Boost Intentions and Facilitate Action to Promote COVID-19 Booster Take-up

Shortly after COVID-19 bivalent boosters were authorized, UCLA Health emailed its patients encouraging them to get the covid booster either at UCLA Health or at local pharmacies. The investigators plan to conduct a randomized controlled trial, several weeks after this initial communication, to test how text-based reminders that boost intentions and/or facilitate action affect the take-up of COVID-19 bivalent boosters. Among all UCLA Health patients eligible to receive the COVID-19 bivalent booster, the investigators will randomly select 35% for this trial, 40% for another trial (Pre-registration title: "Information Provision and Consistency Framing to Increase COVID-19 Booster Uptake") and 25% for a third trial (Pre-registration title: "Effects of Prompt to Bundle Covid-19 Booster and Flu Shot"). The three trials will be run simultaneously.

Patients randomly selected for this trial will be randomized at an equal share to 6 conditions: an Holdout condition that does not receive a text reminder and 5 additional conditions that receive a text message (Facilitate Action Narrowly, Facilitate Action Broadly, Boost Intention, Boost Intentions+Facilitate Action Narrowly, Boost Intentions+Facilitate Action Broadly; see description in the Arm/Interventions section). Within each arm, patients are further randomized into one of the 33 time slots for when they will receive the text message (i.e., 3 time points per day including 9am, 12pm, 4pm for 11 days). Days 1-4 and 8-11 have slightly fewer patients than Days 5-7.

To test whether reminders that facilitate action by providing appointment scheduling links increase vaccine uptake, the investigators will compare the holdout condition to the combination of the 4 arms that contain appointment scheduling links.

With this trial, the investigators will investigate several additional research questions.

1. The investigators will investigate the value of simultaneously elevating intentions and facilitating action. The investigators will compare vaccine uptake between the holdout arm and (1) the combination of the two action-only messages (i.e., the Facilitate Action Narrowly arm and the Facilitate Action Broadly arm), (2) the intention-only message (i.e., the Boost Intention arm), and (3) the combination of the two intention-action messages (i.e., the Boost Intentions+Facilitate Action Narrowly arm and the Boost Intentions+Facilitate Action Broadly arm). Then the investigators will compare vaccine uptake between (1) the combination of the two action-only messages and (2) the intention-only message versus (3) the combination of the two intention-action messages.the investigators will also compare link click rates between the combination of the two action-only messages and the combination of the two intention-action messages.
2. The investigators will investigate whether facilitating action narrowly or broadly is better at promoting vaccine uptake. For that purpose, the investigators will compare vaccine uptake and link click rates between the combination of the two messages facilitating action narrowly (i.e., the Facilitate Action Narrowly arm and the Boost Intentions+Facilitate Action Narrowly arm) and the combination of the two messages facilitating action broadly (i.e., the Facilitate Action Broadly arm and the Boost Intentions+Facilitate Action Broadly arm). The investigators will also compare vaccine uptake between these two sets of messages and the holdout condition.
3. The investigators will explore how the effects of text reminders encouraging people to get vaccinated at specific (narrow) locations vary by the scope and horizon of vaccine uptake measure. For this purpose, the investigators will compare the combination of the two messages containing links to UCLA/CVS with the holdout condition, and the investigators will analyze both vaccine uptake at UCLA/CVS and vaccine uptake in California and both at 2 weeks and 4 weeks. The investigators will do the same analyses for the additive effect of boosting intentions (vs. only having the links to UCLA/CVS)..
4. The investigators will explore the role of the time when a text message is sent. The investigators will examine how the effect of text reminders facilitating action (i.e., the combination of the 4 arms with appointment scheduling links vs. holdout) varies across days and times. As additional specifications, the investigators will control for the time trend and analyze all patients who have not gotten the bivalent booster in California or scheduled a booster appointment at UCLA Health the day before the trial starts.
5. Given that the investigators are simultaneously conducting three trials in the same population, the investigators may combine data from three trials and compare vaccine uptake between the reminder arms and the average of the holdout arms from the three trials to identify the most effective reminder in a mega-study fashion by following the approach in Milkman et al. (2021).

Analysis:

The investigators will run ordinary least squares regressions (OLS) with robust standard errors to predict the aforementioned outcome variables, except that the investigators will use a Cox proportional hazards model with administrative censoring to predict time of obtaining the booster. The significance level will be 0.05. The regressions will include the following control variables:

• Indicators for one of the 33 time slots
• Participant age (If there will be a missing value, the investigators will replace with the mean and add a dummy variable to indicate patients with missing age)
• Indicators for participant race/ethnicity (Black non-Hispanic, Hispanic, Asian non-Hispanic, white non-Hispanic, other/mixed, unknown)
• Indicators for participant gender (male, female, other/unknown)
• The investigators will test whether the main effect of sending text reminders to facilitate action (i.e., the combination of the 4 reminder arms with appointment scheduling links vs. holdout) is robust regardless of:
• Whether the patient is female or male
• Whether the patient is white non-hispanic or racial/ethnical minority
• Whether the patient is 65+ (including 65) or below 65

The investigators will investigate the following moderators for (1) the effect of sending text reminders to facilitate action, (2) the additive effect of adding appointment scheduling links (vs. only boosting intentions), and (3) the additive effect of boosting intentions (vs. only facilitating action with appointment scheduling links):

• Timing of the first dose of the COVID-19 vaccine primary series
• Whether patients got at least one original booster prior to the trial
• Whether or not the patient received a flu shot in either the 2020-2021 season or the 2021-2022 flu season according to the patient's medical record
• How strongly the patient's neighborhood is in favor of the Republican (vs. Democratic) Party if UCLA Health eventually agrees to provide de-identified addresses (e.g., zip code) and if the sample contains patients living in Republican-leaning counties.

Additional information: Three days into this trial, the investigators learned that the supply of boosters at UCLA Health is very limited and patients could not find appointments at UCLA Health. The investigators verified that local pharmacies continue to have appointments available. Thus, the investigators updated all text messages to (1) explain that UCLA Health has limited booster appointments available, (2) no longer provide a direct link to schedule appointments at UCLA Health, and (3) instead encourage patients to make an appointment at either CVS Pharmacy or local pharmacies (depending on the arm). As a result, instead of analyzing the originally preregistered outcome measures of "booster take-up at UCLA/CVS in 2 weeks" and "booster take-up at UCLA/CVS in 4 weeks", the investigators will measure vaccine uptake at CVS in 2 weeks and 4 weeks. In terms of the outcome measure of "Link click rate in 1 week", the investigators will still look at whether patients click on any link in the text message.