Introduction The main advantages of birth cohort studies may be compromised by the refusals in participation, which tend to increase in time.

Objective To describe the main predictors of refusals in the Portuguese birth cohort 4 years after the assembling.

Methods Geração XXI was assembled in 2005/2006 in five public maternities in Porto Metropolitan area, Portugal. After delivery, data were collected by personal interviews. Sub-samples of mothers were re-assessed at 6, 15 and 24 months and an ongoing evaluation of the entire cohort is being performed at the age of 4/5 years. Refusals at follow-up were compared with evaluated participants considering baseline maternal and childhood characteristics, fitting logistic regression models were adjusted for age, education level and working condition, place of residence and previous follow-ups.

Results Within 6025 participants already contacted, 88% were evaluated and 704 (12%) refused. The major determinant of refusal was the absence of previous evaluations (OR=2.01; 95% CI 1.68 to 2.41). Refusals were more frequent among less educated women (≤9 vs ≥13 years of education: OR=1.23; 95% CI 1.01 to 1.53), as within housewives (vs employed OR=1.60; 95% CI 1.14 to 2.25). No statistical significant differences were found regarding other maternal, birth or newborn characteristics. The main reasons for refusal were unavailability (39.6%), personal/professional reasons (18.9%) and distance to evaluation setting (16.1%).

Conclusions This analysis suggests that keeping regular contacts with participants is probably the best approach to reduce refusals, which were, in this cohort, mainly determined by maternal socio-demographics characteristics.