Experimental Design
Study Site
The project will be carried out in healthcare facilities located in four regions with some of the lowest vaccination rates in Ghana: Volta, Upper West, Ashanti, and Western regions. Located in the southern part of Ghana, the Volta Region had the lowest percentage of persons fully vaccinated against COVID-19 (18.5%). This region is followed by the Western Region with 30.6% and Ashanti region with 34%. We will also select the Upper West Region from Northern Ghana as the least performing region in that part of Ghana with a 47.4% vaccination rate. The choice of this region is to provide a more diverse range of regions, as Northern Ghana usually suffers from worse socio-economic characteristics. Our focus on selecting regions on Northern and Southern Ghana is to provide a nationwide representation of COVID-19 vaccine behaviours and attitudes in the country, as there are linguistic, structural, and health variations between Northern and Southern Ghana.
Study Population
The target population for the RCT project will be patients attending selected health centres in the selected regions. It will also include frontline health professionals who provide care to these patients.
Inclusion/Exclusion Criteria
The target population for the cRCT project will be mentally sound adults who 1) are aged 18 years and above, 2) have not received COVID-19 vaccination, 3) and are hesitant about getting the COVID-19 vaccine.
Sampling
Within each region, we select 10-12 worse performing districts or municipalities. In each district, we conduct a census of health centres, yielding 235 health centres. We exclude health centres which do not offer COVID-19 vaccines any time and / or do not have them in stock at time of census, yielding a sample of 144 centres. We sample from these proportionally to the number of health centres in each region. Finally, we randomise the facilities to the three experimental groups, stratified on region and district vaccine rates.
Data collection
Patient exit interviews. On each facility visit, we will recruit 3 eligible patients. To be eligible to participate in our study, patients must be (1) adult (2) unvaccinated or partially vaccinated and (3) not “feel or look ill”. Once eligibility is confirmed, we will administer a brief baseline questionnaire including information on reason for visit, health insurance status, and some socio-economic questions. To facilitate identification of patients throughout the care pathway, we will give patient a unique ID (e.g. a sticker) and ask them to come back upon exiting the facility for an exit interview. When they exit the facility, we will conduct an exit interview and ask about (1) the content of the conversation with the health worker seen, (2) their satisfaction and trust (in the HW; the health system and the government), (3) their level of knowledge about COVID-19 and the vaccine; (4) their beliefs about COVID-19 and the vaccine. The research team would request to see the vaccination cards of patients who indicated they received the vaccine.
Vaccination records: we will capture the number of vaccination recorded in intervention facilities records over the trial period. To obtain confirmation of the vaccination status of interviewed individuals, we will establish a system to capture the decision to get vaccinated of individuals “referred” by the intervention, and distinguish them from those coming for vaccination outside of the intervention. This will also allow us to capture delayed decisions to get vaccinated, defined in our context as up to one week after a discussion was held.
Facility survey. On the day of the facility visits, we will measure the availability and stocks of COVID-19 vaccines. We will also measure the volume of patients waiting for vaccination.
Provider surveys. Each participating provider will be interviewed twice. First, they will be interviewed at the end of the training to determine their knowledge and beliefs after the intervention, and before the first interactions with patients. Providers will then be interviewed a second time at the end of the study period to measure their motivation and satisfaction with the intervention, their beliefs about patients’ vaccination intentions, and their clinical knowledge (to benchmark the quality of care measure we will record).
Standardised patients. In addition to patient exit interviews, we will use standardised patients to (1) ascertain the extent to which doctors implement the intervention and (2) obtain objective measures of quality of care to check whether the intervention affects it.
Econometric approach
Our general approach will be to regress the outcomes of interest on dummy variables indicating treatment status. Standard errors will be clustered at the facility level to account for the randomization at that level. In addition to this basic specification, we will estimate an additional model that will include individual and facility characteristics measured at baseline. We will run a regression selecting individual controls following the post-double-selection (PDS) methodology.
Sub-group analysis will allow us to determine whether the intervention is more effective in the following groups:
1. Individuals with pre-consultation high levels of vaccine intentions (defined at higher than 5 on a scale out of 10)
2. Individuals with risk factors for COVID-19
3. Individuals who support the presidential party
4. Individuals partially vaccinated.
To account for multiple inference, we will follow two approaches. First, we will define summary indices for family of attitudes and behaviours, constructed following Kling, Liebman et al. (2007) and test whether these indices are affected by the treatments. Second, for individual outcome variables within each group, we will adjust for multiple inference using a family-wise p-value adjustment following Anderson (2008).