Replication and Validation of the Brief Autism Detection in Early Childhood (BADEC) in a Clinical Sample

Background: Because ASD can be reliably diagnosed as early as 18 months, pediatricians are recommended to use ASD screeners during 18 and 24 month wellness checks. Despite this, most children are not diagnosed until after age four, which is partially due to the inconsistent use of screeners. Reasons for this include short appointment times and clinicians’ limited knowledge of available screeners. One screener, the Autism Detection in Early Childhood (ADEC) is a 16-item, play based screener for 12-36 month olds. It has good reliability and validity, requires minimal training, and can be administered in 10-15 minutes. Despite these strengths, the ADEC has not been widely implemented. Since tools such as this it would still consume most of a standard 9-16 minute visit, there is a need to abbreviate it further. The BADEC, a five-minute version of the ADEC was therefore developed using a research sample (Nah et al., 2018).

Objectives: To evaluate the BADEC in a clinical sample by a) calculating the Nah et al. BADEC’s sensitivity and specificity b) replicating their procedures to determine if the same five items from the original ADEC are identified as best predictors, and c) evaluating the screening ability and validity of abbreviated version we identified.

Methods: Participants were 107 children aged 14-36 months (M = 28.70 months, SD = 5.40) with confirmed final diagnosis of ASD (n = 48) or who had ASD ruled out (NASD; n = 59), and had complete ADEC data. Participants were screened and evaluated at a pediatric hospital. Screening ability of the Nah et al. (2018) BADEC was assessed in the current sample. Following Nah et al. (2018), Receiver Operating Characteristic analysis was performed on all 16 ADEC items to identify five items associated with best area under the curve (AUC) values. These were then combined into one overall current BADEC score, and sensitivity, specificity, and concurrent, predictive, and diagnostic validity were calculated.

Results: The following items emerged with highest AUC values: Gaze Monitoring (.82), Task Switching (.75), Response to Name (.74), Reciprocity of a Smile (.74), and Following Verbal Commands (.73). Our findings are in agreement with Nah et al. (2018) on the inclusion of three items. While Nah et al (2018)'s data supported a cutoff of 4, our data supported a cutoff of 5 (Se = .77, Sp = .86, PPV = .82, NPV = .82, AUC = .82). Both versions of the BADEC had strong concurrent, predictive, and diagnostic validity.

Conclusions: An abbreviated version of the ADEC can effectively screen for ASD in children under age three. Short tools such as these are particularly amenable for inclusion in wellness checks given that they are easy to use. This can facilitate timely access to supports for those who go on to receive a diagnosis. There is value in evaluating screening tools in clinical contexts given that the majority of available screeners are developed using research samples. Future research evaluating brief screeners should evaluate whether adherence to screening guidelines increases after pediatricians are trained in its use.