
A predictive model for early detection of psychiatric risk in adolescents
Unmet Need
According to the World Health Organization, nearly 15% of adolescents (10 to 19-year-olds) experience a mental health condition, many of which remain undiagnosed and untreated. Since the beginning of the COVID-19 pandemic, the rate of mental illness has significantly increased in the US and globally. Despite this rise, identification of individuals at high-risk of developing mental health conditions, such as depression or psychosis, remains inconsistent, relying on subjective clinical assessment or broad risk factors. Recent studies have suggested that early detection and intervention in vulnerable groups can prevent disease progression and improve clinical outcomes. However, the lack of accurate, data-driven methods to predict which patients are most likely to develop a psychiatric disorder results in missed opportunities for timely intervention. This gap hinders clinical care and places unnecessary strain on an already overburdened healthcare system, diverting care and resources away from patients at critical risk. There is a need for improved methods to accurately predict at-risk patients earlier to allow for more efficient resource allocation. Such a tool would help prioritize care for patients at need, reduce the duration of untreated mental illness, and enhance clinical outcomes.
Technology
Duke inventors have developed a model designed to predict future psychiatric risk in adolescents. This is intended to be used by clinicians to predict future psychiatric risk and identify patients in need of early intervention. Specifically, this model leverages a unidirectional recurrent neural network (RNN) to estimate an individual’s “p-factor” quartile, categorized into four risk levels: “no-risk,” “low-risk,” “medium-risk,” and “high-risk.” The p-factor, a recently proposed dimension of generalized psychopathology, reflects an individual’s risk of developing mental illness. Unlike traditional models that focus on current symptoms, this innovative model predicts risk by analyzing underlying disease mechanisms and protective factors, offering insights that current symptom-based assessments cannot provide. This has been demonstrated by using data from the ongoing Adolescent Brain and Cognitive Development (ABCD) study, one of the largest longitudinal studies of adolescent cognitive development. In validation testing, the model accurately predicted high-risk conversion, identifying individuals likely to progress from lower-risk groups to higher-risk groups. With further prospective evaluation, this model has the potential to enable early intervention, improving clinical outcomes. Because this model uses the “p-factor,” it is not specific to any one psychiatric disorder, making it broadly applicable across multiple mental health conditions. Moreover, the model leverages highly scalable questionnaires to make its predictions, avoiding the need for expensive or time-intensive assessments. This model represents a significant advancement in adolescent psychiatric risk prediction, offering clinicians a data-driven tool to identify at-risk adolescents early, enabling intervention and transforming mental health outcomes.
Advantages
-
Unbiased model to predict high-risk conversion earlier (i.e., before symptomatic)
-
Broadly applicable across psychiatric disorders
-
Extensive training data from the ABCD study (>9,000 participants)