Prevalence and Impact of Myocardial Injury in Patients Hospitalized with COVID-19 Infection

Study Population

Patients in this study were drawn from five New York City hospitals (Mount Sinai Hospital, in East Harlem; Mount Sinai West, in Midtown Manhattan; Mount Sinai St. Lukes, in Harlem; Mount Sinai Queens, in Astoria; and Mount Sinai Brooklyn, in the Midwood neighborhood of Brooklyn). These hospitals comprise the Mount Sinai Health System (MSHS). We included all patients admitted to a MSHS hospital with a laboratory confirmed SARS-CoV-2 infection who were at least 18 years old and had a troponin measurement within the first 24 hours of admission between February 27th and April 12th, 2020. The Mount Sinai Institutional Review Board approved this research under a regulatory protocol allowing for analysis of patient-level COVID-19 data.

Data Collection

Data was collected from electronic health records (EHR) from the five hospitals. Variables collected included demographics, laboratory measurements, disease diagnoses, comorbidities, procedures, and outcomes (death or hospital discharge). Comorbidities were extracted using International Classification of Disease (ICD) 9/10 billing codes for atrial fibrillation (AF), asthma, coronary artery disease (CAD), cancer, chronic kidney disease (CKD), chronic obstructive pulmonary disease (COPD), diabetes (DM), heart failure (HF), and hypertension (HTN). Troponin I concentrations were assessed via the Abbott Architect method (Abbott, Abbott Park, Illinois) wherein the 99th percentile for a normal population is 0.028 ng/mL. The reference level for normal in MSHS is less than 0.03 ng/mL.

Statistical Analysis

Descriptive analyses were performed by troponin levels stratified into normal (0.00-0.03 ng/mL), mildly elevated (between one and three times the upper limit of normal, or >0.03-0.09 ng/mL), and elevated (more than three times the upper limit of normal, or >0.09 ng/mL). Categorical variables were reported as total count and percentage of patients. Continuous non-troponin laboratory values were reported as median and interquartile range. We used troponin measurements within 24 hours of admission. If multiple troponin measurements were available within 24 hours, the patient’s first measurement was used.

To assess the effects of troponin levels on outcomes, we conducted a survival analysis with the dependent variable of time to mortality, setting time zero at time of hospital admission. Patients were considered to be right-censored if they were (1) discharged from the hospital alive or (2) remained in the hospital at the time of data freeze (April 12th). We fit Cox proportional hazards regression models with mortality as the dependent variable, adjusting for age, sex, race, ethnicity, history of CAD, history of AF, history of HF, history of HTN, history of CKD, history of DM, and in-hospital intubation. Age was modeled as age at time of admission, while gender and history of CAD, AF, HTN, CKD, DM, and in-hospital intubation were modeled with binary variables. Self-declared race was included in the model stratified in indicator variables corresponding to Caucasian, African American, Asian, Pacific Islander, Other, or Unknown with Caucasian as reference. Self-declared ethnicity was included in models as Hispanic/Latino, Non-Hispanic/Latino, and Unknown with Hispanic/Latino as the reference level. We tested for proportional hazards assumption by plotting Martingale residuals from the Cox proportional hazards model vs. linearized predictions. 95% confidence intervals from a LOESS best-fit line fit to the Martingale residuals included 0 for all values of the linear predictions, indicating visually there was not significant violation of the proportional hazards assumption. We then plotted Kaplan-Meier curves for survival stratified by troponin group. All analyses were conducted in R version 3.6.1, and visualizations were produced using the survminer R package (7,8).

For a sensitivity analysis, we conducted a secondary survival analysis where discharge from the hospital was considered to be a competing risk since mortality status could not be assessed after hospital discharge. We used the cmprsk R package for this analysis (9). Hazard ratios for the troponin variables from this analysis were not meaningfully different from our standard survival analysis.

Patient Characteristics and Troponin Levels

During the study period, 3,047 COVID-19 positive patients were hospitalized at one of five MSHS New York City hospitals. Of these, 2,736 (89.8%) had at least one troponin-I measurement within 24 hours of admission. The median age was 66.4 years, 40.7% of patients were over age 70, and 59.6% were male. One-quarter of all patients self-identified as African American and 27.6% self-identified as Hispanic or Latino. A history of CVD, including CAD, AF, or HF, was present in 35% of patients. Risk factors for CVD of DM or HTN, were present in 65% of the cohort (Table 1).

Admission troponin-I concentrations are presented in Figure 1. Notably, 1751 (64%) patients had an initial troponin within the normal range. Few patients (86 patients, 3.1%) had an admission troponin elevation over 1 ng/mL within 24 hours of admission, while 173 (6.3%) had a troponin elevation over 1 ng/mL at any point during their hospital stay. Patient characteristics as well as admission vital signs and laboratory measurements, stratified by admission troponin-I, are also displayed in Table 1. Troponin elevations were categorized as mildly elevated and elevated as previously defined. Higher troponin concentrations were seen in patients who were over the age of 70.

• Download figure
• Open in new tab

Figure 1.

Distribution of maximum in-hospital troponin values for all patients with maximum troponin values below 1.0 ng/mL. Patients with troponin concentrations greater than 1.0 ng/mL are not shown.

The proportion of patients with CVD (defined here as CAD, AF, or HF) increased with higher troponin concentrations. Specifically, in those patients with significant myocardial injury (troponin I>0.09ng/mL), CVD including CAD, AF, and HF, was more prevalent (34.9%, 13.0%, and 25.3% respectively) compared to patients with mildly elevated troponins (21.3%, 10.1% and 14.7% respectively) and those with normal troponins (9.8%, 5.2%, and 4.3% respectively). Figure 2 plots troponin measurements within 24 hours of admission among patients with CVD. Individuals with CVD generally presented with higher initial troponins than those without CVD. Similar trends for myocardial injury were seen in patients with history of HTN, DM, and CKD, but not in those with a history of asthma or cancer. Patient characteristics stratified by history of CVD, risk factors, and no history of either are shown in Table 2.

• Download figure
• Open in new tab

Figure 2.

Plot of longitudinal troponin values over time, stratified by history of cardiovascular disease (CAD, HF, AFib) or no history of cardiovascular disease. Smoothing lines fit via LOESS regression with shaded areas indicating 95% confidence intervals.

Inflammatory markers were higher among patients with more substantial troponin elevations as well. In particular, the median D-dimer, C-reactive protein, lactate dehydrogenase, and procalcitonin were higher in patients with elevated initial troponins (2.54 ug/mL, 149.9 mg/L, 520.0 U/L, 0.81 ng/mL) than those with mildly elevated troponins (1.65 ug/mL, 136.78 mg/L, 456.0 U/L, 0.30 ng/mL) and those with normal troponins (1.17 ug/mL, 114.25 mg/L, 425.0 U/L, 0.15 ng/mL). Patients who had lower hemoglobin, hypo- or hypertension, or tachycardia generally presented with higher troponins than those who did not.

Outcomes

Of 2736 COVID-19 patients included in our study, 506 (18.5%) died, 1132 (41.4%) were discharged, and 1098 (40.1%) remained hospitalized at the time of data freeze for this report. Figure 3 presents cumulative incidence plots displaying probability for three possible outcomes (mortality, discharge from hospital, or continued hospitalization) over time. Milder forms of myocardial injury (e.g. troponin concentration 0.03-0.09 ng/mL) were associated with less frequent discharge and higher risk of death than troponin levels in the reference range after adjustment for clinically relevant covariates (adjusted HR: 1.77, 95% CI 1.39-2.26) (Figure 4a). Troponin concentrations over 0.09 ng/dL were associated with more pronounced risk of death (adjusted HR 3.23, 95% CI 2.59-4.02) after adjustment. This risk was consistent across patients stratified by history of CVD, CVD risk factors such as DM or HTN only, and neither CVD nor risk factors. (Figure 4b). A sensitivity analysis using a competing-risks framework demonstrated similar adjusted hazards ratios for risk of death – HR 2.01 (95% CI: 1.59-2.54) for troponin concentrations >0.03-0.09 ng/mL, and HR 3.60 (95% CI: 2.88-4.52) for troponin concentrations >0.09 ng/mL.

• Download figure
• Open in new tab

Figure 3.

Cumulative incidence plots displaying probability for three possible outcomes (mortality, discharge from hospital, or continued hospitalization) over time.

• Download figure
• Open in new tab

Figure 4a.

Kaplan-Meier plot for survival past hospital admission, stratified by troponin grouping. Patients were considered to be right-censored if they were discharged alive from the hospital or were still hospitalized at the time of data freeze (April 12, 2020). Survival times were significantly different between groups (p<0.001).

• Download figure
• Open in new tab

Figure 4b.

Hazard ratios and 95% confidence intervals calculated by Cox proportional hazards regression models for mortality stratified by comorbidities. Patients with cardiovascular disease had comorbidities of coronary artery disease, heart failure, or atrial fibrillation. Patients with cardiovascular risk factors had comorbidities of DM or HTN, but not cardiovascular disease

Limitations

There are some notable limitations of the present analysis. First, there are limitations inherent to the use of EHR for patient level data in such a large sample size not explicitly verified by manual chart review. Despite these limitations, the use of EHR enabled timely analysis and rapid dissemination of crucial information in a large patient cohort at the epicenter of the pandemic. Second, some patients included had not completed their hospital course at the time of data freeze. We accounted for this by conducting a secondary, complementary survival analysis where hospital discharge was treated as a competing risk as outlined in the Methods section. Results from our competing risks analysis were not meaningfully different from a standard survival analysis where discharged patients were simply considered to be right-censored. Lastly our outcomes analyses were focused upon troponin measurements made at hospital admission and less upon serial troponin measurements obtained over the course of each patient’s hospital stay, although we provide plots demonstrating trends in serial troponin measurements.

Clinical Competency in Medical Knowledge 1

Myocardial injury reflected by troponin elevation is common among patients hospitalized with COVID-19, most often present at low levels, and is associated with higher risk of mortality.

Clinical Competency in Medical Knowledge 2

Patients with a history of cardiovascular disease are more likely to have myocardial injury than individuals with risk factors only and those without cardiovascular disease.

Translational Outlook

Elevated troponin concentrations among patients hospitalized with COVID-19 likely represent non-ischemic or secondary myocardial injury.