Nudging parents to increase HPV vaccine demand in Bogota, Colombia

We implemented two sets of experiments that were designed to increase vaccination rates among unvaccinated girls and girls with incomplete HPV vaccination schedules. All these experiments were based on SMS campaigns. Text messages were sent to parents of our target population using alternative behavioral insights based on social norms, beliefs, emotions, and decision aids.

For the first group, 4 experiments were implemented. The first experiment explored the role of social norms, a second experiment exploited the role of beliefs, a third experiment introduce variation in emotions, and a fourth experiment considered the use of decision aids.

The social norm experiment exploited alternative ways to communicate the norms regarding HPV vaccination in our setting. This includes a static design (descriptive and injunctive norms) and a dynamic one (dynamic and trending norms). The Appendix in the Pre-Analysis Plan (PAP) describes in detail the messages delivered as part of this intervention. As an example, a subset of parents in this experiment received an SMS with a descriptive social norm of the following form:
"Hello NAME. You are part of the 30% of parents in your neighborhood who expose their daughters to cancer by not vaccinating them against HPV. Health Secretariat." All messages include fixed elements across treatment arms that were proven to be effective in other settings. In this case, we added the name of the parent and the information of the sender in all the messages in this experiment.

The second experiment seeks to correct a set of beliefs regarding HPV. Messages about HPV (likelihood and severity), HPV vaccine (effectiveness, safety, and cost), government and health provider support were delivered to a sub-sample of the experimental population. Messages were crafted following the same principles described above.

The third experiment explored the role of emotions. Messages emphasizing anticipated regret, anticipated worry, and soft-shame were crafted to boost vaccination rates.

The fourth experiment designed tools to induce a direct change in behavior, including an appeal to prosocial concerns. Among the instruments to induce a direct change in behavior, messages based on the use of soft-defaults enhanced active choice, and pseudo set were used. We also explore the role of altruism.

The second set of experiments was targeted at girls with incomplete vaccination schedules. The first experiment was based on social norms using similar static and dynamic designs. A minor variation was introduced to incorporate a comparison between qualitative and quantitative dynamic norms. The second experiment introduced a set of tools to close the intention-action gap. It was presumed that parents of girls with incomplete vaccination schedules already expressed their desire of vaccinating their daughters, but failed to complete the process. Reminders, presumptive announcements, priming, and planning tools were used in this case. For each of these tools, variations were introduced to complement these messages. These variations were designed to elicit implementation intentions or to introduce an anchor. In the case of priming, question-behavior effects and mere-measurement effects were used.

All the treatments in the two sets of experiments were cross-randomized with planning tools. These planning tools offer a link or a telephone number where parents can have access to information about the closest vaccination point and other related information about HPV vaccination.

The experiments include 3 control groups. The Pure Control Group does not receive any message. The Experimental Control Group receives placebo messages that include the fixed elements described above. Finally, the Policy Control Group receives the "business as usual" message that the Health Secretariat uses in communications with the target population. These messages are described in the Appendix to the PAP.

2021-10-21

2022-03-31

The primary outcome is a binary measure of whether a parent’s daughter is vaccinated against HPV during the SMS campaign window (8 weeks) and/or during the 8 weeks after the end of the intervention.

This outcome will be constructed from administrative records. The Health Secretariat manages, collects, and organizes all the vaccinations records in both the contributed and subsidized health system.

None.

All the interventions described above were randomized at the individual level using stratified individual cross-randomized experimental designs. Each experiment was stratified by locality and vaccinee's age.

All the experiments were implemented simultaneously.

Randomization by a computer using the rerandomization algorithm proposed by Morgan and Rubin (2012).

Vaccinee.

No

None.

The target sample for the experiments with unvaccinated girls is 131,124 observations. The target sample for the experiments with girls with incomplete vaccination schedules is 43,057.

The sample size by treatment arm is 4,300 observations for the experiments with unvaccinated girls.

The sample size by treatment arm is 2,050 observations for the experiments with girls with incomplete vaccination schedules.

Power calculations were implemented using an effect size of 3 percentage points change in the vaccination rate for an individual randomized design. This was implemented considering a test for differences in proportions (Chi2 test), assuming 90% power, and accounting for multiple using a Bonferroni correction (allowing up to 17 comparisons in a given experiment). A minimum sample size of 13,578 per experiment was estimated using the above parameters. All the implemented experiments had larger sample sizes than this minimum benchmark.

PAP for Nudging parents to increase HPV vaccine demand