Utility of Heart Rate Increase for Prediction of Challenging Behavior Episodes in Preschoolers with Autism
Objectives: The study aim was to determine the predictive utility of increase in heart rate to indicate a challenging behaviour episode in children with autism with frequent challenging behaviors.
Methods: Whilst wearing a ECG monitor, 41 children with autism recruited as part of a larger study aged 2-4 years participated in tasks from the Laboratory Temperament Assessment Battery, which mimic everyday life experiences requiring emotional regulation in low-level stress situations (e.g., waiting for a snack). Coders blind to diagnostic group coded challenging behaviors during the 1-1.5 hour-long sessions (i.e., aggression, self-injury, property destruction, loud noises and non-compliance, n=212) and random non-challenging behaviors (n=106). Only 13/41 participants exhibited challenging behaviors. Baseline-corrected heart rate (HR) was computed for each behaviour. The predictive utility of HR in challenging vs. non-challenging behaviors was examined via Receiver Operating Curve (ROC) analysis and a binary logistic regression model was run to examine the contribution of participant characteristics on the association between HR and challenging vs. non-challenging behaviors.
Results: On average, children with autism showed a 21±10% HR increase from baseline, 58±22 s before the onset of a challenging behaviour. The ROC analysis indicated that the peak HR change predicts fairly well the onset of a challenging behaviour vs. non-challenging behaviors (area under the curve= .71, p< .001, 95% CI= .66 - .77), see Figure 1. However, across children there was considerable variation in area under the curve coefficients (.28, p= .20; - .95, p= 04). Binary logistic regression results indicated that the behavioral outcome (challenging vs. non-challenging behavior) was explained by peak HR change (Nagelkerke R2= .21, p< .001), and additionally by participants’ gender [female] and age [older] (Nagelkerke R2 change= .06, p= .002). Autism severity and developmental ability did not significantly contribute variance to the model (see Table 1).
Conclusions: Results indicate that physiological stress predicts challenging behaviour episodes in preschoolers with autism, particularly for girls and older preschoolers. Given the recent technological advances in wearable biosensing, our results indicate that incorporating HR monitoring in intervention for autism may be helpful for some children. By signalling children’s stress, such wearables may allow parents and teachers to intervene and create learning opportunities for emotional expression and regulation. However, given the strength of the prediction and likelihood of false positives, individualised human-computer interaction and machine learning algorithms may be needed to increase the utility of including such information in moment-to-moment treatment planning.