Vaccine Chatbot for Improving Influenza Vaccination Uptake

Overview

This study aims to assess the impact of a vaccine chatbot on improving influenza vaccination uptake among children aged between 6 and 59 months through a cluster randomized trial. Specifically, the main questions it seeks to answer are whether an AI-enabled vaccine chatbot will increase the uptake of influenza vaccine among children and their family members, and how it will influence parents' literacy and confidence towards influenza vaccine. It will explore the potential role of vaccine chatbot on vaccination services. A cluster randomization will be used to assign children to the intervention and control groups. Parents of children in the intervention group will be invited to use the influenza vaccine chatbot online through WeChat, the mostly widely used social media platform in mainland China, or any web browsers. They can ask any questions related to the influenza vaccine and receive validated answers from the chatbot immediately. The intervention will last one and a half months, with invitations sent every ten days to reinforce the engagement. The control group will not use the chatbot during the intervention duration. After the intervention, the uptake, literacy, and confidence towards influenza vaccine will be compared between the intervention and control groups to evaluate the impact of vaccine chatbot.

This is a cluster randomized trial (CRTs) consisting of two arms to evaluate the effectiveness of an AI-enabled vaccine chatbot on influenza vaccination uptake among children aged between 6 and 59 months. Participants regularly visit primary care clinics when they are invited to participate in this trial, and with clinic-days as clusters, a cluster randomization will be used to assign clinic days to the intervention and control groups. The sample size is calculated based on the primary outcome - the influenza vaccination uptake among children, and the main analysis method, which involves the comparison of differences in vaccination rates between the intervention and control groups after the intervention. According to the vaccination data of study sites, the uptake of influenza vaccine among children aged 6-59 months is around 20% in the previous flu season. In China, influenza vaccination starts in September, and has been available for two months before this trial. Therefore, we assume that the baseline vaccination rate is 10% without the intervention during this trial, and the chatbot intervention would raise this rate by 6 percentage points to 16% at least. We assume a cluster size of 15 children per day per clinic based on routine visiting data. To have 80% power to detect a difference between the group proportions of 0.06, it requires 35 clusters and 525 participants per arm, assuming an intracluster correlation coefficient of 0.005 and a two-sided test with the 0.05 significance level. Assuming at least 10% loss to follow-up, the sample size is 600 participants per arm and 1,200 in total. Multi-stage sampling will be utilized. Firstly, three representative regions (an urban district, a suburban district, and a rural county) will be selected to represent different economic development levels in China. In each region, four clinics will be selected based on geographical location, economic development, and patient volume. In each clinic, the 6-8 working days will be chosen to conduct this trial, resulting in 72-96 clinic-days (clusters) in 12 clinics in total. In these selected clinic-days, children and their parents regularly visit primary care clinics. All eligible children presenting in the selected clinic-days will be invited to participate by medical staff, and one of their parents will be included in this trial. Stratified cluster randomized grouping will be employed. All clinic-days will be randomly allocated into intervention group or control group at a 1:1 ratio, stratified by region and clinic, resulting with 36-48 clinic-days per arm. Approximately 400 participants (200 in intervention group; 200 in control group) are expected to participate in this trial in each region, with a total sample size of 1,200 participants, meeting the sample size requirement. The intervention group will engage with influenza vaccine chatbot for one and a half months, while the control group will receive the standard of care according to the local context, without additional intervention. On the day of the visit, the intervention group will be invited to use the influenza vaccine chatbot online through WeChat or any web browsers, where they can ask any questions related to the influenza vaccine and get validated answers from the chatbot immediately. Staff will be on site to help them use vaccine chatbot for the intervention group. Then, during the 1.5-month intervention, participants in the intervention group will be informed that the chatbot is available for use at their convenience, with coordinators sending the chatbot link every ten days to remind them to use. Conversely, the control group will not use the chatbot, without additional intervention during the trial, but will gain access after the trial ends. At the end of the 1.5-month intervention, all participants from both intervention and control groups will be invited to complete a questionary survey. Three months after the intervention begins, influenza vaccination status of children and their parents will be collected from the vaccination registration system of local CDCs. Difference in outcomes between the intervention and control groups will be assessed using t-tests and/or analysis of variance (ANOVA) for normally distributed continuous variables, and the Chi-square or Fisher's exact test will be used for categorical variables. When continuous variables do not meet normal distribution, the Wilcoxon rank-sum test will be employed. Multivariate regression models will be employed to evaluate the effectiveness of the chatbot intervention on the primary and secondary outcomes, adjusting for potential confounders. Given that participants in the intervention group will have varying frequencies and durations of using the chatbot, a dose-response relationship will be employed to evaluate the intervention effects by intervention intensity.