Stability and predictability of Bayley Scales of Infant and Toddler Development: evidence from a south Indian birth cohort prospective study

Abstract

Objectives: There is limited information about the stability and predictability of Bayley Scales of Infant and Toddler Development (BSID) assessing child development in low- and middle-income settings. The objective of the present study was to analyse stability and predictive validity of BSID using an existing birth cohort.

Design: Prospective birth cohort follow-up study.

Setting and participants: A community-based birth cohort of 251 newborns was recruited and followed-up in urban Vellore, South India, until 9 years of age. Using BSID-III, child development was measured at 6, 15, 24 and 36 months. Cognition was assessed using the Wechsler Preschool Primary Scales of Intelligence at 5 years, and the Malin's Intelligence Scale for Indian Children scale at 7 and 9 years of age. The stability of BSID measurements across time points was expressed by intraclass correlation (ICC) and concordance correlation coefficients. Linear regression was used to describe the predictability of BSID-III of cognition at 5, 7 and 9 years.

Results: The ICC for domain-wise BSID scores between time points of measurement suggested a weak correlation. The BSID scores at 36 months correlated best with Full-Scale Intelligence Quotient (FSIQ) at 5 years (r: 0.40-0.49), 7 years (r: 0.35-0.48) and 9 years (r: 0.36-0.38). BSID scores at 36 months predicted FSIQ better at 5, 7 and 9 years with R2 ranging from 23.3% to 28.6%, when compared with 24 months BSID scores (R2 - 16.0% to 25.9%).

Conclusion: Poor stability and predictability of BSID warrant caution in the predictive projection of early childhood assessments. Better predictability of future cognition of 36 months' BSID scores highlights its advantage over the 24 months' assessment.