Is attrition bias a problem in neonatal follow-up?

To assess whether attrition rate influences outcome in the follow-up of very preterm infants. In a national follow-up study of infants born alive in 1983 in the Netherlands with a gestational age less than 32 weeks and/or a birth weight less than 1500 g, outcome was assessed separately for adolescents who responded early or late to a follow-up invitation at age 14 years. Neonatal data and outcome results of earlier assessments from early and late responders were compared to those of non-responders by univariate and nominal (polytomous logistic) regression analysis. There were 723 (76%) early responders, 130 (14%) late responders and 109 (11%) non-responders. We found significantly more non-Dutch origin and more disabilities and school problems at age 10 years in late- and especially in non-responders. At age 14 years, the health utility index was significantly lower in late responders compared to early responders. School outcome did not show difference in relation to the response groups. The results suggest that the incidence of adverse outcome in very preterm infants is underestimated when follow-up is incomplete and hence response rate is not a negligible problem in the assessment of late outcome. Therefore, follow-up studies should include a drop-out analysis to enable comparison to other studies.