Predicting aggression to others in youth with autism using a wearable biosensor.

Only 20 participants from one inpatient site were studied, so the results may not apply to other settings or broader youth with autism. The authors note possible overfitting because the model design reduced opportunities for cross-training across time windows. Aggression frequency and duration varied greatly across participants. This makes it hard to determine how much training data drove prediction performance. The study tested prediction, not intervention; the authors say effectiveness of a biosensor-based system could be evaluated in future work. Most participants were minimally verbal and had severely impaired adaptive skills, so findings may not generalize to more verbal or outpatient youth.

The paper lists NIH extramural, non-U.S. government, and U.S. government non-PHS research support; no author conflict-of-interest declaration appears in the supplied text.

Do not read the 84% average accuracy as proof that alerts will prevent aggression. The study tested prediction in 20 inpatient youth and did not test an intervention.