This repository hosts the dataset employed for the ‘A behaviourally-informed chatbot increases vaccination rates in Argentina more than a one-way reminder’ paper’s analysis. The paper discusses an intervention involving a new chatbot service to explore if two-way interactive messaging outperforms one-way reminders. We conducted a pre-registered randomised controlled trial involving 249,705 participants in Argentina and recorded vaccinations using Ministry of Health data.

The repository includes two data files:

‘brown2024_analysis_data.rds’ features individual-level data of all trial participants, including basic demographic data, Covid-19 vaccination history, assignment to an experimental arm, a summary of chatbot interaction, and outcome measures: binary variables indicating whether the participant received their next Covid-19 vaccine dose within four weeks.

The data has been de-identified. Information on age has been coded into bins or ranges (18-29, 30-49, 50+) and information on gender has been set as a binary variable (0, 1).

‘20230505_vaccinated_per_day.csv’ depicts the demand for COVID-19 vaccines in the Chaco province from February 2021 to December 2022. It records the daily total of COVID-19 vaccine doses delivered during this period in Chaco. These data were sourced from Nomivac, a comprehensive dataset of all COVID-19 vaccinations nationwide.

Description of the data and file structure

1. File “brown2024_analysis_data.rds”

• randomisation_allocation_name: Name of treatment group
• current_dose: Number of doses of COVID-19 vaccine received
• next_dose: Number of next dose of COVID-19 to be received
• random_number_date: Date of message delivery, assigned at random to individuals in the control group. See materials and methods for details
• outcome_latest_date: Date of most recent dose of COVID-19 vaccine received
• dose_X_date: Date of X dose of COVID-19 vaccination
• vaccinated_message_sent: Whether a message was successfully delivered to the individual
• session_current_level: Level of the message flow where the user stopped interacting with the chatbot
• session_current_branch: Branch within the level of the message flow where the user stopped interacting with the chatbot
• session_selected_hc: Option number chosen by the user when asked to select a health centre, out of a list of a maximum of 9 options
• session_reminder_date: The date for which the reminder message was programmed
• session_reminder_time: Time of day chosen by the user to get vaccinated
• session_initial_response: Users’ answer to the initial message sent by the chatbot
• session_reminder_sent: Whether a reminder message was successfully sent to the user
• history_message_code: Type of initial message sent to the user (“control” is the one-way message; “inicial” is the chatbot’s first message)
• history_created: Timestamp of the first message successfully delivered to the user
• history_timestamp_read: Timestamp of the first message read by the user
• message_date: This variable is used to indicate the date that each individual entered the trial. For individuals in the chatbot and one-way message reminder groups, it is the date that the message was sent. For the control group, this is the date that we randomly pre-assigned each participant to enter the trial (random_number_date). See materials and methods for details
• had_new_dose: Participant received an additional dose of the COVID-19 vaccine after being randomly assigned to the treatment arm
• had_new_dose_after_message: Participant received a vaccine within 4 weeks of entering the trial
• had_new_dose_after_message_2w: Participant received a vaccine within 2 weeks of entering the trial
• had_new_dose_after_message_6w: Participant received a vaccine within 6 weeks of entering the trial
• n_doses_after_entering_trial: Number of doses within four weeks of entering the trial
• primary_outcome: had_new_dose as a factor variable
• age: Age range of participant
• gender_numeric: Gender of the participant, coded as 1 or 0 for de-identification purposes
• period_since_last_dose: Length of time between the first value of message_date in the whole sample and the last dose date before trial

2. File “20230505_vaccinated_per_day.csv”

• delivery_date: Date of vaccination
• count: Total number of residents of Chaco province who received a COVID-19 dose on a given day