A scoping review, published in the Journal of Medical Internet Research, found that most studies of artificial intelligence (AI) using data from wearable sensors in child psychiatry were not randomized controlled trials and had a large heterogeneity in methods.
Investigators from the Mayo Clinic searched publication databases with the goal of understanding how wearable devices were being used in child psychiatry research and how the understanding of pediatric psychiatric disorders has been enhanced by AI technologies. Publication databases were searched through June 2021 and 19 studies were included in the review.
Most studies (n=11) focused on wearable devices among patients with autism spectrum disorder (ASD). A total of 6 studies used electrocardiogram straps to track the autonomic nervous system while completing various tasks such as joint attention stimuli or imitation games. One study found an association between lower respiratory sinus arrhythmia and physiological events, suggesting improved cognitive engagement.
Another study in ASD found that aggression could be predicted 1 minute before aggressive behavior occurred when the child was equipped with a wrist-worn biosensor collecting data 3 minutes before the aggressive behavior. A study using wearable ankle-worn biosensors found that infants who would go on to be diagnosed with ASD exhibited reduced motion complexity compared with infants who were not ultimately diagnosed with ASD.
There were 5 studies which focused on attention-deficit/hyperactivity disorder (ADHD), all of which used an accelerometer device. These studies found that medium intensity movements differed among healthy children, nonmedicated children with ADHD, and medicated children with ADHD. Movement data was also shown to be suggestive of ADHD medication efficacy.
There were 3 studies which investigated internalizing disorders (IDs). They found that children with IDs were worn out more quickly than healthy children while completing a bubble task. Wearable sensors were found to identify depressive, anxious, and trauma-related disorders with a 75% accuracy or to classify behavioral states with a 68% accuracy.
This review may have been limited by not examining the AI technologies in detail and approaches in the underlying studies.
“Our scoping review found large heterogeneity of methods and findings in AI studies in child psychiatry. Overall, the largest gaps identified in this scoping review are the lack of randomized controlled trials, as most studies available were pilot trials. […] Given the growing ubiquity of wearables across the age span (children to parents, guardians, or teachers), our review strongly suggests the incorporation of wearables in child and adolescent psychiatry research […] would be pivotal in facilitating remote monitoring and remote psychiatric services, which will likely help reduce disparities in mental health care access because of a shortage of child and adolescent psychiatrists,” concluded the review authors.
Disclosure: An author declared affiliations with industry. Please refer to the original article for a full list of disclosures.
Welch V, Wy TJ, Ligezka A, et al Use of mobile and wearable artificial intelligence in child and adolescent psychiatry: scoping review. J Med Internet Res. 2022;24(3):e33560. doi:10.2196/33560