A diagnostically promising technique for tallying nominal reference errors in the narratives of school‐aged children with Foetal Alcohol Spectrum Disorders (FASD)

Abstract

Foetal Alcohol Spectrum Disorders (FASD) include the range of disabilities that occur in children exposed to alcohol during pregnancy, with Foetal Alcohol Syndrome (FAS) on the severe end of the spectrum. Clinical research has documented a range of cognitive, social, and communication deficits in FASD and it indicates the need for diagnostic tools that can identify children with diminished communicative capacities resulting from prenatal alcohol exposure. Previous research indicates that analysis of nominal reference errors within narrative discourse may provide such a tool. To demonstrate the potential diagnostic utility of a new tool for tallying nominal reference errors in the oral narratives of school-aged children with FASD by presenting quantitative measurement data that address interrater agreement and predictive accuracy. Retrospective analysis was conducted on spontaneously produced oral narratives from 32 school-aged children (8;5-11;7) with a range of socio-economic and ethnic profiles. Sixteen of the children had been previously diagnosis with an FASD, including five with full or partial FAS (pFAS). The remaining 16 children were considered typically developing (TD). A range of methods for calculating the rate of nominal reference errors (rNRE) were used to predict which narratives were produced by children from each group. Accuracy (sensitivity and specificity) for two predictions (FASD versus TD, and FAS/pFAS versus all others) was quantified using receiver-operating characteristic curve analyses. Pairwise statistical comparisons were made between methods to determine which had the most diagnostic potential. The proposed system for calculating the rNRE was highly accurate at predicting which narratives were produced by children with FASD (versus TD, 88% overall accuracy), and which were produced by children with FAS/pFAS (versus all others, 97% overall accuracy), and outperformed all other methods tested. Agreement on coding decisions between independent judges was high (kappa = 0.90). The strong predictive accuracy demonstrated in this study provides empirical evidence that the system proposed in this feasibility study has sufficient sensitivity and diagnostic utility to warrant further development for use with children suspected of prenatal alcohol exposure. It also points to the potential for the tool to be used with other clinical populations that, even in the absence of a confirmed alcohol exposure, share many of the communication challenges of this complex clinical population.