(350) Predicting Recovery Status of COVID-19 Vaccinated Patients with Natural Language Processing Approaches

Background: The brief period between the declaration of the COVID-19 pandemic and the authorization of vaccines raised public concerns about the safety and effectiveness of the vaccines. The recovery status of vaccinated people from side effects became a crucial focal point, prompting intensified scrutiny and a heightened awareness of the need for transparent communication regarding vaccine safety.

Objectives: This study aims to establish effective natural language processing (NLP) models for predicting the recovery status of vaccinated people by combining structured and unstructured data.

Methods: COVID-19 vaccinated patients with clear answers to their recovery status were included in this study. Patients’ descriptions of their symptoms were processed by text mining and employed for prediction. Age, sex, doctor visit, and emergency room visit were used to predict their recovery status as well. Text mining technique was used for feature extraction to improve the performance of predictive models. This study also employed the synthetic minority over-sampling technique (SMOTE) to balance the data. .

Results: A total of 32,810 patients were included, with 66.82% of them recovering from COVID-19. Compared to not recovered patients, recovered individuals were more likely to be young, male, and have fewer doctor visits and emergency room visits (all P < 0.05). Performance accuracy between 0.82 and 0.88. Age, sex and doctor visits were among the top 10 most important predictors. Other top 10 most important predictors included the keywords “bivalent”, “given”, “administer”, “paxlovid”, “last”, “adverse”, and “still”.

Conclusions: This research establishes effective NLP models to predict the recovery status of individuals who have received COVID-19 vaccinations, and identifying key factors and keywords for prediction. The findings of this study are anticipated to offer valuable insights, aiding healthcare professionals in formulating effective strategies to address and manage COVID-19.