Attractive Flu Shot: A Behavioral Approach to Increasing Influenza Vaccination Uptake Rates

We will run an online hypothetical survey that will include an invitation to receive the flu shot that participants will be asked to imagine they will receive on October 1st by their HMO. The survey will be sent out to a sample of the adult Israeli population (N=3,250). Participants will be randomly assigned into either the control or one of the treatment groups. In our treatment groups (the intervention groups) participants will be introduced with two options to receive the shot: early in the season, which is recommended by the ministry of health and hence ``attractive,'' or later. Three additional layers along which the early shot outperforms the late one will be examined across treatments. In the control, participants will simply be invited to receive the shot regardless of timing.

We will compare vaccination intentions (using the self report of the participants in the survey) in these treatments to those of the control group.

2019-08-25

2019-09-30

Our main dependent variable is the self reported intentions of getting the vaccine or not (in the treatment groups the report will be getting the shot early, late or not at all). Our independent variables will be the treatment to which the participant belongs (we will have random assignment to the different treatments) and a measure of the level of certainty of the participant with respect to his/her intentions of getting the vaccine. As controls we will look at vaccination last year, number of vaccinations in the past 5 years and we will also have demographic information such as age, gender, self reported income and education which is kept in the survey company's data set.

The main variables of interest are straightforward. Subjects will report whether or not they plan to get the shot (in the treatment groups the report will be getting the shot early, late or not at all). The measure of level of certainty will be determined by the question asking how certain the participant feels regarding getting the flu shot which will be answered on a 5-point Likert scale.

The design is mostly described in the intervention section above. The experiment is based on a survey which will be distributed to a sample (N=3,250) of the Israeli adult population through a panel company which has a large number of registered panelists. In the survey participants will be asked to imagine that on October 1st they will receive a message from their HMO inviting them to receive the seasonal influenza vaccine. Invitations in all groups will inform the reader about the flu and the vaccine, mention that receiving the shot is recommended by the ministry of health and state that it is available through the end of March the following year, the date until which flu shots are normally offered in Israel.

In the control, the invitation message will be along the lines of the one used in Israel these days in which only the basic information about the flu and the vaccine is provided. Following this information, we will ask participants in the control group whether they plan on taking the shot (which is free of charge) or not. In the four treatments (intervention groups), following the basic information, two options to receive the vaccination will be introduced: an early option, which is recommended by the ministry of health, and a late option. Three of the four treatments will add another dimension along which the early shot is better than the late one.

Following the invitation in all treatment groups, we will ask our participants to mark whether they plan on getting the vaccine early, late or not at all (this is our main dependent variable). In the control they will simply be asked whether they plan on getting the shot or not.

Following these questions we will ask participants for a free text explanation of their answer and their level of certainty with respect to their intention to get the shot. Finally we will ask them whether or not they received the shot last year and to give their best answer to the question of how many times they received the flu shot in the last 5 winter seasons (a multiple choice question).

The design is mostly described in the intervention section above. The experiment is based on a survey which will be distributed to a representative sample (N=3,250) of the Israeli adult population through a panel company which has a large number of registered panelists. In the survey, participants will be asked to imagine that on October 1st they will receive a message from their HMO inviting them to receive the seasonal influenza vaccine. Invitations in all groups will inform the reader about the flu and the vaccine, mention that receiving the shot is recommended by the ministry of health and state that it is available through the end of March the following year, the date until which flu shots are normally offered in Israel.

In the control, the invitation message will be along the lines of the one used in Israel these days in which only the basic information about the flu and the vaccine is provided. Following this information, we will ask participants in the control group whether they plan on taking the shot (which is free of charge) or not. In the four treatments (intervention groups), following the basic information, two options to receive the vaccination will be introduced: an early option, which is recommended by the ministry of health, and a late option. Three of the four treatments will add another dimension along which the early shot is better than the late one (therefore we will have 5 treatments overall: the control and four intervention treatments). Here are the main dimensions along which the early shot is better than the late one according to the different treatments:

1. The early shot is recommended by the ministry of health.

2. Mentioning that the vaccine is more likely to run out of stock later in the season compared to early in the season.

3. The early shot is free while the late one costs a fee of 20 ILS (roughly 5.5 USD).

4. The early shot carries a monetary benefit of 20 ILS which may be used for doctor visits within the HMO.

Following the invitation in all treatment groups, we will ask our participants to mark whether they plan on getting the vaccine early, late or not at all (this is our main dependent variable).

Following the first question in each treatment (getting the shot early/late or not) we will ask participants 4 more questions:

2. Freely explain why they chose what they did in the first question.
3. How certain they are (on a scale of 1 - 5) regarding their willingness to get the shot.
4. Whether or not they received the vaccine last year.
5. How many times they received the flu shot in the last 5 winter seasons (a multiple choice question which may be answered according to their memory if they are not sure).

A detailed explanation of our hypotheses, planned analysis and rule for number of participants (which is based on a pilot study and the pwr package in R) are given in the data analysis document attached to this pre-registration. All surveys are also attached to this pre-registration (in one document).

Randomization by computer.

Individual randomization into treatments.

No

3,250

3,250

650 subjects per treatment. 650 in the control and 650 in each intervention treatment.

The power calculation has been based on a pilot study which was run using 1,151 participants during 2016 and 2017. The calculation was run using the pwr package in R and took into consideration the average effect size found in the pilot study and alpha=0.5. For more information see the planned analysis document.

Plan for Analysis - Attractive Flu Shot