Peer-reviewed article
Exploring the Prevalence and Key Predictors of Probable Depression Among Parents and Primary Caregivers of Children and Adolescents With Specific Learning Disorders in the Kingdom of Saudi Arabia: A Cross-Sectional Explainable Machine-Learning Study.
Almulla AA, Khasawneh MAS ·
Summary
Written by AI, not yet AICS's view · claude-haiku-4-5-20251001This study examined depression among parents and caregivers of children with specific learning disorders in Saudi Arabia. Using machine learning, researchers found that 35.6% of caregivers (218 out of 612) showed signs of probable depression. The study identified key risk factors including parenting stress, parental anxiety, sleep quality, learning disorder severity, and child behavioural difficulties. The researchers developed predictive models that could help identify caregivers at risk of depression for early referral and support.
Key points
- 35.6% of parents and caregivers of children with specific learning disorders in Saudi Arabia had probable depression
- Parenting stress and parental anxiety symptoms were among the strongest predictors of caregiver depression
- Sleep quality, severity of child's learning disorder, and child emotional/behavioural difficulties were important risk factors
- Machine-learning models achieved high accuracy (sensitivity 0.858, specificity 0.830) for identifying caregivers with probable depression
- Perceived social support and coping capacity were protective factors associated with lower depression risk
Limitations
The study was cross-sectional, so it cannot establish causal relationships or determine whether caregiver depression leads to worse child outcomes or vice versa. The findings are specific to Saudi Arabia and may not generalise to other countries with different health systems and cultural contexts.
Abstract
This study estimated the prevalence of probable depression among parents and primary caregivers of children and adolescents with specific learning disorders (SLDs) in Saudi Arabia and developed explainable machine-learning models for screening-oriented classification. In this multicentre cross-sectional study, 612 caregivers were recruited through clinical, psychoeducational, educational and community pathways. Children had formally documented SLD diagnoses established through qualified clinical, psychoeducational, school psychology or multidisciplinary services. Probable depression was defined as a Patient Health Questionnaire-9 score of 10 or higher. Forty-seven candidate predictors covered parent, child, family, psychosocial, socio-economic, stigma-related and service-access domains. Elastic-net logistic regression, random forest, XGBoost, LightGBM and a soft-voting ensemble were evaluated using stratified 10-fold cross-validation. Probable depression was identified in 218 participants (35.6%; 95% CI 31.8-39.6). XGBoost and the soft-voting ensemble achieved the highest ROC AUC (0.916); XGBoost yielded an average precision (AP) of 0.867 and a Brier score of 0.115. At a screening-oriented threshold of 0.425, XGBoost achieved sensitivity of 0.858, specificity of 0.830 and negative predictive value of 0.913. SHAP analyses identified parenting stress, parental anxiety symptoms, sleep quality, SLD severity, child emotional and behavioural difficulties, perceived social support and coping capacity as the most influential predictive features. Probable depression was common in this service- and community-recruited sample. Explainable models may support family-centred identification and referral, but external evaluation is required before implementation.