Fig. 2: Accuracy metrics and normalized SHAP value analysis.

a, ROC curve illustrating the performance of the model for classifying different diagnostic groups, using all app variables. n = 475 participants; 49 were diagnosed with autism and 98 were diagnosed with developmental delay or language delay without autism. The final score of the M-CHAT-R/F screening questionnaire was used when available (n = 374/377). Error bands correspond to 95% CI computed by the Hanley McNeil method. b, Examples of app administration reports are shown, one for a 25-month-old neurotypical boy and one for a 30-month-old autistic girl, both correctly classified, including each child’s app quality score, confidence score and the contributions of each app variable to the child’s individualized prediction. c, Normalized SHAP value analysis showing the app variables importance for the prediction of autism. The x axis represents the features’ contribution to the final prediction, with positive or negative values associated with an increase in the likelihood of an autism or neurotypical diagnosis, respectively. The y axis lists the app variables in descending order of importance. The blue–red color gradient indicates the relevance of each of the app variables to the score, from low to high values; gray indicates missing variables. For each app variable, a point represents the normalized SHAP value of an individual participant. NT, neurotypical.