Abstract

     Early identification of bipolar disorder may provide appropriate support and treatment, however there is no current evidence for statistically predicting whether a child will develop bipolar disorder. Machine learning methods offer an opportunity for developing empirically-based predictors of bipolar disorder. This study examined whether bipolar disorder can be predicted using clinical data and machine learning algorithms. 492 children, ages 6–۱۸ at baseline, were recruited from longitudinal case-control family studies. Participants were assessed at baseline, then followed-up after 10 years. In addition to sociodemographic data, children were assessed with psychometric scales, structured diagnostic interviews, and cognitive and social functioning assessments. Using the Balanced Random Forest algorithm, we examined whether the diagnostic outcome of full or subsyndromal bipolar disorder could be predicted from baseline data. 45 children (10%) developed bipolar disorder at follow-up. The model predicted subsequent bipolar disorder with 75% sensitivity, 76% specificity, and an Area Under the Receiver Operating Characteristics of 75%. Predictors best differentiating between children who did or did not develop bipolar disorder were the Child Behavioral Checklist Externalizing and Internalizing behaviors, the Child Behavioral Checklist Total t-score, problematic school functions indexed through the Child Behavioral Checklist School Competence scale, and the Child Behavioral Checklist Anxiety/Depression and Aggression scales. Our study provides the first quantitative model to predict bipolar disorder. Longitudinal prediction may help clinicians assess children with emergent psychopathology for future risk of bipolar disorder, an area of clinical and scientific importance. Machine learning algorithms could be implemented to alert clinicians to risk for bipolar disorder.

Introduction

     Pediatric Bipolar Disorder (BP disorder) is a prevalent and morbid disorder estimated to affect at least 2% of youth (Van Meter et al., 2011). Individuals with BP disorder often present subsyndromal symptoms of mood dysregulation during their childhood that eventually develop into a full diagnosis. The full syndromatic diagnosis of BP disorder is associated with increased risks of suicide, substance use disorders, hospitalization, and social dysfunctions for the patients and their family (De Crescenzo et al., 2017; Faedda et al., 1995; Serra et al., 2017). Although longitudinal studies have found the prognosis of early-onset mood disorders to be unfavorable, research has also shown there are effective treatments and therapies that could significantly alleviate the patients’ and their families’ struggles from the diagnoses (DelBello et al., 2022; Pavuluri et al., 2005; West et al., 2014). Thus, early identification of the risks and interventions for early symptoms of pediatric mood disorders is crucial. However, uncertainties remain on how to best predict the development of BP disorder in youth with emergent psychopathology referred to clinical practice (Faedda et al., 1995; Leverich et al., 2007).

Results

Sensitivity and specificity of the model in predicting final BPD status
The model computes the probability that a child will develop BP-I disorder. When using a probability of 0.5 or greater to predict the onset of BP-I disorder, the model accurately predicted the development of BP-I disorder with a median sensitivity of 75% and median specificity of 76%. The area under the receiver operating characteristic curve (ROC-AUC) was with median of 75% and F1 score was with median of 79% (Table 2). Our model has a false positive rate of 21.6% and a false negative rate of 3.1%.

     The average precision-recall curve of all bootstrapped iteration is presented in Fig. 1. This curve plots the positive predictive power against the sensitivity for each possible cut point on the model’s output probability. The ROC-AUC curve is presented in Fig. 2.