Construction and validation of the Oxford Neurodevelopment Assessment (OX-NDA) in 1-year-old Brazilian children

Abstract

Abstract Background Over 250 million children under 5 years, globally, are at risk of developmental delay. Interventions during the first 2 years of life have enduring positive effects if children at risk are identified, using standardized assessments, within this window. However, identifying developmental delay during infancy is challenging and there are limited infant development assessments suitable for use in low- and middle-income (LMIC) settings. Here, we describe a new tool, the Oxford Neurodevelopment Assessment (OX-NDA), measuring cognition, language, motor, and behaviour, outcomes in 1-year-old children. We present the results of its evaluation against the Bayley Scales of Infant Development IIIrd edition (BSID-III) and its psychometric properties. Methods Sixteen international tools measuring infant development were analysed to inform the OX-NDA’s construction. Its agreement with the BSID-III, for cognitive, motor and language domains, was evaluated using intra-class correlations (ICCs, for absolute agreement), Bland-Altman analyses (for bias and limits of agreement), and sensitivity and specificity analyses (for accuracy) in 104 Brazilian children, aged 12 months (SD 8.4 days), recruited from the 2015 Pelotas Birth Cohort Study. Behaviour was not evaluated, as the BSID-III’s adaptive behaviour scale was not included in the cohort’s protocol. Cohen’s kappas and Cronbach’s alphas were calculated to determine the OX-NDA’s reliability and internal consistency respectively. Results Agreement was moderate for cognition and motor outcomes (ICCs 0.63 and 0.68, p < 0.001) and low for language outcomes (ICC 0.30, p < 0.04). Bland-Altman analysis showed little to no bias between measures across domains. The OX-NDA’s sensitivity and specificity for predicting moderate-to-severe delay on the BSID-III was 76, 73 and 43% and 75, 80 and 33% for cognition, motor and language outcomes, respectively. Inter-rater (k = 0.80-0.96) and test-rest (k = 0.85-0.94) reliability was high for all domains. Administration time was < 20 minutes. Conclusion The OX-NDA shows moderate agreement with the BSID-III for identifying infants at risk of cognitive and motor delay; agreement was low for language delay. It is a rapid, low-cost assessment constructed specifically for use in LMIC populations. Further work is needed to evaluate its use (i) across domains in populations beyond Brazil and (ii) to identify language delays in Brazilian children.