Validation and extension of a risk calculator to predict mood recurrence in young people with bipolar disorder

The episodic nature of bipolar disorder (BD) and the substantial variability in mood recurrences between individuals, emphasize the need for accurate prediction tools. The original Course and Outcome of Bipolar Youth (COBY) recurrence risk calculator (RC) estimates threshold recurrence risk, but its performance for subthreshold recurrences has not been established. This study extended the COBY RC to predict both threshold and subthreshold recurrences and evaluated its performance in an independent sample.

Adolescents and young adults with BD-I/II (N = 51; BD-I: 38, BD-II: 13; 14-24 years old) were assessed with standard instruments at intake and during follow-up on average every 6 months for a median of 54 weeks. Model performance for predicting threshold and subthreshold recurrence was evaluated using area under the receiver operating characteristic curves (AUC), with additional assessments of calibration and variable importance.

The model demonstrated good discrimination of any recurrence within the next six months (threshold AUC = 0.72; subthreshold or worse AUC = 0.77). Calibration analyses indicated systematic risk overestimation in the external sample, plausibly reflecting differences in ascertainment (prospective vs. retrospective) and prior remission length. Recalibration greatly improved calibration without reducing discrimination.

The COBY RC showed good discrimination for both threshold and subthreshold recurrences in an independent young adult cohort, extending prior youth and adult validations. These findings support its potential utility for personalized monitoring and early intervention across developmental stages and symptom severity.