Electrocardiographic (ECG) findings are commonly used to evaluate myocardial infarction (MI), yet their ability to distinguish Type 1 from Type 2 MI in critically ill patients remains uncertain. We conducted a retrospective observational study using the MIMIC-IV database to examine associations between ECG features and MI subtype among adult intensive care unit patients with elevated troponin and ICD-10–defined Type 1 or Type 2 MI. Structured ECG interpretation features, including ST segment changes, T-wave abnormalities, QTc prolongation, and arrhythmia were analyzed using multivariable logistic regression. Among 2,159 patients, ST segment elevation was uncommon but significantly associated with Type 1 MI (9.4% vs 3.1%; adjusted odds ratios 2.48, 95% CI 1.31 to 5.15), whereas QTc prolongation was highly prevalent—particularly in Type 2 MI (66.2% vs 52.9%)—and independently associated with lower odds of Type 1 MI (adjusted odd ratios 0.63, 95% CI 0.49 to 0.81). Overall model discrimination was modest (area under the curve 0.73), but diagnostic performance was characterized by near-perfect sensitivity and very low specificity due to substantial class imbalance favoring Type 1 MI. In conclusion, ECG features alone provide limited clinical and mechanistic utility for differentiating Type 1 from Type 2 MI in critically ill patients and should be interpreted within the broader clinical and physiologic context.
In the United States, there are an estimated 805,000 myocardial infarctions (MI) every year, of which 605,000 are new events. , The total annual economic burden of MI is approximately $84.9 billion, encompassing both direct medical expenditures and indirect costs related to lost productivity and wages. The substantial mortality, morbidity, and economic impact of MI on individuals and the healthcare system underscore the critical need to improve its understanding, diagnosis, and management.
MI is classically defined as myocardial cell death resulting from ischemia. It is further classified into 5 distinct subtypes based on underlying pathophysiology. , Type 1 MI (T1MI), the most common subtype, is caused by an atherosclerotic process leading to coronary thrombosis. Type 2 MI (T2MI) is categorized as myocardial injury caused by a mismatch of oxygen supply and demand in the absence of an atherothrombotic plaque. ,,, Types 3 to 5 are much less common and are defined as sudden death from presumed cardiac origin lacking biomarker evidence (Type 3), percutaneous coronary intervention induced cardiac injury (Type 4), and coronary artery bypass grafting related MI (Type 5). ,
Understanding the distinct pathophysiological frameworks of MI are important as the underlying mechanisms directly influence diagnostic evaluation and treatment, particularly when distinguishing T1MI and T2MI. While evidence-based strategies for T1MI management are well established, optimal management strategies for T2MI remain poorly defined. ,, This likely reflects the heterogenous etiologies and triggers associated with T2MI, which frequently necessitate individualized approaches that address the underlying cause. Despite this, prior studies demonstrated that patients with T2MI experience worse outcomes and are less likely to receive cardioprotective therapies compared to patients with T1MI. ,
Given these challenges, accurate diagnostic strategies are essential to guide appropriate clinical decision-making and plans, particularly in critically ill patients with complex comorbidities. Electrocardiograms can help detect MI by identifying ST segment changes, T-wave abnormalities, and/or Q-wave developments; however, these findings primarily reflect ischemic burden rather than an underlying mechanism. The ability of electrocardiograms to differentiate T1MI versus T2MI in the context of critically ill patients in the intensive care unit (ICU) remains uncertain. In this study, we use the MIMIC-IV databank from the Beth Israel Deaconess Medical Center to evaluate the associations between electrocardiographic (ECG) findings and myocardial subtype among ICU patients.
Methods
Data source and study design
This retrospective observational study used data from the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 2.2) database and its linked electrocardiography module (MIMIC-IV-ECG), both publicly available through PhysioNet. , MIMIC-IV contains deidentified electronic health records of patients admitted to the Beth Israel Deaconess Medical Center between 2008 and 2019, while the ECG module provides matched diagnostic electrocardiograms with structured interpretation fields. Access to the database required completion of the Collaborative Institutional Training Initiative training and approval of the data use agreement.
Study population
Adult patients (≥18 years) admitted to the ICU with evidence of myocardial infarction were eligible for inclusion. Patients were identified based on elevated cardiac troponin levels and International Classification of Diseases, Tenth Revision (ICD-10) diagnosis codes corresponding to Type 1 MI (I21.*) or Type 2 MI (I21.A1, I21.A9). For patients with multiple ICU admissions, only the first ICU stay was included to avoid duplication. Patients were excluded if ECG interpretation data, key clinical covariates, or outcome data were unavailable. After applying inclusion and exclusion criteria, the final analytic cohort consisted of 2,159 ICU patients.
Data extraction and variables
Data were extracted using Structured Query Language with PostgreSQL and processed in R (R Foundation for Statistical Computing). Clinical data were obtained from MIMIC-IV tables including admissions, icustays, chartevents, labevents, and diagnoses_icd. ECG features were obtained from the MIMIC-IV-ECG module, including structured interpretation, and machine-derived measurements. The primary exposure variables were ECG features, including ST segment elevation, ST segment depression, T-wave abnormalities, QTc prolongation, and arrhythmia. ECG features were defined using standardized interpretation fields provided in the ECG module. QTc prolongation was defined according to institutional reference thresholds. Clinical covariates included age, sex, anemia, sepsis, hypotension, arrhythmia, renal failure, hypertension, and heart rate. Anemia was defined using hemoglobin thresholds (<13 g/dl in men and <12 g/dl in women). Sepsis was identified using ICD-9 and ICD-10 diagnostic codes consistent with Sepsis-3 criteria. Renal failure, arrhythmia, and hypertension were defined using diagnostic codes. Maximum troponin level during the ICU stay was recorded. All dates in the database were shifted by a patient-specific random offset to preserve temporal relationships while maintaining deidentification, and patients aged ≥90 years were top-coded in accordance with HIPAA requirements.
Outcome definition
The primary outcome was MI subtype (Type 1 vs Type 2 MI), as defined by ICD-10 diagnostic coding. In clinical practice, MI subtyping is determined based on various information including clinical presentation, lab values, image findings, and ECG readings. Consequently, to mitigate a potential diagnostic circularity from ECG driven ICD-10 coding, we performed a sensitivity analysis excluding ST elevation MI. Although this sensitivity analysis cannot entirely eliminate diagnostic circularity, it supports the robustness of observed association by reducing the direct dependence on ST elevation for MI classification.
Statistical analysis
Baseline characteristics were summarized by MI subtype and compared using the Wilcoxon rank-sum test for continuous variables and the chi-square test for categorical variables. Univariate logistic regression models were used to evaluate associations between individual ECG features and MI subtype. Multivariable logistic regression models were constructed to assess the independent association between ECG features and Type 1 MI after adjustment for relevant clinical covariates.
Model discrimination was assessed using receiver operating characteristic analysis and area under the curve (AUC). Diagnostic performance metrics were reported descriptively to contextualize observed associations and were not intended to support clinical classification. Diagnostic performance was summarized using sensitivity, specificity, positive predictive value, and negative predictive value. Although receiver operating characteristic analysis identified an optimal cutoff probability of 0.818, this threshold resulted in trivial classification due to class imbalance. Therefore, a conventional cutoff of 0.5 was used for reporting diagnostic performance metrics. Given substantial class imbalance favoring Type 1 MI, diagnostic performance metrics should be interpreted as descriptive of observed associations rather than as evidence of clinically applicable classification performance. Results are reported as odds ratios (OR) with 95% confidence intervals (CI). A 2-sided p-value <0.05 was considered statistically significant.
A sensitivity test excluding patients with ST elevation on ECG was used to address possible diagnostic circularity from ECG influenced MI assignment. A multivariable regression model was reassessed in this non-ST–elevated ECG cohort to evaluate whether associations between ECG characteristic and MI subtype were preserved.
Ethical considerations
All analyzes were performed using RStudio with standard statistical packages. Because MIMIC-IV is a publicly available, fully deidentified database, this study was exempt from institutional review board approval.
Results
Study population
A total of 2,159 ICU patients with MI met the inclusion criteria, including 1,801 (83.4%) with Type 1 MI and 358 (16.6%) with Type 2 MI ( Table 1 ). The mean age was similar between groups (69.0 ± 13.95 vs 69.8 ± 12.90 years, p = 0.30). Male sex was more common among patients with Type 1 MI (65% vs 59%, p = 0.026). Sepsis (47% vs 17%, p < 0.001) and renal failure (72% vs 55%, p < 0.001) were more prevalent in Type 2 MI. Hypertension was more common in Type 1 MI (32% vs 20%, p < 0.001). Maximum troponin levels were significantly higher in Type 1 MI (3.29 ± 4.76 vs 0.97 ± 1.35, p < 0.001) ( Table 1 ).
Table 1
Baseline characteristics of ICU patients by myocardial infarction type (N = 2,159)
| Characteristic | Type 2 (N = 358) | Type 1 (N = 1,801) | p-value |
|---|---|---|---|
| Age (years) | 69.79 (12.90) | 69.00 (13.95) | 0.30 |
| Sex | 0.026 | ||
| Female | 147 (41%) | 628 (35%) | |
| Male | 211 (59%) | 1,172 (65%) | |
| Anemia | 48 (13%) | 253 (14%) | 0.70 |
| Sepsis | 170 (47%) | 312 (17%) | <0.001 |
| Hypotension | 87 (24%) | 454 (25%) | 0.70 |
| Arrhythmia | 215 (60%) | 891 (50%) | <0.001 |
| Renal Failure | 258 (72%) | 991 (55%) | <0.001 |
| Hypertension | 71 (20%) | 579 (32%) | <0.001 |
| ST Elevation | 11 (3.1%) | 170 (9.4%) | <0.001 |
| ST Depression | 7 (2.0%) | 35 (1.9%) | >0.90 |
| T-wave Abnormality | 7 (2.0%) | 87 (4.8%) | 0.015 |
| QT Prolongation | 237 (66%) | 952 (53%) | <0.001 |
| Heart Rate (bpm) | 90.34 (23.93) | 83.10 (21.62) | <0.001 |
| Max Troponin | 0.97 (1.35) | 3.29 (4.76) | <0.001 |
Mean (SD); n (%).
Prevalence of ECG findings
QT prolongation was the most common ECG abnormality in both groups and was more frequent in Type 2 MI (66.2% vs 52.9%, p < 0.001). ST segment elevation occurred more often in Type 1 MI (9.4% vs 3.1%, p < 0.001). T-wave abnormalities were also more frequent in Type 1 MI (4.8% vs 2.0%, p = 0.015), whereas ST depression occurred at similar rates between groups (1.9% vs 2.0%, p > 0.90) ( Table 7 ).
Univariable associations
In univariable logistic regression analysis, ST segment elevation was associated with increased odds of Type 1 MI (OR 3.29, 95% CI 1.85 to 6.49, p < 0.001), as was T-wave abnormality (OR 2.55, 95% CI 1.26 to 6.10, p = 0.019). QT prolongation was associated with lower odds of Type 1 MI (OR 0.57, 95% CI 0.45 to 0.73, p < 0.001). Sepsis (OR 0.23, 95% CI 0.18 to 0.30, p < 0.001), arrhythmia (OR 0.65, p < 0.001), and renal failure (OR 0.48, p < 0.001) were also associated with lower odds of Type 1 MI ( Table 2 ).
Table 2
Univariable logistic regression analysis for predictors of Type 1 vs Type 2 myocardial infarction (outcome: Type 1 MI)
| Variable | Odds Ratio (95% CI) | p-value |
|---|---|---|
| ST Elevation | 3.29 (1.85–6.49) | <0.001 |
| ST Depression | 0.99 (0.47–2.46) | 0.989 |
| T-wave Abnormality | 2.55 (1.26–6.10) | 0.019 |
| QT Prolongation | 0.57 (0.45–0.73) | <0.001 |
| Heart Rate | 0.99 (0.98–0.99) | <0.001 |
| Age | 0.996 (0.987–1.00) | 0.323 |
| Male Sex | 1.30 (1.03–1.64) | 0.027 |
| Anemia | 1.06 (0.76–1.49) | 0.747 |
| Sepsis | 0.23 (0.18–0.30) | <0.001 |
| Hypotension | 1.05 (0.81–1.37) | 0.714 |
| Arrhythmia | 0.65 (0.52–0.82) | <0.001 |
| Renal Failure | 0.48 (0.37–0.61) | <0.001 |
| Hypertension | 1.92 (1.46–2.55) | <0.001 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
