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

Publikation: Beitrag in FachzeitschriftForschungsartikelBeigetragenBegutachtung

Beitragende

  • Gregor Kohls - , Klinik und Poliklinik für Kinder- und Jugendpsychiatrie, Deutsches Zentrum für Kinder- und Jugendgesundheit (DZKJ) - Standort Leipzig/Dresden (Autor:in)
  • Erik M Elster - , Klinik und Poliklinik für Kinder- und Jugendpsychiatrie, Deutsches Zentrum für Kinder- und Jugendgesundheit (DZKJ) - Standort Leipzig/Dresden (Autor:in)
  • Peter Tino - , University of Birmingham (Autor:in)
  • Graeme Fairchild - , University of Bath (Autor:in)
  • Christina Stadler - , Universitäre Psychiatrische Kliniken Basel (Autor:in)
  • Arne Popma - , Amsterdam University Medical Centers (UMC) (Autor:in)
  • Christine M Freitag - , Universitätsklinikum Frankfurt (Autor:in)
  • Stephane A De Brito - , University of Birmingham (Autor:in)
  • Kerstin Konrad - , JARA-Brain Institute II Molecular Neuroscience and Neuroimaging (Autor:in)
  • Ruth Pauli - , University of Birmingham (Autor:in)

Abstract

BACKGROUND: Theoretical models of conduct disorder (CD) highlight that deficits in emotion recognition, learning, and regulation play a pivotal role in CD etiology. With CD being more prevalent in boys than girls, various theories aim to explain this sex difference. The "differential threshold" hypothesis suggests greater emotion dysfunction in conduct-disordered girls than boys, but previous research using conventional statistical analyses has failed to support this hypothesis. Here, we used novel analytic techniques such as machine learning (ML) to uncover potentially sex-specific differences in emotion dysfunction among girls and boys with CD compared to their neurotypical peers.

METHODS: Multi-site data from 542 youth with CD and 710 neurotypical controls (64% girls, 9-18 years) who completed emotion recognition, learning, and regulation tasks were analyzed using a multivariate ML classifier to distinguish between youth with CD and controls separately by sex.

RESULTS: Both female and male ML classifiers accurately predicted (above chance level) individual CD status based solely on the neurocognitive features of emotion dysfunction. Notably, the female classifier outperformed the male classifier in identifying individuals with CD. However, the classification and identification performance of both classifiers was below the clinically relevant 80% accuracy threshold (although they still provided relatively fair and realistic estimates of ~ 60% classification performance), probably due to the substantial neurocognitive heterogeneity within such a large and diverse, multi-site sample of youth with CD (and neurotypical controls).

CONCLUSIONS: These findings confirm the close association between emotion dysfunction and CD in both sexes, with a stronger association observed in affected girls, which aligns with the "differential threshold" hypothesis. However, the data also underscore the heterogeneity of CD, namely that only a subset of those affected are likely to have emotion dysfunction and that other neurocognitive domains (not tested here) probably also contribute to CD etiology.

CLINICAL TRIAL NUMBER: Not applicable.

Details

OriginalspracheEnglisch
Aufsatznummer105
FachzeitschriftBMC psychiatry
Jahrgang25
Ausgabenummer1
PublikationsstatusVeröffentlicht - 6 Feb. 2025
Peer-Review-StatusJa

Externe IDs

ORCID /0000-0002-1059-3415/work/177871509
ORCID /0000-0003-2408-2939/work/177871520
unpaywall 10.1186/s12888-025-06536-6
Mendeley cad151dc-0fa8-3746-9bc9-f84ddeff8694

Schlagworte

Schlagwörter

  • Humans, Adolescent, Sex Factors, Emotional Regulation/physiology, Male, Female, Emotions/physiology, Sex Characteristics, Conduct Disorder/psychology, Machine Learning, Child, Conduct disorder, Sex differences, Youth, FemNAT-CD, Emotion processing, Machine learning, Emotion dysfunction