Is attrition bias a problem in neonatal follow-up?