Social patterning and prediction of parent-reported behaviour problems at 3 years in a cohort study

To study the determinants of parent-reported behaviour problems at 3 years and the value of behaviour problems reported at 8 and 18 months as predictors of behaviour at 3 years

A whole year birth cohort

Coventry

Parent-reported behaviour problems at 3 years

A total of 2580 infants were enrolled into the Coventry Cohort Study at the birth visit by their family health visitor. Data on parent-reported behaviour at all three ages (8 months, 18 months and 3 years) were available on 775 infants.

Living in rented accommodation [adjusted OR 2.38 (95% CI 1.36, 4.21); OR, odds ratio; CI, confidence intervals] and living in a smoking household [adjusted OR 2.47 (95% CI 1.53, 3.99)] were independently associated with parent-reported behaviour problems at 3 years after controlling for other sociodemographic variables in logistic regression. The risk of behaviour problems at 3 years was increased in those reporting behaviour problems at 8 months [OR 3.77 (95% CI 1.73, 8.20)] and 18 months [OR 5.84 (95% CI 3.34, 10.23)] after adjustment for sociodemographic variables and other health problems. Behaviour at 8 months as a predictor of behaviour at 3 years had a sensitivity of 13.9%, a specificity of 95.7%, a positive predictive value of 32.6%, and positive and negative likelihood ratios of 3.23 and 0.90. For behaviour at 18 months, the sensitivity was 35.6%, the specificity 92.9%, the positive predictive value 42.9%, and positive and negative likelihood ratios 5.01 and 0.69.

Parent-reported behaviour problems at 8 and 18 months are highly specific but not particularly sensitive as predictors of behaviour problems at 3 years. Assuming the availability of an effective early intervention, use of a question such as that in the Warwick Child Health and Morbidity Profile at 8 and 18 months will identify, respectively, 21% and 36% of children at risk of parent-reported behaviour problems at 3 years.