This was a retrospective study in which demographic, clinical, and angiographic data, as well as data on management and in-hospital complications, had been prospectively collected and recorded in an electronic database. Subjects were all patients with a diagnosis of ACS admitted consecutively to our hospital between January 2004 and June 2010. The initial cohort consisted of 4645 patients, of whom 274 died during the in-hospital phase. Of the 4371 discharged patients, those in whom ACS was triggered in the context of surgery, sepsis, trauma, or cocaine consumption (n=41), and those missing data for any variable of the GRACE risk score (n=67), were excluded. Of the remaining 4263 patients, one-month follow-up was completed in 99.2% (34 patients without follow-up data). Thus, the final cohort was composed of 4229 patients. The study complies with the Declaration of Helsinki, and was approved by the Clinical Research Ethics Committee of our hospital.

The diagnosis of ACS was validated, through retrospective chart review, if the patient had new onset symptoms suggestive of myocardial ischemia and any of the following criteria: cardiac biomarkers above the upper limit of normal, ST-segment deviation on electrocardiogram, in-hospital stress testing showing ischemia, or known history of coronary artery disease. Patients were classified as having ST-elevation myocardial infarction (STEMI) or non-ST elevation ACS (NSTE-ACS) (non-ST elevation myocardial infarction or unstable angina). The diagnosis of unstable angina was based on suggestive symptoms together with objective evidence of myocardial ischemia on stress testing or detection of a ≥50% culprit lesion on coronary angiography, in addition to cardiac biomarkers below the upper normal limit. Left ventricular ejection fraction (LVEF) was determined by two-dimensional echocardiography. Depressed LVEF was defined as values ≤40%. In accordance with the World Health Organization criteria, anemia was defined as hemoglobin concentration below normal (<13 g/dl in men and <12 g/dl in women). Occurrence and severity of in-hospital bleeding were recorded using the Thrombolysis In Myocardial Infarction (TIMI) scheme.4 Major bleeding was defined as intracerebral hemorrhage or clinically overt bleeding associated with a drop in hemoglobin of ≥5 g/dl, while minor bleeding was defined as a drop in hemoglobin of 3–5 g/dl. In our study, severe bleeding was defined as TIMI major or minor bleeding unrelated to coronary artery bypass grafting.

The study endpoint was the combination of 30-day post-discharge mortality and readmission due to reinfarction, heart failure or stroke. All patients were followed for 30 days or until any event involving the combined endpoint. Follow-up methods involved one or more of the following: use of hospital records, hospital visits, telephone call to the patient's general physician, and telephone call to the patient.

The two versions of the GRACE risk score (in-hospital and six-month) were calculated for each patient from the sum of the individual scores assigned to each of the corresponding variables, as previously described5,6; thus, the sum of these scores corresponded to the total GRACE risk score as a continuous variable. In addition, patients were categorized in different risk groups according to cutoff points and intervals established by the GRACE risk score. Accordingly, three risk categories (low, intermediate and high) were established.

All analyses were performed using SPSS (version 17.0, SPSS Inc., Chicago, IL). Continuous variables were described as mean ± standard deviation (SD) or as median and interquartile range. The Student's t test or Mann–Whitney U test, as appropriate, were used for comparisons of continuous variables between two groups of patients. Discrete variables were expressed as frequencies and percentages, and were compared with the chi-square test. A parsimonious logistic regression model was used to estimate the odds ratios (OR) and 95% confidence intervals (CI), assessing the performance of the GRACE risk score to predict 30-day events in a multivariate model. In the initial model variables that resulted in significant predictors of 30-day events in the univariate model were included. Multicollinearity was assessed by examining pairwise correlations between all continuous predictor variables and by assessing the variance inflation factor for each predictor variable. The contribution of each significant predictor in the multivariate model was ranked by its F-value, and variables with the smallest contribution to the model were sequentially eliminated. The final model was composed of 11 variables: age, diabetes, LVEF ≤40%, grade 3–4 mitral regurgitation, in-hospital infection, anemia at discharge, absence of coronary stenosis, percutaneous coronary intervention (PCI) with drug-eluting stents, beta-blockers at discharge, statins at discharge, and six-month GRACE risk score. A p value <0.05 was considered statistically significant.

Measures of fit and discrimination were used to assess performance of the GRACE risk score, including the Hosmer–Lemeshow (HL) goodness-of-fit test,7 in which higher p values indicate better fit. The GRACE risk score was entered into a logistic regression model to generate the individual probability of death. The HL statistic from the regression modeling was used as an indicator of the goodness-of-fit of the GRACE risk score as an overall predictor variable. Discriminatory power was assessed by the C-statistic, equivalent to the area under the receiver-operating characteristic curve.8 Negative and positive predictive values for the GRACE risk score were also computed for the high-risk group.

Demographic, clinical and procedural characteristics are shown in Table 1, comparing groups of patients with and without 30-day events. Overall, 2.7% of patients presented events within 30 days of discharge: 30 died (0.7%), 42 had reinfarction (1.0%), 57 were admitted for heart failure (1.3%), and nine had a stroke (0.2%).

In the present cohort, the mean six-month GRACE score was 112.9±33.4. Using the mortality risk stratification proposed for ACS, 29.4% (n=1245) of patients were classified as low risk, 32.7% (n=1383) as intermediate risk, and 37.9% (n=1601) as high risk. The GRACE score stratification (low, intermediate and high) was correlated with a progressive increase in 30-day mortality, reinfarction, heart failure and stroke (Figure 1). The sensitivity and specificity of the high risk GRACE classification for 30-day events were 78.1% (95% CI 69.2%–85.1%) and 63.3% (61.8%–64.7%), respectively. Although the positive predictive value was low (5.6%, 95% CI 4.5%–6.8%), with a high false positive rate, the negative predictive value was extremely high (99.1%, 95% CI 98.6%–99.4%). As a continuous variable, a higher six-month GRACE score was found in the group that reached the study endpoint of events within 30 days of discharge (148.4±30.4 vs. 111.9±33.0, p<0.001), and in the subgroups of 30-day mortality (161.1±26.2 vs. 112.6±33.2, p<0.001), reinfarction (141.7±30.8 vs. 112.6±33.3, p<0.001), heart failure (153.6±27.6 vs. 112.4±33.2, p<0.001), and stroke (144.8±39.8 vs. 112.9±33.4; p=0.004) (Table 2). Assessment of the fit and overall performance of the GRACE score demonstrated a good fit (p=0.83 for the HL goodness-of-fit test) and high discriminatory power (C-statistic: 0.79±0.02) for the combined endpoint and for each event separately (Table 3). The fit and discrimination of the GRACE risk score were also good in both the PCI and non-PCI groups, with HL p values of 0.849 and 0.765, and C-statistics of 0.78±0.03 and 0.80±0.03, respectively. In Figure 2 a continuous model of the interaction between GRACE risk score and probability of 30-day events is shown.

Figure 1.

Risk stratification by GRACE score risk categories (low, intermediate, high) for 30-day events (death, reinfarction, heart failure, stroke, and the combination).

(0.36MB).

Figure 2.

Progressive increase in the probability of 30-day events (death, reinfarction, heart failure or stroke) with six-month GRACE risk score.

(0.13MB).

After multivariate analysis, the six-month GRACE risk score was shown to be an independent predictor of early events within 30 days of ACS (OR as a continuous variable 1.02, 95% CI 1.01–1.03; p<0.001), with a three-fold increased risk of 30-day events in the high risk GRACE group (Figure 3).

Figure 3.

Forest plot showing multivariate predictors of 30-day events (death, reinfarction, heart failure or stroke). Adjusted odds ratios, 95% confidence intervals (CI), and p values for each variable derived from multivariate logistic regression analyses are shown. LVEF: left ventricular ejection fraction.

(0.32MB).

The use of the six-month GRACE risk score is crucial in the context of increasing interest in using hospital readmission as a quality metric, linked to correct clinical practice. Although the GRACE score can categorize ACS patients into predicted risk groups for early mortality and readmission after discharge, it does not distinguish between preventable or unpreventable events. However, it can identify high-risk patients for whom there is the greatest potential for preventing early events. Future efforts should be devoted to developing methods for specifically identifying those cardiovascular events that could have been prevented through improved quality of care.24

These data should be interpreted in the context of this study's limitations. First, it is a retrospective analysis of clinical data from a single center. Although a multivariate model was used to adjust for potential confounders, unmeasured or residual confounding may remain. Second, we have no data regarding medication compliance or socioeconomic and educational variables, which can affect the occurrence of early events. Third, the 30-day post-discharge mortality and cardiovascular readmission rates were very low, which limits the analysis of each of the endpoints (mortality, reinfarction, heart failure, and stroke) separately.

The authors declare that the procedures followed were in accordance with the regulations of the relevant clinical research ethics committee and with those of the Code of Ethics of the World Medical Association (Declaration of Helsinki).

The authors declare that they have followed the protocols of their work center on the publication of patient data.

The authors have obtained the written informed consent of the patients or subjects mentioned in the article. The corresponding author is in possession of this document.