Feasibility of respondent-driven sampling to recruit participants with recent abortion experiences in humanitarian...

This study aims to assess the feasibility of respondent-driven sampling (RDS) to recruit participants with recent abortion experiences in humanitarian contexts, and describe the composition of the study sample generated with this sampling method.

This was a three-phase mixed-methods community-engaged research study employing an exploratory and explanatory sequential approach. We conducted in-depth interviews, focus group discussions, an interviewer-administered questionnaire on abortion experiences and a health facility assessment.

Bidibidi Refugee Settlement, Uganda and Kakuma Refugee Camp, Kenya from November 2021 to December 2022.

Using RDS, we recruited 600 participants in Kakuma and 601 participants in Bidibidi with recent abortion experiences. In Kakuma, participants were primarily from Burundi, the Democratic Republic of the Congo and South Sudan; participants in Bidibidi were primarily from South Sudan. Most participants in both sites had completed at least some primary school and were not employed.

RDS recruitment dynamics: convergence and bottlenecks on key sociodemographic variables, recruitment and population homophily, reciprocity of social ties, success and experiences recruiting.

There were minor violations of RDS assumptions, particularly regarding assumptions of reciprocity of ties and seed composition independent of sample. In addition, there was a strong tendency of participants to recruit those from the same home country and living within the same camp zone. However, sample proportions for age, home country, marital status, zone of residence and student status reached equilibrium (stabilised) by around 500 participants at each site, and we were able to quickly attain the study sample size.

While the true representativeness of our sample remains unknown, RDS is a practical and effective recruitment method in humanitarian contexts for sensitive topics, particularly for research questions in which no data or sampling frames exist. However, attention to representativeness and community engagement is essential to optimising its application and ensuring success.