Design and maintenance of community-based cohorts in sub-Saharan Africa: a longitudinal evaluation of participant attrition in SchistoTrack.

Abstract

Introduction: Understanding participant attrition in longitudinal studies is essential for maintaining cohorts, establishing targeted interventions and assessing potential biases introduced in study analyses. Yet, limited metrics, models and long-term assessments exist for the design and evaluation of community-based cohorts in sub-Saharan Africa.

Methods: We prospectively assessed participant attrition in the SchistoTrack cohort. A total of 2885 individuals aged 5-92 years, who were recruited in 2022, were examined from 1445 randomly sampled households across three rural Ugandan districts. Baseline data on sociodemographics, medical history, spatial factors and clinical examinations were collected, with annual and seasonal follow-ups to 2025. Profiles of attriters and rejoiners were established with logistic regressions. Community engagement, retention strategies and study expansion to 4321 participants were evaluated.

Results: Overall attrition rates were stable across the years, ranging from 20.9% to 27.7%. Attriter profiles were established within the first year and were stable predictors of the recurrent attrition. Older individuals were less likely (OR 0.987, 95% CI 0.983 to 0.991), whereas more educated individuals were more likely to be lost to follow-up (OR 1.086, 95% CI 1.057 to 1.117). Individuals belonging to larger households and homes closer to government health centres were less likely to have an attrition event (ORs 0.920-0.924). Fishermen were not more likely than other individuals to have an attrition event, either overall or seasonally. 63.3% (374/591) of participants who dropped out from the first major follow-up and 54.1% (725/1341) of all participants who had any loss to follow-up for reasons other than death later rejoined the study. Schistosome infection and medical referrals were not associated with later attrition. Communicating clinical findings, adjusting incentives and study logistics did not influence attrition.

Conclusions: By providing metrics and models for tracking attrition, our attrition analysis framework can guide the design and evaluation of community-based cohorts in rural sub-Saharan Africa.