Sex Differences in Sentiment-Scored Vocabulary Patterns in VAERS Free-Text Narratives of Four Major Vaccine...

Free-text narratives in vaccine adverse event reports contain symptom descriptions and contextual wording that can be characterized using text mining. However, lexicon-based sentiment scores in medical narratives should be interpreted as vocabulary patterns rather than direct measures of patients' emotional states. We evaluated sex-specific and vaccine-specific patterns in sentiment-scored and emotion-category assignment vocabulary in Vaccine Adverse Event Reporting System (VAERS) narratives for coronavirus disease 2019 (COVID-19), influenza, human papillomavirus (HPV), and measles-mumps-rubella (MMR) vaccines. Public VAERS reports received from 2010 to 2024 were analyzed. VAERSDATA, VAERSVAX, and VAERSSYMPTOMS were linked by VAERS_ID, with duplicate records removed. Report-level AFINN scores were normalized per 100 words. Sensitivity analyses excluded common clinical and symptom-related terms. Effect sizes, 95% confidence intervals, adjusted linear regression models with HC3 robust standard errors, and NRC emotion-category assignment profiles were evaluated. After duplicate handling, 1022460 COVID-19, 148489 influenza, 41212 HPV, and 47459 MMR reports were analyzed. Female-associated reports generally showed slightly lower primary AFINN scores than male-associated reports, although effect sizes were negligible. Excluding clinical and symptom-related terms attenuated or reversed these differences. Adjusted regression models showed only small associations between sex and AFINN scores. NRC analyses demonstrated clearer vaccine-specific than sex-specific vocabulary patterns. VAERS narratives showed only small sex-related differences but distinct vaccine-specific patterns in sentiment-scored and emotion-category assignment vocabulary. These findings should be interpreted as lexicon-derived vocabulary patterns rather than direct measures of psychological responses, vaccine perceptions, or clinical outcomes.