7.8%-9.3%) of patients in the overall cohort were readmitted within 30 days of discharge, accounting for 10.5% (95% CI, 9.0%-11.9%) of patients in the United States and 7.7% (95% CI, 6.9%-8.6%) of patients outside the United States.

Table 1 and Table 2 report baseline characteristics stratified by 30-day readmission status. Age and sex were similar between cohorts. Patients readmitted within 30 days had higher rates of comorbidities, including previous coronary artery disease (20.9% vs 15.5%; 95% CI, 17.7%-24.1% vs 14.4%-16.5%; P < .001), hypertension (56.7% vs 48.0%; 95% CI, 52.9%-60.6% vs 46.6%-49.4%; P < .001), and diabetes (18.7% vs 15.2%; 95% CI, 15.7%-21.7% vs 14.2%-16.2%; P = .02). Patients in the readmission cohort were more likely to have multivessel disease (57.1% vs 38.2%; 95% CI 53.2%-60.9% vs 36.8%-39.5%; P < .001) and noninferior myocardial infarction (63.9% vs 58.1%; 95% CI, 60.1%-67.6% vs 56.7%-59.4%; P = .005). More patients in the readmission group had patient-level LOS exceeding 6 days. Rates of discharge prescription of evidence-based medication, including aspirin, ticlopidine or clopidogrel, β-blocker, and statin, were similar between groups. A larger proportion of patients in the readmission cohort were from the US vs the no readmission cohort (39.1% vs 29.4%; 95% CI, 35.3%-43.0% vs 28.1%-30.7%; P < .001).

eTable 1 reports rates of in-hospital complications stratified by 30-day readmission status. Although overall rates of in-hospital complications were low, patients who were readmitted had higher rates of in-hospital complications, including congestive heart failure (4.3% vs 1.8%; 95% CI, 2.7%-5.9% vs 1.4%-2.1%; P < .001) and atrial fibrillation (7.0% vs 3.8%; 95% CI, 5.0%-9.0% vs 3.3%-4.3%; P < .001).

Patient characteristics stratified by US vs non-US patients are shown in Table 3. Compared with outside of the United States, the United States enrolled slightly younger patients (median age 58 vs 62 years; interquartile range 50-69 vs 54-71 years; P < .001) and a higher proportion of black patients (7.1% vs 0.2%; 95% CI, 5.9%-8.3% vs 0.1%-0.4%; P < .001). Comorbidities were similar, with the exception of a slightly higher prevalence of previous coronary artery bypass grafting surgery and coronary artery disease among US patients and slightly lower prevalence of chronic inflammatory conditions. Patient-level LOS was significantly lower for US patients vs patients outside of the United States (P < .001). Length of stay was 3 days or fewer for 60.0% (95% CI, 57.7%-62.4%) of US patients vs 15.9% (95% CI, 14.9%-17.0%) of patients outside of the United States (P < .001). Length of stay was 6 days or longer for 16.6% (95% CI, 14.8%-18.4%) of the US patients vs 54.0% (95% CI, 52.4%-55.6%) of those outside of the United States (P < .001). Patients in the United States had higher rates of readmission for elective PCI (4.4% vs 2.0%; 95% CI, 3.4%-5.3% vs 1.5%-2.4%; P < .001) and lower rates of readmission for elective coronary artery bypass grafting (0.6% vs 1.2%; 95% CI, 0.2%-1.0% vs 0.8%-1.5%; P = .046). Patients in the United States had higher rates of discharge with β-blocker and nitrate therapy, modestly reduced rates of discharge with ticlopidine or clopidogrel and statins, and reduced rates of discharge with angiotensin-converting enzyme inhibitors or aldosterone antagonists. Rates of aspirin prescription at discharge were similar between groups.

Predictors of 30-day readmission after adjustment for baseline characteristics are presented in Table 4. The 3 strongest predictors based on model χ² values were multivessel disease, US enrollment (vs rest of the world), and baseline heart rate. Patients with multivessel disease had almost twice the odds of readmission compared with those without (OR, 1.97; 95% CI, 1.65-2.35). Patients in the United States had 68% increased odds of readmission vs those outside the United States (OR, 1.68; 95% CI, 1.37-2.07), and baseline heart rate per 10/min increase was associated with a 9% increased odds of readmission (OR, 1.09; 95% CI, 1.05-1.15). Other independent predictors of readmission were noninferior myocardial infarction location (OR; 1.28; 95% CI, 1.07-1.52), recurrent ischemia (OR, 1.58; 95% CI, 1.01-2.47), chronic obstructive pulmonary disease (OR, 1.51; 95% CI, 1.07-2.14), chronic inflammatory conditions (OR, 1.75; 95% CI, 1.04-2.94), and a history of hypertension (OR, 1.28; 95% CI, 1.07-1.52). After adjustment for baseline characteristics, patient-level LOS was not a significant predictor of 30-day readmission (LOS >6 days vs 1-2 days: OR, 1.36; 95% CI, 0.97-1.91).

The odds of 30-day readmission for each country in the study (vs the United States) and adjusted readmission rates are displayed in the Figure. Several countries had significantly lower odds of 30-day readmission vs the United States, including Italy (OR, 0.26; 95% CI, 0.15-0.43), Germany (OR, 0.28; 95% CI, 0.07-0.46), Canada (OR, 0.33; 95% CI, 0.20-0.56), Portugal (OR, 0.41; 95% CI, 0.21-0.80), the Netherlands (OR, 0.50; 95% CI, 0.30-0.84), and the Czech Republic (OR, 0.59; 95% CI, 0.44-0.78). Only Denmark and Sweden had higher odds of readmission, which was not statistically significant.

Median LOS and adjusted 30-day readmission rate by country are reported in eFigure 1 and the Figure, respectively. Country LOS ranged from 3 days (interquartile range, 2-4 days) (United States) to 8 days (interquartile range, 6-11 days) (Germany). Risk-adjusted 30-day readmission rates varied by country, from 4.4% (95% CI, 2.3%-6.4%) (Italy) to 19.2% (95% CI, 12.9%-25.5%) (Denmark). After the addition of country-level median LOS to our readmission model (Table 4), US enrollment was no longer significantly associated with readmission (OR, 1.18; 95% CI, 0.87-1.61); however, country-level median LOS was associated with a 14% reduction in the odds of readmission for each additional in-hospital day (OR, 0.86; 95% CI, 0.78-0.96).

Baseline characteristics stratified by 30-day nonelective readmission are shown in eTable 2. Predictors of nonelective readmission are shown in Table 5. Independent predictors of 30-day nonelective readmission with largest χ² values were baseline heart rate (OR, 1.12; 95% CI, 1.07-1.18), age (for each year >55 years) (OR, 1.02; 95% CI, 1.01-1.03), and US enrollment (OR, 1.53; 95% CI, 1.20-1.96). When only nonelective readmissions were considered, multivessel disease was no longer associated with 30-day readmission.

After adjustment for country-level median LOS, the results of our model examining predictors of 30-day nonelective readmission are shown in Table 5. After adjustment for country-level LOS, US enrollment was no longer a significant predictor of 30-day readmission. Country-level median LOS was associated with a 17% reduction in the odds of 30-day nonelective readmission for each 1-day increase in LOS (OR, 0.83; 95% CI, 0.74-0.93).

To account for the competing risk of postdischarge 30-day mortality and its association with 30-day readmission, we also analyzed variables associated with the combined endpoint of 30-day death or readmission. There were 59 (1.1%) deaths within 30 days of hospital discharge. Overall, 674 (12.1%) patients died or were readmitted within 30 days of hospital discharge. After adjustment for baseline characteristics, multivessel disease and hospitalization in the United States remained the 2 leading predictors of 30-day death or readmission, according to χ² value (OR, 1.94 and 1.67; 95% CI, 1.63-2.31 and 1.36-2.05, respectively).

After adjustment for covariates, the adjusted odds of death in US vs non-US patients were not statistically significant for in-hospital death (OR, 0.88; 95% CI, 0.60-1.30) or 30-day postadmission death (OR, 1.0; 95% CI, 0.72-1.39).

The frequency of readmission according to postdischarge date (stratified by US or non-US enrollment) is illustrated in eFigure 2. In general, readmission frequency among all patients was greatest earlier postdischarge.

Given increasing focus on 30-day readmission rates among patients with common medical conditions, including myocardial infarction, we performed an analysis in a large international clinical trial of STEMI patients to determine predictors of all-cause and nonelective 30-day readmission. A substantial proportion of US patients (14.5%) were readmitted within 30 days of discharge, even though patients treated with primary PCI tended to be younger and generally have lower risk. In comparison, the reported median risk-standardized 30-day readmission rate after acute myocardial infarction (non-STEMI or STEMI) among elderly Medicare beneficiaries is 19.9%.⁷ After multivariable adjustment, multivessel disease and residing in the United States were the strongest predictors of 30-day all-cause readmission according to model χ² values. After exclusion of elective readmissions for PCI or coronary artery bypass grafting, multivessel disease was no longer significantly associated with readmission, but US enrollment remained predictive. Country-level median LOS was also an independent predictor of 30-day readmission rate. When country-level median LOS was added to a revised model of all-cause and nonelective readmissions, this variable eliminated the association between US enrollment and 30-day readmission. Patients residing in other countries vs the United States all had either decreased (6 countries) or statistically similar (10 countries) odds of all-cause readmission, as did patients residing in the United States. Further analysis to investigate whether our results were, in part, due to differential inpatient and short-term mortality between US and non-US patients revealed that there was no difference between groups in the odds of inpatient or 30-day mortality. These findings reinforce that regional differences in readmission rates observed in our study do not seem to be influenced by the competing risk of short-term mortality.

Although the pathophysiology of non-STEMI can be heterogeneous,¹¹ STEMI is nearly always related to plaque rupture, leading to an occlusive thrombus in an epicardial coronary vessel.^16- 17 Hence, isolating our analysis to the STEMI subset better controls for variation in acute pathogenesis, allowing us to focus on other patient factors and care patterns that increase risk for readmission. Our findings indicate that, in addition to patient comorbidities, disease severity (as indicated by the presence of multivessel disease), baseline heart rate, and high-risk electrocardiogram changes are important predictors of subsequent short-term readmission. There was a 68% increase in the risk of readmission among patients residing in the United States vs patients outside the United States in our study. Thirty-day postdischarge readmission rates in the United States were lower in this study than in previously reported Medicare data.^7,18 This discrepancy likely reflects the relatively young clinical trial population with less comorbidity, as well as the highly effective treatment our patients received with primary PCI.¹⁹

As one might predict, when we investigate only readmissions unrelated to elective PCI or coronary artery bypass grafting, multivessel disease is no longer associated with readmission. Being hospitalized inside the United States remains associated with readmission, but the association is diminished, which suggests that differential rates in early readmission for elective revascularization may be associated with the higher rates of all-cause readmission observed in the United States. Curtis et al¹⁸ examined readmission rates among Medicare beneficiaries post-PCI and found that almost one-third of those who were readmitted within 30 days underwent revascularization—and primarily PCI. Although their data could not distinguish between elective and urgent revascularization, our data suggest that a significant proportion of patients in the United States who are readmitted within 30 days after STEMI undergo elective PCI or, less commonly, coronary artery bypass grafting. Moreover, the rates of readmission for elective revascularization after STEMI are higher in the United States than in other countries studied. These findings highlight the variation and potential clinical uncertainty regarding the management of non–infarct-related coronary artery disease, which is the focus of several ongoing clinical studies.

Perhaps the most intriguing finding was the relationship between readmission rates and LOS. Once we adjusted our models of all-cause and nonelective readmissions for country-level median LOS, then the United States was no longer significantly associated with readmission. In general, one notable difference in care patterns outside the United States is longer LOS for cardiovascular disease and post–myocardial infarction,²⁰ as in our study. Length of stay post-STEMI has declined dramatically in the United States during the past quarter century.^6,21 In our study, the United States had the lowest median LOS among all countries, results similar to those of previous analyses of clinical trial populations.²⁰ Some are now suggesting that LOS has declined too far in the United States, resulting in suboptimal outcomes.²¹ This trend may be driven by a health care system that financially rewards early discharge. However, studies have not been able to show an association between shorter LOS and increased risk of readmission for acute myocardial infarction.²¹ Yet we did find an inverse relationship between country-level LOS and 30-day postdischarge readmission rates. This variable offset the strength of being a US vs non-US patient as a predictor of readmission. These data raise the possibility that higher readmission rates in the United States may be an adverse effect of the short LOS practice. A consequence is that patients become eligible for readmission closer to the index event, when risk of subsequent events is still elevated. Nevertheless, because this analysis is exploratory, country-level LOS may be a marker for a combination of differences in health care patterns across different countries, even though we attempted to adjust for some of the factors.

Other system-related differences outside the United States often include near-universal health coverage for all citizens through a single-payer state-run system, private compulsory health insurance, or some combination of the two,²² which may afford quicker and easier access to primary care postdischarge follow-up, as well as more accessibility to medications such as clopidogrel. Recent US data have shown decreased risk of 30-day readmission among heart failure patients who follow up with a physician within 7 days of discharge.²³ Previous research in the United States has also shown decreased risk of readmission among patients with early follow-up after acute myocardial infarction.²⁴

Significant attention has been focused on reducing acute myocardial infarction readmission rates in the United States as a means of reducing health care costs, according to the assumption that readmission is (at least in part) preventable. Our analysis shows that readmission may be preventable because rates are nearly one-third lower in other countries, suggesting that the US health care system has features that can be modified to decrease readmission rates. Understanding these international differences may provide important insight into reducing such rates, particularly in the United States. At the same time, the economic tradeoff between prolonging index hospital stays and reducing readmissions needs further research because the former may simply decrease overall efficiency without a significant effect on outcomes or overall resource use. Comparative analysis of health care systems may allow discovery of the optimal in-hospital processes that minimize readmissions without excessively increasing costs. Examining the total inpatient days for a 30-day episode of care may be a useful construct for this economic analysis; however, those data were not available for the present analysis. Patient-level factors associated with decreased risk for early discharge post–myocardial infarction have been studied and could be explored to reduce LOS without risking increased readmissions.^19,25

Our study has several limitations. First, it is a retrospective analysis of clinical trial data. Although we have used a multivariable model to adjust for potential confounders, there may remain unmeasured or residual confounding. Second, although the Assessment of Pexelizumab in Acute Myocardial Infarction trial had few exclusion criteria (including, for example, patients of any age, as well as those with renal failure and cardiogenic shock), the population was a selected clinical trial population; therefore, our results may not be generalizable to the STEMI population at large. Third, some countries enrolled relatively few patients, limiting the statistical power to make comparisons vs the United States. The analysis of median LOS was limited by the number of countries evaluated and included only as a linear predictor. This relationship requires further consideration in data sets with additional health care systems. Finally, our study compares only a small proportion of all countries globally, and this limits a true global interpretation of our results.

Thirty-day readmission after STEMI occurs in greater than 1 in 10 optimally treated patients. Rates are higher in patients with multivessel disease and in patients treated in the United States. After multivariable adjustment, the United States continued to be an independent predictor of a higher likelihood of 30-day all-cause hospital readmission. This association may be related to differential rates of early readmission for elective revascularization and shorter median LOS in the United States. In particular, country-level median LOS attenuates the relationship between the United States and early readmission. Further research is needed to better understand the relationship between LOS and readmission rates and define and optimize overall efficiency of care internationally.