(292) Tree-Based Data-Mining for COVID-19 Vaccine Safety Assessment

Publication Number: 1337

Background: Except for spontaneous reporting systems, vaccine safety monitoring generally involves pre-specifying health outcomes and post-vaccination risk windows of concern. In contrast, we used data-mining with TreeScan software to look broadly for possible adverse events after COVID-19 vaccination.

Objective: To detect possible adverse events after COVID-19 vaccination.

Methods: Enrollees in the Vaccine Safety Datalink’s 8 sites receiving ≥ 1 dose of COVID-19 vaccine in 2021 were followed for 70 days after Dose 1 of Pfizer or Moderna and 56 days after Janssen vaccination. Incident diagnoses in inpatient or emergency department settings were analyzed for clustering within both the hierarchical ICD-10 code structure and the post-vaccination follow-up period. We used the self-controlled tree-temporal scan statistic, conditioning on the total number of cases for each diagnosis. Monte Carlo simulation was used to obtain statistical significance, and P=0.01 was the pre-specified cut-off for statistical significance.

Results: There were more than 4, 2.5, and 0.4 million Pfizer, Moderna, and Janssen vaccinees, respectively. Outcomes with statistically significant clusters after Pfizer vaccination fell into 5 groups. For 4 of these outcome groups, the strongest clusters were mostly after Dose 2: 1) unspecified adverse effects; 2) common systemic or local reactions to vaccination; 3) myo/pericarditis; and 4) less specific cardiac or respiratory symptoms, including chest pain. The fifth group was viral pneumonia, sepsis, or respiratory failure, all in the first 3 weeks after Dose 1. Results for Moderna were similar but without a significant myo/pericarditis cluster. Further investigation indicated the pneumonia/sepsis signal resulted from SARS-CoV-2 infection before the vaccine conferred protection. Among the Janssen vaccinees were clusters of unspecified or common vaccine reactions, gait/mobility abnormalities, and muscle weakness. The latter two were deemed false signals arising from practices at one large site.

Conclusions: We detected clusters of unspecified adverse effects; common vaccine-associated adverse events, such as fever, myalgia, and headache; and, for the mRNA vaccines, chest pain and palpitations, as well as myo/pericarditis after Pfizer Dose 2. We also saw mRNA vaccine effectiveness, manifested as clustering of COVID-19-like cases within 21 days after Dose 1. The untargeted nature of the method and its inherent adjustment for the many diagnoses and risk intervals evaluated are unique advantages; limitations include the possibility of missing adverse events that do not cluster strongly in time during follow-up.