Objective: To study the validity and accuracy of differing cutoff scores of the Ages and Stages Questionnaires-Chinese (ASQ-C) for screening infants and toddlers in comparison with the gold standard, Bayley Scale of Infant Development, Second Edition (BSID II).
Method: The 269 samples were enrolled from the normative children, aged 3 - 31 months, of the ASQ-C in Shanghai. The age-appropriate ASQ-Cs were completed by parents/caregivers and the BSIDIIwas administered by professionals. The cutoff scores of -2 standard deviation (s), -1.5 s, and -s for the ASQ-C were examined against BSID II with the cutoff scores set at -2 s as the standard of developmental delay, -s as the standard of suspected developmental delay and developmental delay respectively. Agreement between the classifications of the ASQ-C (i.e., typical, suspected, delay) was compared with the classification of the BSID II (typical, suspected, delay), sensitivity, specificity, Youden Index and area under ROC curve of ASQ-C were examined. The statistical analysis was carried out using SPSS 13.0.
Result: When the cutoff score for BSID II was -2 s, the -2 s cutoff score for ASQ-C exhibited the following properties: the highest agreement of 83.64%, the sensitivity and specificity both above 80% being respectively 88.46% and 83.13%, the highest Youden Index of 0.72 and the largest area of 0.86 under ROC curve. The -1.5 s cutoff score for ASQ-C showed the following properties: 71.75% agreement, 100% sensitivity, 68.72% specificity, Youden Index = 0.69, the area under ROC curve = 0.84. The -s cutoff score for ASQ-C showed the following properties: the lowest agreement of 55.02%, 100% sensitivity, the lowest specificity of 50.21%, the lowest Youden Index of 0.50, and the smallest area of 0.75 under ROC curve. When the cutoff score for BSID II was set at -s, the -2 s for ASQ-C showed the following properties: the highest agreement of 85.87%, the lowest sensitivity of 68.57%, the highest specificity of 91.96%, Youden Index = 0.61, the smallest area = 0.77 under ROC curve. The -1.5 s for ASQ-C showed the following properties: the agreement of 80.67%, the sensitivity and specificity both above 70% being respectively 85.71% and 78.89%, the highest Youden Index of 0.65, the largest area of 0.82 under ROC. The -s cutoff score for ASQ-C showed the following properties: the lowest agreement of 68.40%, the highest sensitivity of 94.29%, the lowest specificity of 59.30%, the lowest Youden Index of 0.54, and the area under ROC curve = 0.80. When the cutoff score for BSID II was -1 to 2 s, the identifying percentages of the -2 s, -1.5 s and -s for the ASQ-C were 56.82%, 77.27% and 90.91%, respectively.
Conclusion: For developmental delay identification, the -2 s cutoff score for ASQ-C produces the most robust validity and highest accuracy; for the identification of suspected developmental delay and developmental delay, the -1.5 s cutoff score for ASQ-C has the highest screening accuracy with appropriate sensitivity and specificity; for identifying the suspected developmental delay, the -s cutoff score for ASQ-C has the highest percentage of the identification. It is necessary to add 1 - 2 s to the ASQ-C's cutoff scores as the standard for screening suspected developmental delays.