The outcome at age 2 years of preterm babies recruited into a three-arm randomised controlled trial of prophylactic volume expansion was ascertained in two ways: from a neurodevelopmental assessment performed by a paediatrician and from responses on a brief questionnaire completed by the child's personal health visitor. Of 776 babies recruited into the trial, 604 survived to the age of 2 years and the findings of a paediatric assessment were available for all survivors. Questionnaires were sent to the health visitors of 601 of the survivors; 513 (85.4%) were returned. There was sufficient information on the returned questionnaires to categorise 449 of the children as normal, impaired, moderately disabled or severely disabled. We were unable to detect a response bias by severity of disability. Agreement on individual questions ranged between 86.3% and 98.4%. There was some mismatch in the reporting of vision (weighted kappa = 0.71) and hearing (weighted kappa = 0.73), with differences in perception of level of severity of sensory loss. Health visitors tended to underestimate the child's functional level compared with the paediatrician. However, of 56 children classified as severely disabled by the paediatrician, 48 were classified similarly and eight as moderately disabled on the basis of the questionnaire. The end point of the trial was death or severe disability at 2 years of age. There was close similarity in the trial results whether based on the paediatric assessment or on the questionnaire. Further refinement of the questionnaire is needed, but this methodology may be useful in ascertaining the frequency of severe disability in large cohorts of babies.