COVID-19 Vaccination Take-Up in a County-Run Medicaid Managed Care Population

We will test the impact of financial incentives, the highlighting of a convenient scheduling link, messaging from race and gender concordant vs. discordant providers and message type on COVID-19 vaccine uptake.

2021-05-24

2021-09-20

Vaccine uptake at 1-month
Vaccine intentions at time of survey

Vaccine uptake at 1-month is measured from a state registry that feeds into our partner's data system.
Vaccine intentions are measured in our survey

Vaccine take-up at 6 months; vaccine take-up at 1-year.

Difference between vaccine intention and vaccination

Time to first vaccination

Full vaccination status

We will also study heterogeneous treatment effects by race, gender, age and other characteristics.

From administrative data, we may study health care utilization and COVID-19 testing.

We will test the impact of financial incentives, messages from race and gender concordant vs. discordant providers, as well as message type on COVID-19 vaccine uptake.

We aim to recruit 10,000 adult subjects from the Contra Costa Health Plan (CCHP) via a baseline survey. Subjects who complete the baseline survey will be randomized to the following arms:
1. Control Arm (n=2,500)
2. Informational Arm: state/treatment as usual message vs. safety and effectiveness information vs. information on consequences of going unvaccinated, race and/or gender concordant or discordant [7,500]

Each of these arms will be interacted with a financial incentive of $10 (N=2,500) or $50 (N=2,500) and, separately with a convenient link to the county public vaccine appointment scheduling system highlighted for participants (N=5,000).

The above treatments are designed to test the role of the following on vaccine take-up:

• Financial incentives [N=5,000] vs. no financial incentives [N=5,000]
o 2,500 will be randomized to a $10 incentive and 2,500 to a $50 incentive
• Convenient scheduling link highlighted [N=5,000] vs. not [N=5,000]
• Messaging [N=7,500] vs not [2,500]
o Message type: treatment as usual [N=2,500] vs. safety and effectiveness [N=2,500] vs. consequences of not vaccinating [N=2,500]
• Race concordant [N=2,500] vs. race discordant messenger [N=2,500]
• Gender concordant [N=2,500] vs. gender discordant messenger [N=2,500]

We will obtain survey data on preventative health behaviors, including mask-wearing, hand washing, and willingness to vaccinate. We will obtain data on vaccine take-up from the California's Immunization Registry (CAIR) and the EMR.

computer

members of Contra Costa Health Plan

No

The goal is to recruit 10,000 unvaccinated members of Contra Costa Health Plan

10,000

We will randomize the 10,000 respondents to one of four arms:

1. Control arm [2,500]
2. Messaging/Information Arm 1: no information/emotional message [N=2,500]
3. Messaging/Information Arm 2: provider safety and effectiveness information [N=2,500]
4. Messaging/Information Arm 3: information on consequences of going unvaccinated [N=2,500].

Each of these four arms will be interacted with a financial incentive of $10 (N=2,500) or $50 (N=2,500) and, separately with a convenient link to the county public vaccine appointment scheduling system highlighted for participants (N=5,000). Provider messages will also be randomized by race and gender concordance.

We power our study for our primary outcome, vaccine take-up. For the purposes of this power calculation, we assume that among the vaccine hesitant only 10% of the population will get vaccinated in the absence if our interventions. In unadjusted comparisons, we will be able to detect a change in vaccinations of 1.68 percentage points (16.8% off the mean) for our financial incentive (5,000) vs. control (5,000) or scheduling link vs. control. For any message (7,500) vs. no message (N=2,500) we can detect take-up changes of 1.94 percentage points (19.4% off the mean). For our 3-message type or race or gender concordance (N=2,500 each) comparisons we can detect take-up changes of 2.38 percentage points (23.8% off the mean). Accounting for statistical controls as well as our randomization strata and assuming these increase the R-squared on take-up to 0.25, the MDEs decline to 0.146 percentage points for financial incentives/scheduling link, 1.68 percentage points for any message and 2.1 percentage points for message type or race or gender concordance comparisons. To benchmark these comparisons, we note that Alsan et al. (2019) finds increases in flu shot take-up among African American men, a vaccine hesitant group, of about 22 percentage points for a $5 or $10 incentive. A key difference with our study is Alsan et al. (2019) provided vaccinations on site. Nonetheless, our study will be well-powered if our interventions have impacts even 1/10th the size as those in Alsan et al. (2019). Alsan M., Owen G. and Graziani G. 2019. "Does Diversity Matter for Health? Experimental Evidence from Oakland." American Economic Review 109(12): 4071-4111. DOI: 10.1257/aer.20181446.

Pre-analysis+plan+Encouraging+Vaccinations+for+COVID19+v7.docx