Background Developmental outcomes of very preterm (gestational age ≤32 weeks) or very low birthweight (<1500g) children are commonly reported before age 3 years.
Aims To determine the validity of early developmental assessments in predicting school-age cognitive deficits in this population.
Methods English-language studies, identified through MEDLINE, in which at least 2 serial developmental/cognitive assessments (at ages 1–3 years and ≥5 years) were reviewed. For included studies, cross-tabulations of cognitive deficit (defined as test scores one SD below the population mean) identified by each assessment were constructed. With school-age assessment as the reference standard, the sensitivity and specificity of early assessment for cognitive deficit were calculated. We fitted a hierarchical summary receiver operator characteristic (HSROC) curve to obtain pooled estimates of sensitivity and specificity. Meta-regression was used to evaluate study and population characteristics associated with assessment validity.
Results Twenty-two studies (n = 2681 children) were included. Early assessment tools were Bayley Scales of Infant Development (14 studies), Griffiths Mental Development (5), Stanford-Binet Intelligence (2) and Brunet-Lezine (1) Scales; 11 different school-age cognitive tests were used. Assessments were conducted at 18–40 months for early and 5–18 years for school-age assessments. There was significant between-study heterogeneity in the reported sensitivities and specificities (figure 1, p < 0.01). Gestational age, birthweight, age at assessment and time difference between assessments did not explain between-study heterogeneity in results. From the HSROC curve (figure 2), pooled sensitivity (95% CI) was 0.55 (0.44–0.64) and specificity 0.84 (0.77–0.89).
Conclusion Early developmental assessment had poor sensitivity but good specificity at predicting school-age cognitive deficits.