Progressing efforts towards leveraging data science and artificial intelligence for COVID-19 response and recovery, LAISDAR recently wrapped up a longitudinal survey involving around 7000 participants in Rwanda aimed at collecting data on a range of topics related to COVID-19. From face mask use and hand hygiene, to social distancing and socio-economic impact, the topics were covered in three languages, namely Kinyarwanda (the mother tongue in Rwanda), French and English. This blog is a reflection on the approach and team learnings that emerged from this data collection process.
The LAISDAR project was designed to inform public health policy making related to current and future infectious diseases in Rwanda. With respect to COVID-19, one of the key goals was to gather data on COVID-19 that is currently fragmented across the health system, as well as collect new and enriched longitudinal data. Here we focus on the latter.
The data collection was conducted over several months in the form of a survey, from November 2021 to March 2022. Relying on the National Institute of Statistics of Rwanda (NISR) database, participants over the age of 18 were randomly selected. Informed consent was built into the design of the data collection, and so it was important to get each participant’s consent upfront. In total around 7000 people in the country took part, of which nearly 220 of them were contacted biweekly by phone with further questions. This was a massive undertaking, as shown in the tally of around 38,000 calls that were made throughout the survey period.
Around 38,000 calls were made by the data collectors throughout the survey period
A team of data collectors was mobilized to aid in this effort. The data collectors included university graduates who were selected on a competitive basis, and received training from LAISDAR’s research team on how to navigate the questionnaire and conduct phone interviews. With one supervisor overseeing three data collectors, the objective was to ensure that each data collector had the opportunity to ask clarifying questions to an informed resource person who would be available. Having a dedicated budget to compensate the data collectors was crucial and helped with expediting the recruitment process.
Data collectors participating in a training exercise
In designing the survey instrument, we recognised the importance of using effective language that was clear and succinct. This of course meant making sure that participants could comfortably understand the questions, and likewise respond to them in their chosen language. Accordingly, the questionnaire was written in three languages: Kinyarwanda (the mother tongue in Rwanda), French and English. It contained 59 questions covering a range of categories such as: demographics; face mask use; hand hygiene; adherence to social distancing and risk minimization measures; COVID-19 test results; mental health indicators and anxiety, and social-economic impact, amongst others.
To ensure efficiency and data privacy, this questionnaire was incorporated in a mobile application, known as SoGoSurvey – a cloud-based SaaS platform that enables creation, distribution and multilingual analysis of surveys. Each data collector was assigned a tablet, on which the application was installed. By design, all the calls were done over the phone to make sure participants with limited internet connectivity could still be reached. Calls were done by data collectors on a bi-weekly basis, and the responses were recorded using SogoSurvey thus ensuring that only the research team could access the data. We found the use of the application especially useful as it reduced human errors (for instance transferring written notes to digital) and added an extra layer of built-in data protection.
The first week of the data collection was an essential step for validating the tool, providing useful insights on what needed to be adjusted and potential glitches that both the participants and data collectors needed to anticipate. Each data collector interviewed the same participants throughout the weeks of the survey, and as part of the monitoring and quality control, each supervisor was required to provide a daily report on the performance of each data collector to the research coordinator for review. In addition, there was a research meeting held weekly to present and discuss the data quality.
One of the key learnings from this data collection process is that remote data collection requires persistence, excellent verbal communication to build trust with respondents and empathy. For instance, with network connectivity issues, several calls had to be rescheduled and the data collectors had to allow some flexibility. There were instances where some participants were busy at the time they were called, and so with the time constraint, the options were to either conduct the interview in a short period or reschedule the call. In general this required empathy on the part of the data collectors, not least because the data collection was being conducted in the middle of the pandemic where some respondents had to be juggling home care responsibilities.
We are pleased with the progress we have made thus far and recently convened a general research meeting with multiple stakeholders to share the preliminary results. Next up, the research time will be conducting further analysis of the survey data and preparing a manuscript that discusses the key results and findings.
This LAISDAR project is funded by Canada’s International Development Research Centre (IDRC) and the Swedish International Development Agency (Sida) as part of the Global South AI4COVID program. We extend thanks to the Government of Rwanda, all consortium institutions who have helped to facilitate project activities and to the funders. To learn more about our work at LAISDAR, please visit our website at https://rbc.gov.rw/laisdar, twitter account: @laisdar and Facebook: Laisdar LaisdarTags: #datascience, #survey, #datacollection, #citizenparticipation