Prediction of vaccine hesitancy based on social media traffic among Israeli parents using machine learning strategies

INTRODUCTION: Vaccines have contributed to substantial reductions of morbidity and mortality from vaccine-preventable diseases, mainly in children. However, vaccine hesitancy was listed by the World Health Organization (WHO) in 2019 as one of the top ten threats to world health. AIM: To employ machine-learning strategies to assess how on-line content regarding vaccination affects vaccine hesitancy. METHODS: We collected social media posts and responses from vaccination discussion groups and forums on leading social platforms, including Facebook and Tapuz (A user content website that contains blogs and forums). We investigated 65,603 records of children aged 0–6 years who are insured in Maccabi’s Health Maintenance Organization (HMO). We applied three machine learning algorithms (Logistic regression, Random forest and Neural networks) to predict vaccination among Israeli children, based on demographic and social media traffic. RESULTS: Higher hesitancy was associated with more social media traffic, for most of the vaccinations. The addition of the social media traffic features improved the performances of most of the models. However, for Rota virus, Hepatitis A and hepatitis B, the performances of all algorithms (with and without the social media features) were close to random (accuracy up to 0.63 and F1 up to 0.65). We found a negative association between on-line discussions and vaccination. CONCLUSIONS: There is an association between social media traffic and vaccine hesitancy. Policy makers are encouraged to perceive social media as a main channel of communication during health crises. Health officials and experts are encouraged to take part in social media discussions, and be equipped to readily provide the information, support and advice that the public is looking for, in order to optimize vaccination actions and to improve public health SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s13584-021-00486-6.