Epidemiology of myocarditis following COVID-19 or influenza and use of diagnostic assessments

Study design and data source

This was a retrospective comparative cohort study using secondary healthcare data from Optum’s deidentified Clinformatics Data Mart Database (CDM). CDM is derived from a database of administrative health claims for members of large commercial and Medicare Advantage health plans. CDM is statistically deidentified under the Expert Determination method consistent with the Health Insurance Portability and Accountability Act and managed according to Optum customer data use agreements. The population is geographically diverse, spanning all 50 states. At the time of analysis, data were available until 31 March 2023.

Study population

For the COVID-19 cohort (exposure group), patients were included if they had an incident COVID-19 diagnosis (ie, index date) between 1 January 2020 and 31 December 2021. For the influenza cohort (referent group), patients were included if they had an incident influenza diagnosis (ie, index date) between 1 January 2016 and 31 December 2018. Our cohort includes both inpatient and outpatient populations. Different time periods for exposure and referent groups were selected in order to clearly differentiate between influenza and COVID cases. Due to protection measures, for example, social distancing and mask wearing, implemented during the COVID-19 pandemic, influenza rates significantly decreased.14 As such, it was felt that a prepandemic period was more appropriate for the influenza cohort.

Patients were excluded if they met any of the following criteria in the 365 days prior to the index date: diagnosis of myocarditis, diagnosis of COVID-19 (exposure group only) or diagnosis of influenza (referent group only), or had missing age or gender at index date. No patients were excluded from the analysis on the basis of racial, socioeconomic or regional factors.

Two nested cohorts were defined using patients from the main comparative cohorts. The eligibility criterion was a diagnosis of myocarditis within the 365 days after index date in the respective group (exposure or referent). The date of first recorded myocarditis diagnosis was the nested cohort index date.

Each nested cohort described was composed of two subpopulations, adults (patients aged 18 years old and above at index date and paediatrics (patients aged <18 years old at index date).

COVID-19, influenza and myocarditis diagnoses, along with relevant comorbidities, were identified in the claims using relevant International Classification of Diseases (ICD)-10 and ICD-9 codes (online supplemental eTable 1).

Patients in the main comparative cohorts were followed for a maximum of 1 year, disenrolment or death, whichever occurs first. Patients in the nested cohorts were followed until the end of the assessment period, occurrence of outcome, end of data, disenrolment or death, whichever occurs first. No minimum follow-up was required for any cohort.

The study involved no personally identifiable information and the data used in this study were deidentified and anonymised before use. Our study practices were performed in accordance with the Declaration of Helsinki guidelines and followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline. There was no patient or public involvement in this study.

Outcomes

The endpoint for the primary objective was the first recorded occurrence of myocarditis following either COVID-19 or influenza. These conditions were defined based on a recorded diagnosis in inpatient and outpatient records documented according to ICD code (online supplemental eTable 1). Incident myocarditis was considered as a record of myocarditis within 365 days after the index date.

A 12-month follow-up period was employed based on established precedents in the literature, particularly from two large-scale US healthcare database studies on the long-term cardiovascular outcomes of COVID-19,8 15 both of which employed a 12-month follow-up. While we note that a shorter follow-up period may capture a more direct temporal relationship between COVID-19 and myocarditis but would also limit our ability to detect late-onset cases, potentially underestimating the true burden of myocarditis related to COVID-19.

The endpoints for the secondary objectives were the occurrence of the following diagnostic assessments: ECG, echocardiography, blood tests (brain natriuretic peptide, C reactive protein, troponin, sedimentation rate, erythrocyte, myocardial antibody screen), fluorodeoxyglucose-positron emission tomography, invasive coronary angiography, endomyocardial biopsy and CMR (contrast-enhanced and non-contrast enhanced) (online supplemental eTable 2).

Covariates

The following baseline characteristics of patients for all cohorts were assessed at index date or over the 365 days prior to the index date: age, sex, race (white, black, Asian, Hispanic) hypertension, hyperlipidaemia, tobacco use, obesity, history of coronary artery disease (CAD), myocardial infarction (MI), diabetes, (online supplemental eTable 3), Charlson Comorbidity Index (CCI) score (online supplemental eTable 4). Relevant ICD codes were used to define these measures. A full list of relevant codes for the patient characteristics and comorbidities is included in online supplemental file.

Statistical analysis

Rates of myocarditis are reported per 1000 person-years using a Poisson model (figure 1).

Competing risk regression models were used to calculate the association of COVID-19 (exposure group) with incident myocarditis, relative to the influenza population (reference group). The event of death was considered a competing risk. In step 1 of the model-building procedure, the outcome was adjusted for age group and gender (figure 2). In step 2, an extended number of variables was accounted for by including all 11 baseline covariates into the model (figure 2). In step 3, additional sensitivity analyses were conducted, controlling for influenza vaccination status, influenza infection in the COVID-19 group and COVID-19 vaccination status in the COVID-19 group separately (online supplemental eFigures 1–4). Results are reported as subdistribution HRs (SHRs) with 95% CIs and p values.

The secondary endpoint (diagnostic assessments) was described for the nested cohorts by crude and relative numbers of patients in each of the diagnostic assessment categories and is reported as stratified by population. Only the first assessment in each category is considered for this analysis (table 3). The contrast media used was reported for all recorded CMR procedures (online supplemental eFigure 5).

Both endpoints were analysed for the adult subpopulation. Due to the low number of paediatric myocarditis cases, only the crude myocarditis rates were analysed for the paediatric subpopulation.

Patient and public involvement

Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

Myocarditis incidence rates

669 (0.06%) adult COVID-19 patients and 91 (0.02%) adult influenza patients received a diagnosis of myocarditis in the follow-up period while 44 (0.04%) paediatric COVID-19 patients and 11 (0.01%) paediatric influenza patients received a diagnosis of myocarditis in the follow-up period (table 2).

In the adult population, the myocarditis rate per 1000 person-years was 0.73 (95% CI 0.67 to 0.78) for COVID-19 patients and 0.24 (95% CI 0.19 to 0.28) for influenza populations. By age group, the myocarditis rate per 1000 person-years for COVID-19 patients was highest in the group aged 18–29 years (young adults) at 1.16 and highest for the influenza group in the group aged over 70 years (older adults) at 0.42 (figure 1). In the paediatric population, the myocarditis rate per 1000 person-years was 0.53 (95% CI 0.37, 0.68) for COVID-19 patients and 0.06 (95% CI 0.03, 0.10) for the influenza population.

By gender, the myocarditis rate per 1000 person-years was higher for males with COVID-19 at 0.84 than females with COVID-19 at 0.63, while there was no difference in rates for males with influenza at 0.23 and females with influenza at 0.24.

Myocarditis rates by index date (year) showed a slight decrease for COVID-19 between 2020 at 0.82 and 2021 at 0.67, while rates following influenza were highest in 2017 at 0.29 compared with 2016 at 0.22 and 2018 at 0.20, presumably driven by different variants.

SHR model

Step 1: In step 1, the SHR model was limited to exposure (COVID-19 and influenza), sex and age group. In the age-adjusted and sex-adjusted model, COVID-19 infection was associated with almost threefold greater risk of myocarditis (SHR 2.85; 95% CI 2.37, 3.43) compared with influenza infection.

Step 2: In step 2, the SHR model was expanded to include race and clinical comorbidities. The additional covariates did not meaningfully change the results.

Step 3: In step 3, additional sensitivity analyses were conducted including stratification by influenza vaccination status, censoring the COVID-19 population in the event of influenza diagnosis, controlling for COVID-19 vaccination status by excluding patients based on prior COVID-19 vaccination during baseline and censoring on COVID-19 vaccination during follow-up (online supplemental eFigures 1–4).

These additional sensitivity analyses did not alter the direction of the results, and COVID-19 infection was consistently the most strongly associated risk factor for the development of myocarditis.

Diagnostic assessments

The majority of all patients who received a myocarditis diagnosis underwent some diagnostic assessments, with 94% of adult COVID-19 patients, 79% of adult influenza patients, 98% of paediatric COVID-19 patients and 90% of paediatric influenza patients having at least one of the procedures recorded in the period around their myocarditis diagnosis (table 3). To note, categories with small sample counts (N<5) were masked to ensure no reidentification of patients.

The most commonly used diagnostic assessments were ECGs and echocardiography. Blood tests were also common, of which troponin was the most frequently used.

Among patients with myocarditis diagnosis, CMR was conducted in approximately 25% of COVID-19 and 22% of influenza patients and in 41% of paediatric COVID-19 patients and 27% of influenza patients and the most commonly used contrast agent was Gadobutrol (Gadavist, online supplemental eFigure 5).

Limitations

The reasons that led to an individual receiving a diagnosis of influenza in the period 2016–2018 might differ from the reasons that led to an individual receiving a diagnosis of COVID-19 in the period 2020–2021. Widespread testing for COVID-19 was implemented in the period 2020–2021, potentially resulting in individuals with a broader spectrum of disease severity being diagnosed with COVID-19. Conversely, individuals who received a diagnosis of influenza in the period 2016–2018 might have proactively sought out medical treatment for their illness, and therefore, might represent a subgroup with more severe or acute disease. As such, the rates of complications such as myocarditis may be overestimated in the influenza cohort compared with the COVID-19 cohort, in which case the true relative risk of COVID-19-related myocarditis would be even greater than that described here. Alternatively, greater awareness of the link between COVID-19 and myocarditis might have led to a surveillance bias17 and overestimation for the COVID-19 population, although the lower rates of myocarditis following COVID-19 in 2021 compared with 2020 count against this.

While both COVID and influenza vaccination status were ascertained via ICD coding, we recognise that these data may be incomplete and thus the impact of vaccination on our findings may not be fully accounted for. Additionally, while we recognise that laboratory and imaging data to confirm myocarditis diagnosis was not performed in all patients, we note that our study encompasses a large sample size from a national database and thus has the potential to overcome biases that may be present in single-centre studies.

One of the limitations inherent in administrative claims databases is that data are primarily collected for billing purposes and not for clinical research. Coding errors and variability in data quality among different healthcare providers can affect the accuracy of the data and outcomes recorded are often limited to what is billed and might not accurately reflect clinical efficacy or the patient’s quality of life. The population included in the database may also not be representative of the entire population. For example, it might predominantly consist of individuals with commercial insurance or Medicare Advantage, excluding those without insurance or those covered by Medicaid.

In addition, diagnoses are recorded by healthcare professionals who may have used a variety of diagnostic methods. The exact diagnostic criteria employed to make a diagnosis and test results are not detailed within the database making it impossible to verify the accuracy of the diagnosis or to limit the cohort to those confirmed via a specific diagnostic approach. In the current study, limiting the analysis to patients with CMR-confirmed myocarditis or biopsy-proven cases would increase the specificity of the analysis, but this is not feasible given the data limitations. Future studies with access to more detailed clinical data, including CMR and biopsy results, are needed to validate our observations.

A 12-month follow-up was employed, based on precedents observed in the literature8 ,15 . We acknowledge that while some cases may be directly attributable to viral infection, others could result from heightened susceptibility to myocarditis from other causes postinfection. However, the same follow-up period was applied for both COVID-19 and influenza populations and it is not expected that different follow-up periods would alter the proportional difference in myocarditis cases observed between the two populations.