Using the PDD Behavior Inventory as a Level 2 Screener: A Classification and Regression Trees Analysis

Abstract

In order to improve discrimination accuracy between Autism Spectrum Disorder (ASD) and similar neurodevelopmental disorders, a data mining procedure, Classification and Regression Trees (CART), was used on a large multi-site sample of PDD Behavior Inventory (PDDBI) forms on children with and without ASD. Discrimination accuracy exceeded 80 %, generalized to an independent validation set, and generalized across age groups and sites, and agreed well with ADOS classifications. Parent PDDBIs yielded better results than teacher PDDBIs but, when CART predictions agreed across informants, sensitivity increased. Results also revealed three subtypes of ASD: minimally verbal, verbal, and atypical; and two, relatively common subtypes of non-ASD children: social pragmatic problems and good social skills. These subgroups corresponded to differences in behavior profiles and associated bio-medical findings.

Level 2 Screening With the PDD Behavior Inventory: Subgroup Profiles and Implications for Differential Diagnosis

“…The partitioning procedure is repeated recursively at each node. The General Regression and Classification Trees module from the Statistica 12 package (Dell, 2015) was used (where more details about its methods can be found), and more details regarding how the analysis was carried out for the current study can be found in Cohen et al (2016). Domains from the PDDBI and whether the (Cohen & Sudhalter, 2005).…”

“…When using ASD-specific instruments such as the PDDBI to aid in diagnosis, sensitivity (agreement of the instrument with clinical diagnosis of ASD) and specificity (agreement of the instrument with clinical diagnosis of not ASD) are of importance. However, as noted previously (Cohen et al, 2016), these metrics depend, in part, on the degree to which the person being diagnosed displays behaviors that relate to ASD or to behaviors typical of disorders that can overlap with ASD, for example, social anxiety, obsessivecompulsive behaviors, language delay, and so on (Eaves, Wingert, Ho, & Mickelson, 2006;Havdahl, von Tetzchner, Huerta, Lord, & Bishop, 2016;Muratori et al, 2011).…”

“…Based on a large, multisite sample of children with ASD, typically developing children and those with disorders other than ASD, Cohen et al (2016) reported that when the PDDBI was analyzed with CART, sensitivity and specificity results were quite good, showing agreement with Autism Diagnostic Observation Schedule (ADOS) groupings, and generalized across age groups and diagnostic assessment sites.…”

“…Described below is a brief summary of the characteristics of the sample used to develop the decision tree and how the CART analyses were carried out in the original study (Cohen et al, 2016) where more details can be found. This is followed by examination of the behavioral profiles provided on the parent and teacher versions of the PDDBI.…”

Screening in toddlers and preschoolers at risk for autism spectrum disorder: Evaluating a novel mobile‐health screening tool

“…Accuracy depends on achieving an ideal balance of sensitivity (the measure's ability to correctly recognize the presence of a disorder) and specificity (the ability of a measure to correctly recognize when a disorder is absent). Numerous studies have examined the efficacy of screening measures in ASD [e.g., Cohen et al, ]. In sum, these have typically found that the efficacy varies depending on several factors, including whether it's Level 1 or Level 2, broadband or narrowband, the content of the tool, the age range that is being targeted, and what aspect is being optimized (e.g., sensitivity or specificity).…”