Carla J DeVincent1, Kenneth D Gadow. 1. Cody Center for Autism and Developmental Disabilities, Department of Pediatrics, Putnam Hall, South Campus, Stony Brook University, Stony Brook, New York 11794-8790, USA.
Abstract
OBJECTIVE: The present study compared three separate Child Symptom Inventory-4 (CSI-4) scoring algorithms for differentiating children with autism spectrum disorder (ASD) from youngsters with attention-deficit/hyperactivity disorder (ADHD). METHOD: Parents/teachers completed the CSI-4, a DSM-IV-referenced rating scale, for 6 to 12-year-old clinical referrals with ASD (N = 186) and ADHD (N = 251). Algorithms were based on either all CSI-4 items (forward logistic regressions) or the 12 DSM-IV symptoms of pervasive developmental disorder (PDD) included in the CSI-4. RESULTS: ROC analyses indicated generally good to excellent values for area under the curve, sensitivity, specificity, and positive predictive power. Algorithms for parent ratings were superior to teacher ratings. The algorithm based solely on PDD symptoms evidenced the greatest generalizability. CONCLUSION: Although algorithms generated from regression analyses produced greater clinical utility for specific samples, the PDD-based algorithm resulted in greater stability across samples.
OBJECTIVE: The present study compared three separate Child Symptom Inventory-4 (CSI-4) scoring algorithms for differentiating children with autism spectrum disorder (ASD) from youngsters with attention-deficit/hyperactivity disorder (ADHD). METHOD: Parents/teachers completed the CSI-4, a DSM-IV-referenced rating scale, for 6 to 12-year-old clinical referrals with ASD (N = 186) and ADHD (N = 251). Algorithms were based on either all CSI-4 items (forward logistic regressions) or the 12 DSM-IV symptoms of pervasive developmental disorder (PDD) included in the CSI-4. RESULTS: ROC analyses indicated generally good to excellent values for area under the curve, sensitivity, specificity, and positive predictive power. Algorithms for parent ratings were superior to teacher ratings. The algorithm based solely on PDD symptoms evidenced the greatest generalizability. CONCLUSION: Although algorithms generated from regression analyses produced greater clinical utility for specific samples, the PDD-based algorithm resulted in greater stability across samples.
Authors: Elizabeth N Allred; Olaf Dammann; Raina N Fichorova; Stephen R Hooper; Scott J Hunter; Robert M Joseph; Karl Kuban; Alan Leviton; Thomas Michael O'Shea; Megan N Scott Journal: J Neuroimmune Pharmacol Date: 2017-04-12 Impact factor: 4.147
Authors: Hyunsik Kim; Cara M Keifer; Craig Rodriguez-Seijas; Nicholas R Eaton; Matthew D Lerner; Kenneth D Gadow Journal: J Child Psychol Psychiatry Date: 2017-02-14 Impact factor: 8.982
Authors: Jelske W van der Burg; Elizabeth T Jensen; Margot van de Bor; Robert M Joseph; T Michael O'Shea; Karl Kuban; Elizabeth N Allred; Megan Scott; Scott Hunter; Stephen R Hooper; Olaf Dammann; Alan Leviton Journal: Early Hum Dev Date: 2017-08-17 Impact factor: 2.079