Summary: A new international study shows that emotional and behavioural difficulties in five-year-old children can be predicted using risk factors present during pregnancy and the newborn period. Using artificial intelligence to analyse data from nearly 6,000 children in a UK longitudinal study, researchers identified a set of early-life indicators—including maternal smoking during pregnancy, low birth weight and lack of breastfeeding—that reliably predict later emotional and behavioural problems. The analysis also uncovered important sex-specific differences: boys appeared more vulnerable to the effects of maternal smoking, while girls were more strongly affected by early infant fussiness.
These findings underline the value of early, preventive strategies and the potential for gender-responsive screening tools. Experts involved in the research emphasize acting early, even before symptoms appear, to support children’s mental health trajectories.
Key Facts:
- Predictive AI: Machine learning methods identified 14 prenatal and neonatal risk factors that help predict emotional and behavioural difficulties by age five.
- Gender Differences: Boys showed greater sensitivity to maternal smoking during pregnancy, while girls were more affected by infant fussiness and regulatory problems.
- Early Action: The results support investments in preventive care and monitoring starting in pregnancy and continuing through the newborn period.
Source: University of Helsinki
The researchers identified strong and often overlooked early risk factors linked with emotional and behavioural difficulties at age five.
“Lack of breastfeeding, low birth weight and maternal smoking during pregnancy are factors that can be used to react very early, even before the child shows symptoms,” says the study’s lead author, doctoral researcher Xu Zong from the University of Helsinki.
Published in the Journal of Affective Disorders, the study applied advanced artificial intelligence techniques to data drawn from the UK Household Longitudinal Study. The research team set out to determine whether information available during pregnancy and the newborn period could reliably predict emotional and behavioural difficulties when children reach five years of age.
The AI-driven analysis allowed the researchers to detect complex relationships among many variables and to rank the relative importance of different risk factors. While the three most significant predictors were lack of breastfeeding, low birthweight and maternal smoking during pregnancy, the model also highlighted infant regulatory problems—such as excessive crying or feeding and sleeping difficulties—as meaningful contributors to later outcomes.
A notable outcome of the study was the identification of sex-specific differences in predictive risk. The models indicated that maternal smoking during pregnancy carried a stronger predictive weight for boys’ later emotional and behavioural difficulties, whereas measures of fussing and regulation in infancy were more predictive for girls. These differences point to the potential benefit of tailoring early screening and intervention strategies to account for gender-specific vulnerability patterns.
“Our findings are particularly relevant in a time when concern for children’s mental health and demand for early intervention are increasing,” Zong adds. The authors argue that adding targeted screening based on prenatal and neonatal risk profiles could help identify children at higher risk long before problematic behaviours fully emerge.
Beyond the immediate prediction of risk, the study reinforces broader public health priorities: investing in prenatal care, reducing maternal smoking, promoting breastfeeding where possible, and providing early support for newborn regulatory problems. Such measures could contribute to better mental health outcomes across childhood by addressing modifiable risk factors early in life.
The research was led by the University of Helsinki in collaboration with Stockholm University, Karolinska Institutet (Sweden) and the University of Essex (UK). By combining large-scale longitudinal data with interpretable machine learning methods, the team produced findings intended to inform both clinical screening practices and public health policy.
About this AI and neurodevelopment research news
Author: Xu Zong
Source: University of Helsinki
Contact: Xu Zong – University of Helsinki
Image: The image is credited to Neuroscience News
Original Research: Open access. “Predicting children’s emotional and behavioural difficulties at age five using pregnancy and neonatal risk factors: Evidence from a longitudinal study of UK households” by Xu Zong et al., Journal of Affective Disorders.
Abstract
Predicting children’s emotional and behavioural difficulties at age five using pregnancy and neonatal risk factors: Evidence from a longitudinal study of UK households
Emotional and behavioural difficulties in childhood can significantly affect later life outcomes, which makes identifying early-life risk factors that predict these difficulties an important public health goal. Much prior research has focused on diagnosing problems after they appear or on adolescent populations, leaving a gap in predictive work for younger children.
This study applies machine learning techniques to construct an interpretable predictive model based on data from the UK Household Longitudinal Study. The research examined maternal behaviours during pregnancy and parent-reported information on birth outcomes, breastfeeding, and newborn regulatory problems to determine which factors most strongly predict emotional and behavioural difficulties by age five.
The analysis identified lack of breastfeeding, low birth weight and maternal smoking during pregnancy as the three most influential predictors. Infant regulatory issues also emerged as important. Further heterogeneity analysis revealed gender differences: maternal smoking during pregnancy was a stronger predictor for boys, while fussiness and regulatory problems in infancy had a larger predictive impact for girls.
The study emphasizes the need for comprehensive prenatal and postnatal care, supports the case for early screening based on readily available risk information, and highlights the potential value of gender-specific assessment and intervention strategies to better prevent and address emotional and behavioural difficulties in children.