An international study co-led by Joaquim Raduà, head of the Imaging of Mood- and Anxiety-related Disorders (IMARD) group at IDIBAPS, together with Gonzalo Salazar de Pablo (King's College London), Daniel Guinart (Hospital del Mar) and Christoph Correll (Charité - Universitätsmedizin, Berlin) have developed a new tool capable of estimating the risk that an adolescent hospitalized for mental health issues will develop bipolar disorder in the years following that hospitalization. The study, published in Molecular Psychiatry, uses machine learning techniques to generate a risk calculator that, according to the authors, can contribute to the early detection of this disorder and therefore to optimizing clinical monitoring.
Bipolar disorder is a serious mental illness characterized by intense mood swings, alternating between episodes of depression and hypomania or mania. It usually takes years to be diagnosed, which makes early intervention difficult. In order to address this challenge, the research team followed 105 adolescents hospitalized for affective, anxiety, or conduct disorders, but without a diagnosis of bipolar disorder or psychosis, and monitored them for up to five years. The cumulative risk of developing bipolar disorder was 5% over the first year, 22% over the first two years, 29% over the first three years, and 36% over the first four years.
A predictive tool based on clinical data
From this follow-up, the researchers applied advanced statistical models to identify the factors that best anticipated the transition to bipolar disorder.
Among the symptoms that contribute most to predicting this risk are mild but significant manifestations of hypomania, such as:
-
Inflated self-esteem or grandiosity (which was the most potent predictor)
-
Accelerated thoughts
-
Excessive speech
-
Increased energy
These elements allow the new calculator to estimate an individual's risk of developing bipolar disorder, especially during the two years following hospitalization, where it has an accuracy rate of between 72% and 86% (depending on the type of validation).
As Joaquim Raduà explains, “our tool allows us to identify adolescents who, although they do not have bipolar disorder, present a clinical pattern associated with a higher probability of developing it. Detecting these cases in advance can make a difference in terms of greater prevention as well as earlier and therefore more effective treatment.”
Towards earlier and more personalized Intervention
The clinical implementation of this calculator could help mental health professionals tailor monitoring and offer preventive interventions to adolescents at higher risk. This includes psychoeducation programmes, specific psychological treatments, or closer monitoring of symptom progression.
As Gonzalo Salazar de Pablo, a co-author of the study, highlights, “this tool is not a diagnosis, but a support for clinicians. It allows us to estimate who may need closer monitoring, thus improving opportunities for prevention and early intervention.”
The research also underscores the importance of continuing to validate the model in larger cohorts and in other clinical contexts, to ensure its robustness and facilitate its adoption.
A step forward in precision psychiatry
This study represents a significant advance in the field of precision psychiatry, which seeks to introduce predictive tools to tailor treatments to the individual characteristics of each patient. Risk calculators, like the one co-developed by the IMARD group, help identify clinical trajectories before symptoms become evident, facilitating more effective interventions and potentially preventing the onset of severe disorders.
With this new tool, IDIBAPS reinforces its commitment to translational research and the improvement of mental health care for adolescents.
