Machine learning reveals sex differences in distinguishing between conduct-disordered and neurotypical youth based on...

Area of Science:

• Neuroscience
• Developmental Psychology
• Psychiatry

Background:

• Conduct disorder (CD) is more prevalent in boys, with theoretical models implicating emotion recognition, learning, and regulation deficits.
• Existing research has not supported the
• differential threshold
• hypothesis, which posits greater emotion dysfunction in conduct-disordered girls compared to boys.

Purpose of the Study:

• To investigate sex-specific differences in emotion dysfunction among youth with conduct disorder (CD) using machine learning (ML).
• To compare ML-based classification of CD status by sex using neurocognitive emotion features.

Main Methods:

• A multivariate ML classifier was employed to analyze multi-site data from 542 youth with CD and 710 neurotypical controls (ages 9-18).
• Emotion recognition, learning, and regulation tasks data were used to distinguish between CD and control groups, analyzed separately for males and females.

Main Results:

• Both male and female ML classifiers accurately predicted CD status above chance levels based on emotion dysfunction features.
• The ML classifier for females demonstrated superior performance in identifying individuals with CD compared to the male classifier.
• Classification accuracy for both sexes was below the clinically significant 80% threshold, likely due to sample heterogeneity.

Conclusions:

• The findings confirm a strong association between emotion dysfunction and CD in both sexes, with a more pronounced link in girls, supporting the "differential threshold" hypothesis.
• The study highlights the significant neurocognitive heterogeneity within conduct disorder, suggesting that emotion dysfunction characterizes only a subset of affected individuals.
• Other neurocognitive domains likely contribute to the etiology of conduct disorder beyond emotion processing deficits.