Cardiovascular disease (CVD) is the most common cause of mortality worldwide and substantially contribute to loss of health and excess health system costs [1][2]. Acute myocardial infarction (MI) is a common clinical manifestation of CVD. The mortality of patients with acute MI has been decreasing over the last 50 years [3], in part because of improvement in acute MI prevention, diagnosis, and treatment [4]. In-hospital mortality has decreased from 29% in 1969 [5] to <7% today [6][7]. However, around 9 million people worldwide die after acute MI each year (in 2019, 16% of total death in the world) 1, and in-hospital mortality varies substantially across hospitals [7]. Therefore, study of factors predicting the risk of death for a particular medical institution, adapted to its material and technical base and the existing flow of patients with acute MI, is an urgent problem.

Although several risk scores of in-hospital mortality were developed for patients with AMI, including the "Global Registry of Acute Coronary Events" (GRACE), "Thrombolysis in Myocardial Infarction" (TIMI), and "Can Rapid risk stratification of Unstable angina patients Suppress ADverse outcomes with Early implementation of the ACC/AHA Guidelines" (CRUSADE) scores [8-10], few included a representative sample from routine clinical care. Moreover, continued improvement in acute MI care requires periodic update to the risk models so that hospitals can assess their quality as contemporary care continues to evolve.

This study sought to determine factors associated with for in-hospital mortality in patients presenting with acute MI and develop a prognostic model for in-hospital mortality in a contemporary settings.

Study design and participants

A prospective, single-center observational study at Vinogradov Municipal Clinical Hospital (Moscow, Russia) was conducted. From January 2017 to December 2018, a total of 712 consecutive adults >18 years admitting with acute MI and undergoing coronary angiography (CAG) <24 hours after symptom onset were recruited. The exclusion criteria were type 3, 4, or type 5 MI. The diagnosis of acute MI was based on the Third universal definition of MI [11]. All patients provided written informed consent. The study was approved by the local Ethics Committee of RUDN University, and carried out in accordance with the Declaration of Helsinki.

Data collection and outcomes

Variables of interest included clinical characteristics, cardiovascular risk factors, comorbidities, physical examination findings, blood test results, and imaging data (electrocardiography, echocardiography, and CAG). Cardiac troponin I levels were measured using the Access 2 Immunoassay System (Beckman Coulter, USA), with 99th percentile upper reference limit being 0.02 ng/L. The GRACE 2.0 score was used to risk stratification for MI patients [12]. Anemia was defined as a hemoglobin concentration of less than 130 g/L for men or less than 120 g/L for women 2. The left ventricular ejection fraction (LVEF) was estimated by the modified biplane Simpson method [13] along with a transthoracic echocardiogram with using the Vivid 7 ultrasound system (General Electric Healthcare, USA).

The primary endpoint was all-cause death during the index hospitalization for acute MI, obtained from medical records.

Statistical analysis

Statistical analysis was performed using IBM SPSS Statistics 25.0 (SPSS Inc., Chicago, IL, USA) and R software (version 3.6.3). Descriptive statistics are presented as frequencies and percentage for categorical variables and as means ± standard deviation for continuous variables if variables adhering to a normal distribution or median (Me) and interquartile range (IQR) for skewed variables. Comparisons of categorical variables were performed by Chi-Square or Fisher’s exact tests, while for continuous variables by using the unpaired Student t-test and Mann-Whitney U test. Logistic regression analysis was used to identify factors associating with in-hospital mortality. Odds ratios (OR) and 95% confidence intervals (CI) were presented. The receiver operating characteristic (ROC) curve analysis was performed to determine cut-off level of continuous variables associated with in-hospital mortality. Based on the estimated factors from the multivariable logistic regression model, a nomogram was constructed to assess risk of in-hospital death ("rms" package in R). The discrimination accuracy of a nomogram was assessed the area under the ROC curve (AUC) [14]. The calibration was assessed by Hosmer-Lemeshow chi-square statistics for goodness of fit. A two-tailed p-value <0.05 were considered statistically significant.

Among patients, 434 (61%) were male with median age of 65 (IQR: 56-74) years, 47.8% presented with ST elevation. During the follow-up period, 36 patients (5.06%) died.

The baseline characteristics of the patients are shown in Table 1. Compared with survivors, age, proportion of female gender, ST-elevation, coronary artery disease (CAD), diabetes mellitus, atrial fibrillation, chronic kidney disease, anemia, dyspnea, Killip class II-IV, creatinine level, three-vessel CAD, GRACE score were higher in non-survivors. Hemoglobin, LVEF were lower in non-survivors. There were no significant differences in terms of arterial hypertension, previous MI, previous revascularization, previous heart failure (HF), previous cerebrovascular accident, peripheral artery disease, chronic lung disease (asthma and/or chronic obstructive pulmonary disease), peptic ulcer disease, troponin level, non-obstructive CAD, percutaneous coronary intervention (PCI) rate.

In the structure of in-hospital mortality, majority of patients died from acute HF (75%), followed by myocardial rupture with cardiac tamponade (11.1%) and multiple organ failure (8.3%). Other causes of in-hospital death are presented in Table 2.

In univariate analysis, age, female gender, ST-elevation, diabetes mellitus, atrial fibrillation, Killip class, hemoglobin at admission, estimated glomerular filtration rate (eGFR), LVEF and three-vessel CAD were associated with in-hospital mortality (all p-value < 0.05) (Table 3). The cut-off for LVEF in ROC curve analysis, was 36% (AUC = 0.692; 0.5548–0.836, p=0.004) for in-hospital mortality.

In multivariate analysis, age (OR 1.07, 95% CI 1.01-1.14, p=0.027), Killip class (OR 2.95, 95%CI 1.67-5.23, p<0.001), hemoglobin at admission (OR 0.97, 95% CI 0.95-0.99, p=0.006), and LVEF ≤36% (OR 8.87, 95% CI 2.95-26.69, p<0.001) were independently associated with the primary endpoint.

Based on the estimated variables in the multivariate model, a nomogram was developed to predict the risk of in-hospital death in patients with MI (Fig. 1). The risk of in-hospital death in MI patients was assessed according to the following equation:

Risk = 1/(1 + e-Z), where Z = -6.683 + 0.083 x Age — 0.028 x Hemoglobin at admission + 1.065 x Killip class + 2.377 x LVEF ≤36%.

A total score was generated by using the number of the points of each factor on the corresponding axis with drawing a vertical line to "Points" axis. Summary the points of all the factors and drawing a vertical line to the "Risk of in-hospital mortality" line to determine the individual’s probability of death during hospitalization.

In ROC curves analysis of a developed nomogram for predicting in-hospital mortality, the AUC was 0.949 (95% CI: 0.925-0.972, p<0.001). The sensitivity and specificity were 91.3% and 89.9%, respectively (Fig. 2). A nomogram showed good calibration (chi-square 3.064, p for Hosmer-Lemeshow test 0.93)

The present study demonstrated that age, hemoglobin at admission, Killip class and LVEF were independently associated with in-hospital death. Based on results from multivariate analysis, we developed a contemporary risk model to predict in-hospital mortality in patients with acute MI. The model had high predictive potential in predicting in-hospital death.

Age is one of the most important risk factors for a poor outcome in hospitalized patients with acute coronary syndrome (ACS) [15][16]. Median of age among non-survivors in our study was 77.5 years and was an independent factor for in-hospital dearth. Aging causes patients to be physically weaker and frailer. Therefore, old age is an independent predictor of poor prognosis in ACS patients [17]. Additionally, age serves as a prognostic variable in various risk stratification scores for ACS patients [8-10]. In the TIMI III registry, patients over the age of 75 with unstable angina (UA)/non-ST-segment elevation MI (NSTEMI) had more adverse outcomes both in-hospital and within the first six weeks after discharge than those less than 75 years of age [16]. Age and gender are closely associated with in-hospital mortality in acute MI with women being generally older than men at hospitalization for MI. In our study, the risk of in-hospital mortality for women was 2.3 times higher than for men.

The in-hospital mortality rate in our study was 5.06%, which is lower than the 8.8% reported in a recent meta-analysis involving 615,035 MI patients (comprising both ST-segment elevation MI (STEMI) and NSTEMI/UA) [18]. During hospitalization, acute HF emerged as the primary cause of death, accounting for three-fourths of the cases. These results are consistent with those of O. G. Sivkov, who examined 1120 patients with acute MI. The author reported an in-hospital mortality rate of 6.0% (n=67), with acute HF contributing to 70.1% of the deaths (32 cases with acute HF and 15 cases with cardiogenic shock), followed by multiple-organ failure at 28.3% [19].

Our multivariate analysis affirmed the continued relevance of Killip classification and LVEF for effective risk stratification in these patients. Killip classification categorizes patients based on physical examination findings that indicate left ventricular dysfunction and HF [20-22]. According to our results, an increase of in Killip degree was associated with an increased odds ratio of 2.95 for in-hospital mortality. These findings are in the line with an analysis of international data from the GRACE registry involving 3917 NSTEMI patients and 4960 UA patients, where Killip class II or III on admission independently predicted in-hospital death (OR, 2.2; p<0.0001) [23]. A broader analysis of data from 26,090 patients across the GUSTO IIb, PURSUIT, PARAGON A, and PARAGON B trials revealed a significant correlation between Killip class and survival [24]. Patients with Killip class II and III/IV had significantly higher mortality rates at 30 days (3% versus 9% and 14%) and six months (5% versus 15% and 23%) compared to patients with Killip class I. Killip class has proven to be a valuable predictor for adverse outcomes in acute MI patients, incorporated into several risk scores [8-10].

The baseline measurement of LVEF has been identified as the most influential determinant of both in-hospital [25][26] and long-term mortality [27, 28], making its inclusion imperative in risk models for maximal predictive accuracy. In a study by T. Sato et al. [26], involving 1102 acute MI patients undergoing primary PCI, multivariable analysis revealed that LVEF ≤33% independently predicted in-hospital death, with an AUC of 0.79 (p<0.001). Further, incorporation of LVEF into the TIMI score, as opposed to the TIMI risk score alone, resulted in significant enhancements in predicting in-hospital death (AUC: 0.854 vs. 0.803, p=0.033) in a cohort of 673,673 patients with STEMI [27]. Regarding long-term risk stratification, LVEF has proven to be a valuable predictor in various scores. For instance, LVEF was included in the CADILLAC score for predicting one-year mortality after primary PCI for acute AMI, with a cutoff value of LVEF <40% [29]. Consistent with these findings, H. K. Kim et al. developed the new KAMIR score from a cohort of 3,997 hospital-discharged patients with acute MI. The KAMIR score, incorporating six independent variables, including LVEF, demonstrated significant differences in predictive accuracy for one-year mortality compared to the GRACE score (AUC 0.83 vs. 0.76, p=0.0089, respectively). LVEF <40% in this study was associated with a hazard ratio (HR) of 2.24 (95% CI 1.47–3.41) for one-year mortality [30]. In our current study, Killip class at presentation, consistently featured in previous scores remained an independent predictor of reduced in-hospital survival, alongside baseline LVEF. This underscores the significance of clinical examination for signs of mild to moderate HF, even when left ventricular function is preserved.

Anemia is increasingly recognized as a condition strongly predictive of adverse outcomes in patients with acute MI, both in the short term [31-33] and long term [33-35]. In a study by J. J. González-Ferrer et al., involving 542 high-risk ACS patients, anemia was present in 147 patients (27.1%) at admission [32]. The study found that the admission hemoglobin level (OR=1.4 for each 1 g/dL below normal; 95% CI, 1.1-1.8; p=0.003) was independently associated with in-hospital all-cause mortality or cardiogenic shock, after adjusting for other variables. Another study by M. G. Colombo et al. [36] was performed in 2011 consecutively hospitalized acute MI patients. Mild anemia (defined as hemoglobin concentration of 11 to <12 g/dL in women and 11 to <13 g/dL in men) and moderate to severe anemia (defined as hemoglobin concentration of <11 g/dL) were found in 183 (9.1%) and 100 (5%) patients, respectively. The Cox regression analysis showed significantly increased mortality risks in both patients with mild anemia (HR 1.74, 95% CI 1.23–2.45) and moderate to severe anemia (HR 2.05, 95% CI 1.37–3.05) compared to patients without anemia. Anemia represents a potentially modifiable risk factor, and its incremental prognostic value demonstrated in the current analysis highlights the imperative for tailored therapeutic strategies aimed at mitigating its impact and improving overall prognostic outcomes in affected patients.

The study limitations

Our observational study has such limitations as non-randomization and unmeasured confounding factors. For instance, the exclusion of the variable "previous cerebrovascular accident" despite its known association with increased mortality risk in acute MI patients [37] may compromise the comprehensiveness of our findings. Being single centered with a small sample size is another limitation, though the study’s prospective nature adds strength. Thirdly, treatment decisions were left to attending cardiologists, making it difficult to assess the benefits of individualized therapy. At last, the lack of external validation for the nomogram in another acute MI patient cohort limits the generalizability of our findings and reduces the model’s clinical implementation. External validation is necessary to confirm the nomogram’s effectiveness in real-world scenarios.

Age, Killip class, hemoglobin at admission, and LVEF were identified as factors associated with higher in-hospital mortality in patients with acute MI. The most common etiology of in-hospital mortality was acute HF. The nomogram for predicting in-hospital mortality demonstrated high prognostic potential, allowing for the identification of patients at high-risk of adverse outcome, and targeted therapeutic strategies may be needed to improve the survival of patients with acute MI.

Relationships and Activities. None.

Funding. The study was performed with the support of Peoples’ Friendship University of Russia.

1 World Health Organization. The top 10 causes of death. World Health Organization, 9 December 2020. Available from: https://www.who.int/news-room/fact-sheets/detail/the-top-10-causes-of-death2 World Health Organization. Nutritional anaemias: report of a WHO scientific group [meeting held in Geneva from 13 to 17 March 1967]. World Heal Organ Tech Rep Ser. 1968;405:1-37. Available from: https://iris.who.int/handle/10665/40707.