Electrodermal Variability and Symptom Severity in Children with Autism Spectrum Disorder

Data cleaning procedures included visual inspection of NSCRs as well as implementation of procedures to detect extreme scores. The wrist sensors did not tend to indicate very high EDA levels (e.g. >60 us) or positive spikes that occurred more rapidly than physiological processes might induce (e.g. >10–30 us/3 s). Only a single 9 s artifact was observed in the dataset, which involved both high EDA and a rapid rise greater than 10–30 us/3 s. To address the possibility of the sensors momentarily losing connection and mimicking an NSCR upon re-establishing contact, we identified EDA-level decreases of both 1 SD or more and 0.5 SD (based on a particular child’s mean EDA) over the course of 3 s. Both types of decreases were infrequent, and corrected data that excluded NSCRs preceded by an identified drop were virtually identical to the original data. For example, a near perfect correlation was observed between uncorrected and corrected NSCR scores during the SB5, the longest sampling period, r = .999).

Additional HLM analyses were performed to control for mean-level movement data in examining the association between EDA and ASD symptom severity. The pattern and significance of findings for EDA variability and for NSCRs in relation to the ADOS-2 comparison score remained unchanged. Given sample size considerations, the uncontrolled models are presented.