This population-based, retrospective cohort study of very low birth weight infants was undertaken to: 1) identify factors associated with nonattendance for follow-up, 2) estimate the prevalence of cerebral palsy at age > or = 18 months, and 3) model the prognostic association between prenatal and perinatal risk markers and cerebral palsy. The sample included 496 surviving very low birth weight infants born in 32 hospitals in Southwest Ontario between January 1982 and December 1986. Multivariate analyses were performed using the proportional odds regression model. Loss to follow-up was more likely among those with mothers < 20 years of age, those with unmarried mothers, and those not born in a tertiary center. Loss to follow-up was less likely for those with neonatal anemia and those of lower birth weight. Motor development of the 369 children who were followed at least 18 months was classified into one of three categories: normal, suspect, or cerebral palsy. Multivariate analysis revealed that factors predictive of poorer outcome were intraventricular hemorrhage, unmarried mother, male sex, recurrent apnea, and hydrocephalus. The finding that unmarried status was associated with loss to follow-up and was also an important predictor of cerebral palsy suggests that it is important for follow-up clinics to identify ways of assisting this population to remain in contact with the clinic.