Graduate Student College of Pharmacy, Chungbuk National University, Cheongju, Republic of Korea Cheogjusi, Republic of Korea
Background: COVID-19 vaccination provided substantial protection against serious illness, hospitalization, and death. With the increasing vaccination rate, the number of reports of side effects also increased. Among side effects, carditis cases, including myocarditis and pericarditis, were reported to increase. However, controversy over the correlation between vaccination and carditis and the risk factors of this event is not well-known.
Objectives: This study aimed to discover the risk factors of carditis incidence after the COVID-19 vaccination by developing a prediction model.
Methods: This study investigated the correlation between COVID-19 vaccination and carditis using National Health Insurance Service data from 2008 to 2021 in two study designs (case-crossover and nested case-control). The case group in the case-crossover study was defined as patients with the first diagnosis of carditis within the vaccination date range, and logistic regression was conducted to validate the correlation between vaccination and carditis. In the nested case-control study, patients diagnosed with carditis were defined as the case group, and propensity score matching was used for the control group. The study utilized ICD-10 codes and ingredient codes from Korea to engineer features for every disease and medication. Machine learning-based feature selection was used to increase the model's explainability, and the best algorithms and features were selected using TPOT and Boruta with stepwise logistic regression. Data processing, analysis, and machine learning procedures were conducted using SAS Enterprise Guide 8.3 and Python 3.9.
Results: In the case-crossover study, 4,541 patients were selected as the case group and had an adjusted odds ratio of 20.80 (95% CI: 18.59 – 23.27, p-value < 0.001) for vaccination during the hazard period. The nested-case control study selected 3,215 patients as both case and control groups. The most important factor in predicting carditis was the time length between vaccination and diagnosis (aOR: 0.233, 95% CI: 0.193 – 0.282). Other factors that increased the risk of carditis included inflammation in other body parts, gastro-esophageal reflux disease, angina pectoris, cardiac arrhythmias, anxiety disorder, and certain medications. Female gender and the ChAdOx1-S vaccine injection decreased the risk of carditis.
Conclusions: In our data, the COVID-19 vaccination showed a high correlation with carditis. We discovered several factors that might affect the incidence of carditis by screening all the possible features and selecting 17 features that might effectively predict carditis incidence after the COVID-19 vaccination.
