Besides the health sector, COVID-19 has had a far-reaching impact across different spheres of life. The work of the AI4COVID research team in Sri Lanka and Malaysia has focused on highlighting some of the socio-economic impacts of the pandemic, especially amongst diverse groups such as women, children, and underprivileged groups. In this light, a key endeavour of the team has been to create a dataset relying on survey data collected from 3000 households in Sri Lanka on the social and economic impacts. Here we share our experiences and lessons we learned along the way. 

Developing and testing the questionnaire 

A semi-structured questionnaire was developed for the purpose of capturing this data. A series of pilot tests were then conducted using different remote methods (such as via online platforms and telephone interviews) considering the strict travel restrictions imposed at the time. In addition, face-to-face interviewing was later tested as an alternate approach when the situation allowed it. Combined with expert advisory, the feedback we received during the piloting phase was used to refine the language in the questionnaire and the design of the survey tool itself. For increased efficiency and accuracy, the data was collected using a uniquely developed Computer-Assisted Personal Interviews (CAPI) system that allows responses to be digitally recorded in real-time during an interview. Given the diversity in languages spoken within the targeted areas, the questionnaire was translated into three languages: Sinhala, Tamil, and English. 

Getting prior permissions and approvals 

Prior to commencing the study, the team sought and received clearance to carry out its activities from the Ethical Approval Committee at the Faculty of Arts, University of Peradeniya (the lead organization involved in this project). In Sri Lanka, permission from the village-level administrative officer, known as Grama Niladhari (GN), was required to conduct village-level data collection of our scope. Therefore, we took the approach of first requesting approval from the Ministry of Home Affairs that is responsible for the administrative sector in Sri Lanka, which would then ensure support from the respective administrative officers in targeted locations. 

As part of this approval process, prior notices and requests were sent to the Divisional Level and GN level, with considerable attention given to this task to ensure a smooth data collection. Getting local buy-in from the legal authorities also helped with encouraging participation from people in the community due to familiarity and a certain degree of trust. In cases where people were reluctant to welcome outsiders to their homes, the local chaperones were helpful in facilitating the interview that ensured we could access a broad range of respondents. Furthermore, detailed discussions with village-level administrative officers and assistants about the scope and objectives of this study made sure there was a clear mutual understanding that then informed the sample selection. 

Selecting individual respondents 

The questionnaire contained questions about various aspects of impact, such as social (educational, health and social stigma); economic (impact on income and food intake); cultural; mobility-related and socio-political. The expectation was that different sections would be more appropriate for different household members. As an example, the education impact component was intended for full-time or part-time students. Accordingly,  all the enumerators were trained to identify and map each respondent with the appropriate component. This training also included a practical session to mimic actual scenarios that was guided by the project’s research investigators. 

For the most part, we selected households with at least four members. Households with fewer than three members were not considered, unless the household was significantly impacted by COVID-19 (an assessment which relied on the knowledge and advice of each village’s GN). One of the key lessons was that: different households have diverse scenarios, and any generalization about particular aspects of impact should be kept at bay. Therefore, the training the enumerators received was quite critical in helping them to identify these nuances. 

Keeping in mind cultural and religious aspects 

In some households, women were not permitted to respond to a male outsider unless her husband or another male household member was present. Understanding this sensitivity, female enumerators were specifically assigned to these households. Regarding language use, enumerators were proficient in speaking both Sinhala and Tamil languages, which was incredibly useful in making sure all respondents could communicate in their preferred language. At times a few individuals were also not willing to respond to parts of the questionnaire due to political sensitivities. Fortunately though with the help of local assistants and religious leaders, the research team was able to navigate these sensitivities. Having a gender balance in the enumerator team emerged as a key necessity, particularly in facilitating discussions with the respondents in line with their cultural, religious and social norms. 

Conducting the interview 

The duration of each interview ran on average 35-40 minutes. Understandably in a home setting, this meant some respondents also had to be juggling domestic activities, such as cooking, religious activities and studying. In those instances, the enumerators had to wait until there was enough time for the respondents to answer the questions.  This in the end reduced the number of interviews that could be scheduled in a day. As the CAPI system is a real-time tool, none of the data could be entered retroactively and it was not feasible to revisit the households to repeat the interview. This feature was incorporated into the CAPI system as an accountability mechanism to make sure enumerators were not able to alter data after the interview concluded. 

One of the key takeaways from the process was that: people are not always able to provide dedicated time to respond to a questionnaire. Therefore, these expected delays should be accounted for in the time planning for the data collection process. We also found that sometimes the respondents would mislead the interviewers, with the hope of convincing them that they were highly impacted by COVID-19 – assuming governments would provide assistance accordingly. Thus, explaining the goal of the study and purpose of the data collection was a key part of informed consent. In addition, whilst on-the-spot inputting of data into the CAPI system was practical, it was sometimes a barrier for enumerators who had to multitask with asking questions, inputting responses and handling the device. 

Ensuring safety first

Conducting fieldwork in the middle of the pandemic came with its own challenges, but ensuring the safety of the enumerators and the respondents was central to all decisions made. For this reason, we diligently followed all the health guidelines provided by the World Health Organization and the Sri Lankan government policies on COVID-19 throughout the data collection. Given the research team’s decision to conduct the data collection through face-to-face interviews, we consciously delayed the in person data collection until the travel restrictions were lifted. As part of the precautions, all enumerators were advised to keep their physical distance from the respondents; avoid carrying any food or drinks; and conduct interviews outdoors with proper ventilation. All the enumerators, supervisors, and members of the research team were supplemented with Personal Protective Equipment such as masks and sanitizers.

In addition to this main data collection that has already been conducted, there will be three other components: i) Focus Group Discussions, to collect qualitative data from the most affected groups of people who were identified through the main survey; ii) in-depth interviews, to collect data on mobility patterns; and iii) mobility tracking using a mobile app, to trace real-time mobility of selected individuals. We have also shared about privacy by design and what it means for our research.

We are thankful to Canada’s International Development Research Centre (IDRC) and the Swedish International Development Cooperation Agency (Sida) for supporting our research through the Global South AI4COVID program. Visit our project website: http://covid.eng.pdn.ac.lk