“…This model had an area under the receiver operating characteristic (ROC) curve (AUC) of 90.7% and a cut-off could be chosen with 100% sensitivity and 72% specificity ( Figure 2). At 18 months, the bivariate logistic regression analysis allowed us to identify 17 Q-CHAT items that were significantly associated to positivity to ASD or BAP (1,2,3,4,5,6,7,8,9,10,14,15,16,17,19, 20, 23 and 25; p < 0.05). Again, starting from these items, we carried out a multivariate logistic regression adopting a stepdown procedure and obtained a model with 4 significantly associated items: item 10 ("Does your child follow where you're looking?…”