Mental health and peer relationship problems in preterm born adolescents: Which factors predict absence of symptoms?

Abstract

Background

Preterm birth is associated with difficulties in mental health and peer relationships in adolescence; however, most preterm adolescents do not experience these difficulties.

Objective

To apply machine learning models to identify key early predictors of better mental health and peer relationships in preterm adolescents.

Methods

The participants of the current study included 1472 preterm and 16,389 full-term individuals from the UK Millennium Cohort Study (2000−02). Early factors included a range of measurements across the following broad categories in infancy and early childhood: sociodemographic, family structure and environment, child-related birth and infancy factors, and early childhood factors. Mental health and peer relationships were assessed at 11, 14, and 17 years using the Strengths and Difficulties Questionnaire.

Results

The prediction model in preterm born adolescents had the highest accuracy for 17 years of age and in hyperactivity/inattention disorders (75 %, 82.7 %, 92 %, at 11, 14 and 17 years respectively) and conduct/oppositional disorders (80 %, 78 %, 87.1 %, respectively). A similar pattern was found in full-term born adolescents. Family structure and environment related factors in early childhood contributed to better mental health and peer relationships problems in both preterm and full-term adolescents. In preterm born adolescents, motor skills in infancy and better cognitive development and emotional regulation in early childhood predicted better mental health and peer relationships.

Conclusions

This study suggests that machine learning can help paediatricians differentiate preterm children who will not develop mental health symptoms and peer relationship problems from those at risk for developing these problems in adolescence.