Abstract
Aims
The Killip classification is a long-established bedside tool for early haemodynamic risk stratification in ST-elevation myocardial infarction (STEMI). However, its prognostic performance in contemporary STEMI populations treated with primary percutaneous coronary intervention (PCI) remains debated. We aimed to re-evaluate the association between Killip class and in-hospital mortality in a modern STEMI cohort.
Methods and results
We conducted a retrospective cohort study including 288 consecutive adults admitted with confirmed STEMI to the Hospital of the Lithuanian University of Health Sciences, Kaunas Clinics, between 1 January 2018 and 31 December 2021. STEMI was diagnosed according to the Fourth Universal Definition of Myocardial Infarction and ESC guidelines. The primary endpoint was in-hospital all-cause mortality. Independent predictors were identified using multivariable logistic regression. Model discrimination was assessed using receiver operating characteristic (ROC) analysis. Overall, in-hospital mortality was 18.2% (52/286 evaluable patients). Mortality increased substantially across Killip classes, from 1.0% in Class I and 3.0% in Class II to 69.6% in Class IV (P < 0.001). In multivariable analysis, Killip Class IV remained an independent predictor of in-hospital mortality (OR 60.94, 95%: CI 15.98–232.46; P < 0.001), together with age, body mass index, troponin level, and asystole. The final model demonstrated excellent discrimination (AUC 0.969, 95% CI: 0.945–0.992).
Conclusion
In this temporary STEMI cohort, Killip Class IV was strongly and independently associated with in-hospital mortality. Although lower Killip classes showed limited prognostic separation, the Killip classification remains a rapid and clinically accessible tool for early risk assessment, particularly in haemodynamically unstable patients.
Keywords: Killip classification, ST-elevation myocardial infarction, Prognostic assessment, In-hospital mortality, Cardiac biomarkers, Ejection fraction
Introduction
Cardiovascular diseases (CVD) remain the leading cause of death worldwide, accounting for an estimated 20.5 million deaths in 2021 and nearly one-third of all global mortality.1,2 Although age-standardized CVD mortality has declined over recent decades due to advances in prevention and treatment, the absolute burden continues to rise, driven largely by population ageing and persistent inequalities in access to timely and effective cardiovascular care. These disparities particularly pronounced in low- and middle-income regions, where more than 80% of CVD-related deaths occur.1,2
Among available prognostic tools, the Killip classification, introduced in 1967, remains one of the simplest and most widely used bedside systems for assessing heart failure severity in patients with AMI. This four-level clinical classification, ranging from no signs of heart failure (Class I) to cardiogenic shock (CS) (Class IV), has been consistently associated with short- and long-term mortality. Several studies have confirmed its independent prognostic value. Cox et al.3 demonstrated that Killip class stratifies long-term mortality beyond the acute phase, while the large registry-based SAIKUMA study showed that Killip Class IV at presentation was associated with an almost 16-fold increase in-hospital-related mortality, despite widespread use of primary percutaneous coronary intervention (PCI) and contemporary stenting techniques.4
Despite the development of more complex risk models—such as GRACE and TIMI—that incorporate biochemical, electrocardiographic, and imaging parameters, the Killip classification remains embedded in contemporary clinical practice and guideline recommendations because of its simplicity, rapid bedside applicability, and validated prognostic strength. Indeed, Killip class continues to be integrated into the GRACE score for grading heart failure severity, underscoring its ongoing clinical relevance.5 Recent studies have further confirmed its value across multiple AMI subtypes. Killip Class ≥II independently predicts adverse outcomes in ST-elevation myocardial infarction (STEMI), even in the era of routine primary PCI,6 and has demonstrated prognostic validity in non-obstructive AMI (MINOCA), where risk stratification options remain limited.7,8
In addition to mortality prediction, higher Killip classes have been associated with reduced left ventricular ejection fraction (LVEF), atrial fibrillation (AF), conduction abnormalities, and elevated biomarkers of myocardial injury.6,9,10 Analyses from large registries such as FAST-MI and CORONOR have shown that diabetes mellitus significantly increases the likelihood of higher Killip class and the risk of heart failure following AMI.9 Importantly, projections suggest a 70% increase in cardiovascular mortality by 2050, with the greatest relative rise expected in Central and Eastern Europe.11 Against this background, cost-effective, rapid, and scalable prognostic tools remain essential for managing patients within increasingly strained healthcare systems.
Despite extensive historical validation, the prognostic performance of the Killip classification has been re-evaluated only infrequently in contemporary real-world cohorts, particularly within underrepresented Eastern European populations. Differences in pre-hospital logistics, comorbidity burden, and access to advanced circulatory support may influence the clinical presentation and outcomes of high-risk patients, potentially modifying the performance of traditional risk stratification tools. Therefore, the aim of this study was to re-evaluate the clinical and prognostic utility of the Killip classification in a contemporary cohort of patients admitted with STEMI to a tertiary care centre in Lithuania. We examined associations between Killip class and in-hospital mortality, cardiac function, biomarker levels, infarct localization, conduction abnormalities, and demographic characteristics. In addition, we assessed the independent predictive value and discriminative performance of the Killip classification within a multivariable logistic regression model. By focusing on a real-world STEMI population with a notably high prevalence of Killip Class IV and an attenuated mortality gradient across lower Killip classes, this study provides updated, region-specific insights into the strengths and limitations of the Killip system in contemporary acute myocardial infarction management.
Methods
Study design and population
This retrospective observational cohort study evaluated the prognostic value of the Killip classification for predicting in-hospital mortality in patients with ST-elevation myocardial infarction (STEMI). Consecutive adults (≥18 years) admitted with STEMI to the Cardiology Department of the Hospital of Lithuanian University of Health Sciences (Kaunas, Lithuania) between 1 January 2018 and 31 December 2021 were screened for eligibility.
STEMI was diagnosed according to the Fourth Universal Definition of Myocardial Infarction and current European Society of Cardiology (ESC) STEMI guidelines. Diagnosis required persistent ST-segment elevation of ≥1 mm in at least two contiguous leads or new left bundle branch block consistent with myocardial ischaemia, accompanied by a rise and/or fall in cardiac troponin above the 99th percentile upper reference limit in the appropriate clinical context.
Inclusion criteria were (i) confirmed STEMI; (ii) age ≥18 years; (iii) admission to the cardiology ward or intensive cardiac care unit during the study period; and (iv) documented Killip class at hospital presentation. Repeat admissions were excluded, retaining only the first admission. Killip class was determined at the time of initial hospital assessment and was not reclassified following subsequent clinical deterioration. Asystole was recorded if documented at presentation or at any time during the index hospitalization; therefore, transient arrhythmic events occurring after the initial haemodynamic evaluation did not alter the baseline Killip class assignment.
The study population consisted of 288 consecutively admitted patients with confirmed STEMI during the study period. The patient selection process is illustrated in Figure 1.
Figure 1.
Flow diagram illustrating identification of the study cohort and deviation of analytical samples.
Killip class was assigned prospectively at the time of initial hospital assessment (emergency department or intensive cardiac care unit) before coronary angiography and prior to any in-hospital clinical deterioration. Classification was based solely on haemodynamic and clinical findings at presentation and defined as follows: Class I (no signs of heart failure), Class II (S3 gallop or pulmonary rales), Class III (acute pulmonary oedema), and Class IV (cardiogenic shock).
Asystole was defined as documentation of a non-shockable rhythm (absence of ventricular electrical activity) recorded either at presentation or during the index hospitalization, as noted in the medical record. Shockable arrest rhythms (ventricular fibrillation or ventricular tachycardia) and the exact timing of cardiac arrest in relation to procedures or clinical events were not consistently recorded and therefore were not analysed separately.
Cardiogenic shock was defined as persistent hypotension (systolic blood pressure <90 mmHg for ≥30 min or requiring vasopressor support) with signs of end-organ hypoperfusion. Although Killip Class IV corresponds clinically to cardiogenic shock, cardiogenic shock was also recorded as a documented clinical diagnosis in the medical record to allow evaluation of potential overlap and documentation variability.
Clinical data collection
Baseline demographic and clinical data were extracted from electronic medical records, including age, gender, body mass index (BMI), smoking status, hypertension, diabetes mellitus, dyslipidaemia, and prior cardiovascular history. Clinical variables recorded at admission included atrial fibrillation, asystole, cardiogenic shock, and out-of-hospital cardiac arrest (OHCA). Reperfusion strategy was documented for all patients and categorized as primary percutaneous coronary intervention (PCI) performed, not performed, or unsuccessful. Unsuccessful PCI was defined as failure to achieve effective coronary reperfusion or inability to complete the procedure for technical or clinical reasons.
Cardiac function and biomarkers
Cardiac function was assessed by left ventricular ejection fraction (LVEF), measured by transthoracic echocardiography during the index hospitalization using standard clinical protocols. High-sensitivity troponin level (TL) (ng/L) was measured prior to percutaneous coronary intervention from hospital laboratory testing. The 99th percentile upper reference limit was 28 ng/L. The troponin concentration obtained before coronary intervention was used for analysis.
Outcome definition
The primary outcome was in-hospital all-cause mortality, defined as death occurring during the index hospitalization for STEMI.
Ethics
This retrospective study was approved by the Bioethics Centre of the Lithuanian University of Health Sciences (Approval No. 2024-BEC2-906; 2 October 2024). In accordance with Lithuanian national regulations governing biomedical research and institutional data protection policies, individual informed consent for medical treatment was obtained before or during the patient’s hospital care. All data were anonymized before analysis. The study was conducted in accordance with the Declaration of Helsinki.
Statistical analysis
Categorical variables are presented as counts and percentages and were compared between Killip classes using Pearson’s χ2 test. Continuous variables were assessed for normality and, due to non-normal distributions, are presented as medians with minimum and maximum values. Comparisons across Killip classes were performed using the Kruskal–Wallis test, with Bonferroni-adjusted pairwise comparisons where appropriate.
The primary endpoint was in-hospital all-cause mortality. Of the 288 included patients, mortality status was available for 286, and these patients were included in the primary endpoint analyses.
To identify independent predictors of in-hospital mortality, a multivariable binary logistic regression analysis was performed. Candidate predictors were selected a priori based on clinical relevance and included age, gender, BMI, smoking status, diabetes mellitus, dyslipidaemia, atrial fibrillation, asystole, Killip class, left ventricular ejection fraction, and troponin level. Variable selection was performed using the backward likelihood ratio (backward LR) method. Due to missing values in selected covariates, complete-case analysis resulted in a final regression sample of 275 patients. Model calibration was assessed using the Hosmer–Lemeshow goodness-of-fit test. Multicollinearity between the variables included in the model was evaluated using variance inflation factors (VIF), tolerance statistics, and pairwise correlation coefficients. VIF values > 2, tolerance values < 0.6, and correlation coefficients |r| > 0.5 were considered as indicative of multicollinearity.
Model discrimination was assessed using receiver operating characteristic (ROC) curve analysis, with calculation of the area under the curve (AUC) and corresponding 95% confidence intervals using DeLong’s method.
Because precise time-to-event data were not consistently available and the endpoint was defined as in-hospital mortality during the index admission, logistic regression was considered the most appropriate modelling approach rather than time-to-event (Cox) regression analysis.
Statistical significance for all analyses was defined as a two-sided P-value <0.05 and was performed using IBM SPSS Statistics (version 29).
Results
A total of 288 consecutively identified patients with confirmed ST-segment elevation myocardial infarction (STEMI) were included in the study cohort. In-hospital mortality status was available for 286 patients, who were included in the primary endpoint analyses. Baseline characteristics of the overall cohort are summarized in Table 1.
Table 1.
Baseline clinical characteristics of patients with ST-elevation myocardial infarction
| Patient characteristics | Value |
|---|---|
| Number of patients, n | 288 |
| Age, years, median (min, max) | 65 (34, 90) |
| Male, n (%) | 196 (68.1) |
| BMI, kg/m2, median (min, max) | 27.8 (17.7, 49.3) |
| Killip class at admission, n (%) | |
| Class I | 105 (36.4) |
| Class II | 101 (35.1) |
| Class III | 11 (3.8) |
| Class IV | 71 (24.7) |
| Cardiogenic shock at admission: yes, n (%) | 43 (14.9) |
| Smoking: yes, n (%) | 127 (44.1) |
| Dyslipidaemia: yes, n (%) | 220 (76.4) |
| Diabetes: yes, n (%) | 44 (15.3) |
| Hypertension: yes, n (%) | 211 (73.3) |
| AF: yes, n (%) | 56 (19.9) (7 MV) |
| Asystole: yes, n (%) | 34 (11.8) |
| Primary PCI, n (%) | (4 MV) |
| Not performed | 38 (13.3) |
| Successful | 223 (78.0) |
| Faileda | 25 (8.7) |
For variables without an indicated count of missing values, data are complete.
BMI, body mass index; AF, atrial fibrillation; PCI, percutaneous coronary intervention; MV, count of missing values.
aFailed PCI defined as the inability to achieve effective coronary reperfusion or procedural failure.
Distribution of Killip classes in patient characteristics
Killip Class I was the most common form, closely followed by Class II (101, 35.1%) and IV, while Class III was less common. The distribution across Killip classes differed significantly [χ2 (3) = 75.5, P < 0.001]. A significant gender disparity was observed, with males predominating in Class I (77.1%), whereas females were more frequently represented in Class IV (43.7%) [χ2 (3) = 11.8, P = 0.008] (Table 2).
Table 2.
Distribution of patient characteristics according to the Killip classes
| Patient characteristics | Killip class | Test (sig.) | |||
|---|---|---|---|---|---|
| I | II | III | IV | ||
| Gender (M), n (%) | 81 (77.1%) | 65 (64.4%) | 10 (90.9%) | 40 (56.3%) | χ 2 (3) = 11.8, P = 0.008 |
| Smoking (yes), n (%) | 53 (50.5%) | 51 (50.5%) | 6 (54.5%) | 17 (23.9%) | χ 2 (3) = 15.6, P = 0.001 |
| Dyslipidaemia (yes), n (%) | 89 (84.8%) | 76 (75.2%) | 7 (63.6%) | 48 (67.6%) | χ 2 (3) = 8.2, P = 0.042 |
| AF (yes), n (%) | 11 (10.9%) | 15 (15.3%) | 3 (27.3%) | 27 (38.0%) | χ 2 (3) = 21.4, P < 0.001 |
| Asystole (yes), n (%) | 2 (1.9%) | 4 (4.0%) | 0 (0%) | 28 (39.4%) | χ 2 (3) = 69.4, P < 0.001 |
| CS (yes), n (%) | 0 (0.0%) | 7 (4.2%) | 3 (8.3%) | 33 (66.0%) | χ 2 (3) = 125.2, P < 0.001 |
| Age (years), median (min, max) | 61 (34, 89) | 63 (39, 87) | 64 (42, 75) | 75 (48, 90) | H (3) = 49.7, P < 0.001 |
Results are based on the overall sample (n = 288).
M, males; AF, atrial fibrillation; CS, cardiogenic shock at admission.
Lifestyle and metabolic risk factors exhibited variable distributions across Killip classes. Smoking prevalence declined with increasing clinical severity {from 50.5% in Classes I and II to 23.9% in Class IV [χ2 (3) = 15.6, P = 0.001]}. Dyslipidaemia was common across the cohort (76.4%) but more prevalent in Class I (84.8%) compared to Class IV (67.6%) [χ2 (3) = 8.2, P = 0.042].
Hypertension and diabetes mellitus were frequent comorbidities (73.3 and 15.3%, respectively), but neither showed statistically significant variations across Killip classes (P > 0.05). In contrast, atrial fibrillation (AF) increased significantly with Killip severity—from 10.9% in Class I to 38.0% in Class IV [χ2 (3) = 21.4, P < 0.001]. The presence of asystole showed an even steeper gradient, increasing from 1.9% in Class I to 39.4% in Class IV [χ2 (3) = 69.4, P < 0.001].
Age also increased significantly with Killip class, from a median of 61 years in Class I to 75 years in Class IV [H (3) = 49.7, P < 0.001]. Body mass index (BMI), on the other hand, did not differ significantly between the classes [median 27.8–28.4 kg/m2; H (3) = 5.6, P = 0.13] and remained in the overweight category throughout the entire cohort.
Cardiogenic shock was documented in 43 patients; among the 71 patients in Killip Class IV, 33 (66.0%) had documented cardiogenic shock.
Cardiac function and biomarker trends
Two important cardiac indicators [left ventricular ejection fraction (EF) and troponin level (TL)] were analysed across Killip classes (Table 3). Both showed statistically significant correlations with increasing clinical severity.
Table 3.
Results on clinical indicators of cardiac function
| Cardiac function | Class I | Class II | Class III | Class IV | Test (sig.) |
|---|---|---|---|---|---|
| EF (%), median (min, max) 40 (10, 68) |
43 (15.6, 68) | 40 (15, 60) | 25 (17, 40) | 25 (10, 55) | H (3) = 92.5, P < 0.001 |
| TL (ng/L), median (min, max) 11.9 (0.2, 180) |
4.1 (0.2, 180) |
9.6 (0.2, 180) |
21.2 (0.2, 155) |
48.4 (0.28, 180) |
H (3) = 54.0, P < 0.001 |
Medians (min, max) in the cardiac function column refer to overall sample descriptives (n = 288).
EF, ejection fraction; TL, high-sensitivity troponin level.
The ejection fraction (EF) decreased progressively with Killip class, from a median value of 43% in Class I to 25% in Classes III and IV [H (3) = 92.5, P < 0.001]. Post hoc comparisons confirmed that patients in Killip Classes III–IV had significantly lower EF compared to Classes I–II, indicating marked systolic dysfunction.
Troponin level showed a significant increase from 4.1 ng/L in Class I to 9.6 ng/L in Class II, peaking at 48.4 ng/L in Class IV [H (3) = 54.0, P < 0.001]. Post hoc tests confirmed that Classes I vs. II, I vs. IV, and II vs. IV differed significantly. Killip class was significantly associated with markers of myocardial injury and systolic dysfunction. However, post hoc comparisons showed that the most pronounced differences occurred between the lower and higher classes, rather than indicating a strictly progressive gradient across all categories.
Infarct localization and mortality
The relationship between the localization of myocardial infarction (MI) and the Killip class was investigated (Table 4). Anterior MI was the most common overall (53.5%), followed by inferior (44.1%) and posterior infarction (14.2%). Anterior and inferior MI were not significantly associated with Killip class (P = 0.108 and P = 0.217, respectively). However, posterior MI was significantly more common in Killip Class I (32.4%) and less common in Class IV (2.8%) [χ2 (3) = 44.9, P < 0.001].
Table 4.
Localization of myocardial infarction and mortality according to Killip classes
| Condition | Class I | Class II | Class III | Class IV | Test (sig.) |
|---|---|---|---|---|---|
| Anterior MI (yes), n (%) 154 (53.5%) |
49 (46.7%) | 55 (54.5%) | 9 (81.8%) | 41 (57.7%) | χ 2 (3) = 6.1, P = 0.108 |
| Posterior MI (yes), n (%) 41 (14.2%) |
34 (32.4%) | 5 (5.0%) | 0 (0.0%) | 2 (2.8%) | χ 2 (3) = 44.9, P < 0.001 |
| Inferior MI (yes), n (%) 127 (44.1%) |
52 (49.5%) | 43 (42.6%) | 2 (18.2%) | 30 (42.3%) | χ 2 (3) = 4.4, P = 0.217 |
| Mortality (yes), n (%) 52 (18.2%) |
1 (1.0%) | 3 (3.0%) | 0 (0.0%) | 48 (69.6%) | χ 2 (3) = 161.6, P < 0.001 |
Percentages in the condition column refer to condition (yes) vs. condition (no). Percentages in other columns are row-based values from condition (yes)–Killip class crosstabs; they reflect condition within each Killip class separately. Two patients had missing in-hospital mortality status; these results are based on n = 286. Values in bold indicate statistically significant results.
Mortality varied statistically significantly between classes ranging from 1.0% in Class I to 3.0% in Class II and to 69.6% in Class IV [χ2 (3) = 161.6, P < 0.001]. Posterior MI was associated with significantly lower mortality [16.7% vs. 3.8%, χ2 (1) = 5.7, P = 0.02], while anterior and inferior infarcts showed no significant differences.
Direct comparison of infarct location with mortality (independent of Killip class) revealed no significant difference in outcomes for anterior MI [53.0% vs. 53.8%, χ2 (1) = 0.01, P = 0.91] or inferior MI [44.0% vs. 46.2%, χ2 (1) = 0.08, P = 0.78], suggesting that there were no significant survival differences based on these MI locations. However, patients with posterior infarction showed a significantly higher survival rate than patients without posterior infarction [16.7% vs. 3.8%, χ2 (1) = 5.7, P = 0.02].
Predictors of mortality and discriminative performance
To identify independent predictors of in-hospital mortality, multivariable binary logistic regression analysis was performed. Candidate predictors were selected a priori based on clinical relevance and included age, gender, BMI, smoking status, diabetes mellitus, dyslipidaemia, atrial fibrillation, asystole, Killip class, left ventricular ejection fraction, and troponin level. Due to missing values in death records (two cases) and other covariates, the model was based on 275 valid cases with a discrepancy from the total sample of 288. From a total sample of 275, there were 52 (18.9%) death cases. The final model was selected using the backward likelihood ratio (backward LR) method.
The resulting model (Table 5) was statistically significant [χ2 (5) = 180.7, P < 0.001], showed a good fit to the data [Hosmer–Lemeshow test: χ2 (8) = 4.2, P = 0.843], and explained 76.5% of the variance in mortality (Nagelkerke R2 = 0.765). The overall classification accuracy reached 93.7%.
Table 5.
Multivariate logistic regression analysis of mortality risk
| Variable | Beta | P-value | OR | 95% CI for OR |
|---|---|---|---|---|
| Age, years | 0.062 | 0.037 | 1.06 | 1.004–1.129 |
| BMI, kg/m2 | 0.148 | 0.016 | 1.16 | 1.03–1.31 |
| Troponin level, ng/L, | 0.014 | 0.003 | 1.014 | 1.005–1.023 |
| Killip Class IV (yes) | 4.11 | <0.001 | 60.94 | 15.98–232.46 |
| Asystole (yes) | 2.705 | <0.001 | 14.95 | 3.53–63.29 |
| Constant | −13.413 | <0.001 | 0.0 | – |
‘Yes’ is the reference category of qualitative variables. Due to missing data in covariates, n = 275.
Five variables remained significant in the final model: age, BMI, troponin level, Killip Class IV status, and asystole. Gender (P = 0.781), smoking status (P = 0.671), diabetes mellitus (P = 0.539), dyslipidaemia (P = 0.873), atrial fibrillation (P = 0.986), left ventricular ejection fraction (P = 0.985), and cardiogenic shock (P = 0.672) were removed. All VIF indexes vary from 1.1 to 1.4, tolerance values vary from 0.7 to 0.9, and pairwise correlation coefficients (|r|) were not higher than 0.4; all these parameters indicate that there is no significant multicollinearity while building the model.
The variables in the model reflect both the baseline characteristics of the patient and the acute clinical deterioration. In particular, Killip Class IV and asystole proved to be the strongest predictors of mortality, underlining the prognostic importance of haemodynamic instability and severe conduction disturbance.
Each additional year of life increased the risk of death by 6% (OR = 1.06, 95% CI = 1.004–1.129). The increase of BMI by 1 kg/m2 is significantly associated with a 16% increase in the risk of death (OR = 1.16, 95% CI = 1.03–1.31). Troponin level was also significantly predictive, with each one-unit increase (ng/L), associated with an increased risk of death by 1.4% (OR = 1.014, 95% CI = 1.005–1.023).
The strongest association with mortality was observed in patients in Killip Class IV, where the odds of death were more than 60 times higher than in patients in lower classes (OR = 60.94, 95% CI = 15.98–232.46). Similarly, patients admitted with asystole had an almost 15-fold higher odds of dying in-hospital (OR = 14.95, 95% CI = 3.53–63.29).
The discriminatory power of the model was excellent. Receiver operating characteristics (ROC) analysis yielded an area under the curve (AUC) of 0.969 (95% CI: 0.945–0.992, P < 0.001) (Figure 2), confirming the model’s strong ability to discriminate between survivors and non-survivors.
Figure 2.
ROC curve showing the accuracy of the model in predicting mortality.
These results confirm the Killip classification as a robust predictor of short-term mortality and support the inclusion of age, biomarkers, and cardiac arrest indicators in early prognostic models for AMI patients. The data emphasize the importance of rapid clinical triage in emergency situations to inform triage and treatment decisions.
Discussion
In this retrospective cohort of patients with ST-elevation myocardial infarction (STEMI), we assessed the prognostic performance of the Killip classification in a contemporary, real-world tertiary care setting in Eastern Europe. Our findings confirm that Killip Class IV remains a strong independent predictor of in-hospital mortality, while prognostic separation among Killip Classes I–III was more modest. Nevertheless, when considered alongside age, biomarkers, and arrhythmic complications, the Killip classification retains clinical relevance as a pragmatic bedside risk stratification tool in modern STEMI care.4,12–14
Large contemporary registries, including JAMIR and SAIKUMA, have consistently validated the prognostic significance of the Killip classification, particularly among patients presenting with cardiogenic shock (Killip Class IV).4 Mello et al.13 (2014) further demonstrated the long-term prognostic relevance of the Killip system in a Brazilian AMI population, supporting its applicability across diverse healthcare settings. In these cohorts, Killip Class IV patients typically represented a relatively small proportion of admissions and experienced in-hospital mortality rates of approximately 30–50% despite contemporary reperfusion strategies.
In contrast, our cohort showed a higher prevalence of Killip Class IV and an in-hospital mortality rate approaching 70% in this group. Several contextual factors may have contributed to this observation. A substantial proportion of patients presented with out-of-hospital cardiac arrest (18%), a recognized determinant of early mortality. Additionally, more advanced haemodynamic compromise at admission and potential delays in pre-hospital stabilization may have influenced clinical severity. Limited access to advanced mechanical circulatory support in resource-constrained settings may also have further affected outcomes. Similar patterns have been described in lower-resource environments, where high Killip class STEMI patients demonstrated increased mortality associated with impaired reperfusion and systemic instability.15
An important observation was the absence of a pronounced mortality gradient across Killip Classes I–III. Although earlier studies reported progressively increasing mortality across all Killip classes,16 mortality rates in our cohort were relatively similar among the lower classes. This may reflect limited subgroup sizes, the predominance of critically ill Killip Class IV patients, and widespread use of guideline-directed therapies in lower-risk presentations. Contemporary evidence indicating that multivariable risk scores, including GRACE, TIMI, and HEART, provide more granular prognostic discrimination by integrating age, comorbidities, ECG findings, and biomarkers.17–19 While the Killip classification remains clinically useful, it may be insufficient as a standalone tool for refined risk stratification among lower-risk STEMI patients in the modern PCI era.5,6,10
Notably, part of the study period coincided with the COVID-19 pandemic. Several reports have described delayed presentation and increased clinical severity among patients with acute coronary syndromes during pandemic waves. Although we did not perform a formal temporal analysis, pandemic-related healthcare system disruptions and delayed care-seeking behaviour may have influenced the severity of presentation in our cohort.
Age was strongly associated with Killip class severity, with patients in Killip Class IV being significantly older than those in lower classes, consistent with prior reports.12,14 Electrical conduction abnormalities, particularly atrial fibrillation and asystole, were more frequent in higher Killip classes and were associated with increased mortality, reflecting advanced haemodynamic compromise.10,20–22 Left ventricular ejection fraction was lower, and troponin levels were higher in patients with advanced Killip class, although post hoc analyses suggested that the most pronounced differences were observed between the lower and highest classes rather than demonstrating a strictly progressive gradient across all four categories.6,10
Other cardiovascular risk factors, including hypertension, diabetes mellitus, dyslipidaemia, smoking status, and gender, varied across Killip classes but were not independently associated with in-hospital mortality, underscoring the dominant prognostic impact of acute haemodynamic and electrical instability at presentation.
In multivariable analysis, age, body mass index, troponin level, Killip Class IV, and asystole emerged as independent predictors of in-hospital mortality. The strong association between asystole and mortality is consistent with previous reports highlighting the prognostic impact of severe conduction disturbances and electromechanical instability in acute coronary syndromes.20–22 The observed association between body mass index and mortality reflects the complex relationship between adiposity and acute cardiovascular outcomes.23
The final model demonstrated excellent discriminative performance (AUC 0.969). These findings suggest that clinically accessible variables, when appropriately integrated, can provide robust prognostic information without reliance on more complex modelling approaches.24,25
Key strengths of this study include a prospective Killip class assignment at presentation, comprehensive characterization using routinely available bedside variables, and analysis of a contemporary STEMI cohort from an underrepresented Eastern European tertiary centre. The exclusive inclusion of STEMI patients ensured cohort homogeneity and avoided heterogeneity associated with mixed acute coronary syndrome populations.
Several limitations should be acknowledged. First, the retrospective, single-centre design and moderate sample size limited statistical power for subgroup analyses, particularly within Killip Classes II and III. Second, the primary endpoint was restricted to in-hospital all-cause mortality, with no 30-day or long-term follow-up data available. Third, several clinically relevant variables, including door-to-balloon times, pre-procedural pharmacotherapy, mechanical complications of myocardial infarction, and advanced circulatory support, were not uniformly documented in the registry. Detailed classification of cardiac arrest rhythms and precise timing of arrhythmic events were also not consistently available; therefore, asystole should be interpreted as a marker of severe electrical and haemodynamic instability, rather than as a comprehensive characterization of arrest mechanisms.
Additionally, creatine kinase (CK) or CK-MB measurements were not systematically available in this dataset. Although high-sensitivity troponin is the current guideline-recommended biomarker for myocardial injury, the absence of CK data may limit comparisons with studies using multimarker approaches or infarct size estimation based on CK kinetics.
Finally, the relatively high proportion of patients who did not undergo PCI or had unsuccessful PCI represents an important contextual limitation. Contemporary STEMI cohorts from high-volume centres typically report higher rates of successful primary PCI. Differences in reperfusion success may influence haemodynamic status at presentation, infarct size, and early mortality risk. Accordingly, caution is warranted when directly comparing our findings with populations characterized by near-universal successful revascularization. This factor may also have contributed to the relatively high proportion of patients presenting with advanced Killip class and to the observed mortality rates.
Although established risk scores such as GRACE and TIMI provide more comprehensive prognostic assessment, the Killip classification offers immediate haemodynamic evaluation at first medical contact. In this real-world cohort, Killip Class IV remained strongly associated with in-hospital mortality, whereas prognostic differentiation among lower classes was more modest. These findings support the continued use of the Killip classification as a rapid bedside tool, particularly in unstable patients and resource-limited settings, when integrated into a broader, multimodal risk stratification strategy.
Conclusion
In this contemporary real-world STEMI cohort, the Killip classification remained a strong and independent predictor of in-hospital mortality, particularly among patients presenting with Killip Class IV. Despite being introduced more than five decades ago, the Killip system continues to reflect clinically relevant haemodynamic compromise and retains prognostic value in modern acute myocardial infarction care.
Multivariable analysis showed that age, body mass index, troponin level, Killip Class IV, and asystole independently associated with in-hospital mortality, and their combined use provided excellent discriminative performance. While mortality risk increased substantially in patients with advanced Killip class, differentiation among lower classes (I–III) was more modest in this contemporary PCI-treated cohort.
These findings support the continued use of the Killip classification as a rapid and standardized bedside assessment tool, particularly in patients presenting with haemodynamic instability. However, optimal risk stratification in modern STEMI care likely requires integration of clinical assessment with objective biomarkers and multivariable risk models. Future studies should focus on external validation in diverse populations and on the development of integrated prognostic frameworks combining clinical, biochemical, and rhythm-based parameters.
Contributor Information
Vita Speckauskiene, Department of Physics, Mathematics and Biophysics, Lithuanian University of Health Sciences, Eiveniu str. 4, Room 417, Kaunas LT-50161, Lithuania.
Diana Meilutyte-Lukauskiene, Department of Physics, Mathematics and Biophysics, Lithuanian University of Health Sciences, Eiveniu str. 4, Room 417, Kaunas LT-50161, Lithuania; Laboratory of Hydrology, Lithuanian Energy Institute, Breslaujos str. 3, Kaunas LT-44403, Lithuania.
Reda Cerapaite-Trusinskiene, Department of Physics, Mathematics and Biophysics, Lithuanian University of Health Sciences, Eiveniu str. 4, Room 417, Kaunas LT-50161, Lithuania.
Jannik Mueller, Department of Physics, Mathematics and Biophysics, Lithuanian University of Health Sciences, Eiveniu str. 4, Room 417, Kaunas LT-50161, Lithuania.
Andrius Macas, Department of Anesthesiology, Lithuanian University of Health Sciences, Eiveniu str. 2, Kaunas LT-50161, Lithuania.
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Data availability
The data supporting the findings of this study are available from the corresponding author upon reasonable request.
References
- 1. World Heart Federation . World Heart Report: confronting the world's number one killer. Available from: https://world-heart-federation.org/resource/world-heart-report-2023; 2023 (accessed 8 Sep. 2025).
- 2. World Health Organization . Cardiovascular diseases. Available from: https://www.who.int/health-topics/cardiovascular-diseases#tab=tab_1 2025 (accessed 8 Sep. 2025).
- 3. Cox LJ, Bata IR, Gregor RD, Johnstone DE, Wolf HK. Trends in event rate and case fatality of patients hospitalized with myocardial infarction between 1984 and 2001. Can J Physiol Pharmacol 2006;84:121–127. [DOI] [PubMed] [Google Scholar]
- 4. Taguchi E, Konami Y, Inoue M, Suzuyama H, Kodama K, Yoshida M, Miyamoto S, Nakao K, Sakamoto T. Impact of Killip classification on acute myocardial infarction: data from the SAIKUMA registry. Heart Vessels 2017;32:1439–1447. [DOI] [PubMed] [Google Scholar]
- 5. Wohlfahrt P, Jenča D, Melenovský V, Stehlik J, Mrázková J, Šramko M, Kotrč M, Želízko M, Adámková V, Piťha V, Kautzner J. Development and validation of a prognostic score integrating remote heart failure symptoms and clinical variables in mortality risk prediction after myocardial infarction: the PragueMi score. Eur J Prev Cardiol 2024;31:1713–1720. [DOI] [PubMed] [Google Scholar]
- 6. Vicent L, Velásquez-Rodríguez J, Valero-Masa MJ, Díez-Delhoyo F, González-Saldívar H, Bruña V, Devesa C, Juárez M, Sousa-Casasnovas I, Fernández-Avilés F, Martínez-Sellés M. Predictors of high Killip class after ST segment elevation myocardial infarction in the era of primary reperfusion. Int J Cardiol 2017;248:46–50. [DOI] [PubMed] [Google Scholar]
- 7. Armillotta M, Amicone S, Bergamaschi L, Angeli F, Rinaldi A, Paolisso P, Stefanizzi A, Sansonetti A, Impellizzeri A, Bodega F, Canton L, Suma N, Fedele D, Bertolini D, Foà F, Pizzi C. Predictive value of Killip classification in MINOCA patients. Eur J Intern Med 2023;117:57–65. [DOI] [PubMed] [Google Scholar]
- 8. Chen L, Fan Y, Fang Z, Liu N. Long-term outcomes and predictors of patients with ST elevated versus non-ST elevated myocardial infarctions in non-obstructive coronary arteries: a retrospective study in Northern China. PeerJ 2023;11:e14958. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Lemesle G, Puymirat E, Bonello L, Simon T, Steg PG, Ferrières J, Schiele F, Fauchier L, Henry P, Schurtz G, Ninni S, Lamblin N, Bauters C, Danchin N. Compared impact of diabetes on the risk of heart failure from acute myocardial infarction to chronic coronary artery disease. Diabetes Metab 2022;48:101265. [DOI] [PubMed] [Google Scholar]
- 10. Sasaki O, Nishioka T, Inoue Y, Isshiki A, Sasaki H. Predictors of in-hospital death in patients with acute myocardial infarction. Cureus 2023;15:e43392. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Chong B, Jayabaskaran J, Jauhari SM, Chan SP, Goh R, Kueh MTW, Li H, Chin YH, Kong G, Anand VV, Wang JW, Muthiah M, Jain V, Mehta A, Lim SL, Foo R, Figtree GA, Nicholls SJ, Mamas MA, Januzzi JL, Chew NWS, Richards AM, Chan MY. Global burden of cardiovascular diseases: projections from 2025 to 2050. Eur J Prev Cardiol 2024;32:1001–1015. [Google Scholar]
- 12. Hashmi KA, Adnan F, Ahmed O, Yaqeen SR, Ali J, Irfan M, Edhi MM, Hashmi AA. Risk assessment of patients after ST-segment elevation myocardial infarction by Killip classification: an institutional experience. Cureus 2020;12:e12209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Mello BH, Oliveira GB, Ramos RF, Lopes BB, Barros CB, Carvalho Ede O, Teixeira FB, Arruda GD, Revelo MS, Piegas LS. Validation of the Killip-Kimball classification and late mortality after acute myocardial infarction. Arq Bras Cardiology 2014;103:107–117. [Google Scholar]
- 14. Zadok OIB, Ben-Gal T, Abelow A, Shechter A, Zusman O, Iakobishvili Z, Cohen T, Shlomo N, Kornowski R, Eisen A. Temporal trends in the characteristics, management and outcomes of patients with acute coronary syndrome according to their Killip class. Am J Cardiol 2019;124:1862–1868. [DOI] [PubMed] [Google Scholar]
- 15. Riaz G, Khan N, Haq EU, Kumari S, Hussain I, Zada S. Mortality predictors and in-hospital outcomes in high Killip class STEMI patients: insights from a tertiary cardiac center. Pakistan heart journal (Karachi) 2025;58:51–58. [Google Scholar]
- 16. Khot UN, Jia G, Moliterno DJ, Michael Lincoff A, Khot MB, Harrington RA, Topol EJ. Prognostic importance of physical examination for heart failure in non–ST-elevation acute coronary syndromes. JAMA 2003;290:2174. [DOI] [PubMed] [Google Scholar]
- 17. Yüzlü H, Işık S, Doğan H. Comparison of GRACE, HEART and TIMI scores in predicting major adverse cardiac events in patients visiting the emergency department. Global Emergency Crit Care 2025. Epub ahead of print. [Google Scholar]
- 18. Cakar MA, Sahinkus S, Aydin E, Vatan MB, Keser N, Akdemir R, Gunduz H. Relation between the GRACE score and severity of atherosclerosis in acute coronary syndrome. J Cardiol 2014;63:24–28. [DOI] [PubMed] [Google Scholar]
- 19. Zhu H, Xue H, Wang H, Chen Y, Zhou S, Tian F, Hu S, Wang J, Yang J, Zhang T. Risk stratification and prognostic value of GRACE and TIMI risk scores for female patients with non-ST segment elevation acute coronary syndrome. Int J Clin Exp Med 2015;8:4038–4044. [PMC free article] [PubMed] [Google Scholar] [Retracted]
- 20. Shinde S, Jadhav N. Study of conduction blocks in ST elevation myocardial infarction—a cross-sectional analysis. Eurasian J Med 2024;56:148–152. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Cercato C, Fonseca FA. Cardiovascular risk and obesity. Diabetol Metab Syndr 2019;11:74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Pokorney SD, Radder C, Schulte PJ, Al-Khatib SM, Tricocci P, Van de Werf F, James SK, Cannon CP, Armstrong PW, White HD, Califf RM, Gibson CM, Giugliano RP, Wallentin L, Mahaffey KW, Harrington RA, Newby LK, Piccini JP. High-degree atrioventricular block, asystole, and electro-mechanical dissociation complicating non-ST-segment elevation myocardial infarction. Am Heart J 2016;171:25–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Luo W, Wen L, Zhang J, Zhao J, Wang Z, Luo X, Pi S, Chen Y, Zhang J, Li T, Zhang Z, Luo D, Qin Z, Yu S. The short-term outcomes and risk factors of post-myocardial infarction ventricular septal rupture: a multi-center retrospective study. Journal Cardiothorac Surg 2024;19:571. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Rosa SA, Timóteo AT, Nogueira MA, Belo A, Ferreira RC. Risk factor paradox in the occurrence of cardiac arrest in acute coronary syndrome patients. Rev Bras Ter Intensiva 2016;28:405–412. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Gupta MD, Goyal D, Kunal S, Shetty MK, Girish MP, Batra V, Bansal A, Mishra P, Shukla M, Kohli V, Chadha A, Fatima A, Muduli S, Gupta A, Yusuf J. Comparative evaluation of machine learning models versus TIMI score in ST-segment-elevation myocardial infarction patients. Indian Heart J 2025;77:133–141. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data supporting the findings of this study are available from the corresponding author upon reasonable request.