Background The widely used Mental Developmental Index (MDI) of the Bayley Scales of Infant Development-II has recently been superseded by the Bayley-III – in which cognitive and language development are assessed separately. Comparing developmental outcomes between cohorts assessed using two different editions of this test is problematic.
Aims To compare MDI and Bayley-III cognitive and language scores, evaluate the agreement between classifications of disability made using the two tests, and develop an algorithm for converting Bayley-III scores into corresponding MDI scores.
Methods 185 children (derived from a sub-cohort of the EPICure-2 study) aged 29–41 months were administered the MDI and Bayley-III concurrently. An average of Bayley-III cognitive and language scores (CB-III score) was calculated for comparison with MDI scores.
Results Bayley-III cognitive, language and CB-III scores were 3-, 10- and 7-points higher respectively than MDI scores, but the relationship was non-linear: the CB-III progressively overestimated developmental scores at lower MDI scores. Using a conventional cut-off score <70, CB-III scores under-detected developmental impairment compared with the MDI (sensitivity 58%, specificity 100%). Predictive validity was improved using a CB-III cut-off <80 (sensitivity 89%, specificity 99%). An algorithm was derived for transforming CB-III scores into MDI-equivalent scores, with further improvement in classification of children with MDI<70 (sensitivity 95%, specificity 97%).
Conclusions Caution should be exercised in classifying developmental outcomes using the Bayley-III for research purposes. To produce equivalent identification of children who would have an MDI <70, we recommend using a CB-III scores <80 or predicted-MDI scores <70.