The value of child health surveillance (CHS) practices in early detection of mild to moderate learning difficulties (LD) was investigated at the North and West Belfast Community Paediatric Unit, Belfast, N Ireland. The prevalence of learning difficulties in this deprived inner city area was 16%. Only 6% of children with LD were identified by the CHS in the preschool period. Perinatal variables associated with LD were lower social class, prematurity, male sex, and birth to an unmarried mother. Risk factors used in the CHS which proved to be insensitive included speech delay, poor parenting, behavior problems, enuresis, poor visual acuity, and otitis media with effusion. 
COMMENT. Child Health Surveillance (CHS) by health visitors, at birth, 6 weeks, 6, 12, and 18 months, 2,3, and 4 years, failed to detect the majority of children with mild learning difficulties and missed 38% of the moderately learning disabled. Retrospective analysis of the child health record failed to identify a model to predict children with later LD. The failure of CHS in this setting was attributed to a combination of poor test sensitivity and the overlap of LD with variants of developmental norms. Nursery school and preschool surveillance, and increased parental and professional awareness could provide more accurate early detection of LD children.