IJCS | Volume 33, Nº4, July and August 2020

382 depression (use of medicine or not) were adjusted for the statistical model. 3 For patients’ stratification according to the Killip classification, the following categories were used: I – normal, II – heart failure, III - acute lung edema, IV – cardiogenic shock. 14 The TIMI Risk score to non- STEMI is based on 7 variables – age ≥ 65 years, ≥ 3 risk factors for coronary artery disease, previous cardiac catheterization (stenosis > 50%), electrocardiography (ST-segment depression ≥ 0.5 mm), anginal symptoms, use of acetylsalicylic acid (ASA) in the last 7 days, and elevated troponin levels. 15 Complete blood count parameters including NRBCs, leukocytes, neutrophils, lymphocyte, platelet, MPV were measured using a Sysmex XE-2100 blood analyzer (Sysmex, Kobe, Japan). A positive NRBC was defined as any value above zero; cut-off level for high MPV was ≥ 10.4 fL, andNRLwas calculate by dividing the neutrophil count by the lymphocyte count, with a high cut-off level of ≥ 3.7, as previously described. 5-7 Blood samples were collected between 24 and 48 hours after admission. Statistical analysis The scoring system was developed in three steps (Figure 1). Multiple linear and multivariate logistic analysis of hematological variables and cardiovascular risk factors were used to identify independent predictors of mortality. In the first step, themagnitude of association of clinical and laboratory parameters with in-hospital mortality were measured by odds ratio (OR), whose statistical significance was estimated by likelihood ratio (Pearson chi-squared test) and represented by p-value. In the second step, the multivariate logistic regression model was conducted with all variables with a p value < 0.05 and the outcomes remained in themodel. In the third step, another adjusted logistic regression was calculated, with hematological parameters with p < 0.05 and a coefficient β (strength of association between variables). A score was attributed to each variable, which was the coefficient β of each variablemultiplied by 10, for the sake of rounding off. To analyze the accuracy of the scoring system, a receiver operating characteristic (ROC) curve was constructed, and sensitivity, specificity, positive and negative predictive values, positive (LR+) and negative (LR-) likelihood ratios, with their respective confidence intervals, were calculated. Continuous variables were expressed as mean ± standard deviation (normal distribution) or median (without normal distribution) and categorical variables were expressed as absolute or percent values, as appropriate. The association of higher levels of NRBC, MPV andNLRwith clinical and laboratory characteristics of the patients were assessed using Pearson Chi-squared test or Mann-Whitney U test. Regression analysis was performed for the variables identified as statistically significant in univariate analysis. The abilities of NRBC, MPV andNLR to distinguish patients withAMI from low or high risk of in-hospital death were evaluated using ROC curve analysis. The overall agreement between the hematological scoring system and Killip / TIMI Risk scores was assessed using Kappa coefficient. Statistical analyses were conducted using the Statistical Program for Social Sciences (SPSS), version 10.0 for Windows. Results A total of 466 patients (mean age 64.2 ± 12.8 years, 61.6%male) were included in this study. Total mortality was 11.8% (55 patients): 43/326 (13.2%) STEMI and 12/140 (8.6%) non-STEMI. Clinical characteristics related to in-hospital mortality among patients with AMI are described in Table 1. The presence of NRBCs in the sample was detected in 9.1% (42 patients), 27 (5.8%) with levels > 200/μL. Mean MPV value was 10.9 ± 0.9 fL and the meanNLR value was 3.71 (2.38; 5.72). The association of in-hospital mortality with the presence of NRBCs and increases in MPV and NLR in peripheral blood is shown in Table 1. We used the univariate model to assess which clinical and laboratory factors were associatedwith in-hospital mortality among patients with AMI (Table 1). To identify independent predictor variables associated with in-hospital mortality, we performed a multivariate analysis model (Table 2). After adjustment, the points assigned to each hematological variable of the scoring system proposed and respective coefficients β are detailed in Table 3. The hematological scoring system proposed had a scale ranging from 0 to 49, where higher values were associated with higher risk of in-hospital death. The better performance was registered for a cut- off value of 26 with sensitivity of 89.1% and specificity of 67.2%, positive predictive value of 26.8% (95% CI: 0.204 – 0.332) and negative predictive value of 97.9% (95% CI: 0.962 – 0.996) (Table 4). The area under the curve for the scoring system was 0.868 (95% CI: 0.818 – 0.918) (Figure 2). A score ≥ 26 points in the scoring system proposed showed an agreement of 82.1% with Monteiro Júnior et al. Hematological scoring system in AMI Int J Cardiovasc Sci. 2020; 33(4):380-388 Original Article