A comprehensive new study has revealed significant shortcomings in current prediction models for attention-deficit hyperactivity disorder (ADHD). Despite the growing interest in using these models to improve the diagnosis and treatment of ADHD, none of the existing models is yet fit for implementation in clinical practice.
The systematic review, published in Molecular Psychiatry, examined 100 prediction models that have been developed over the years to support personalised detection, prognosis, and treatment of ADHD. These models, which use a variety of predictors, including clinical, neuroimaging, and cognitive data, aim to provide a more precise and tailored approach to managing the disorder. However, the review found that most of these models are flawed, particularly in their ability to be applied in real-world clinical settings.
One of the key findings of the review is that while 96% of the models had undergone internal validation, only 7% had been externally validated. This lack of external validation is a critical barrier to the widespread adoption of these models in clinical practice, as it raises concerns about their generalisability and reliability when applied to different populations.
Moreover, the review highlighted that only 8% of the models were considered to be at low risk of bias, with the majority deemed to have a high risk of bias. This is particularly concerning in the analysis domain, where 61% of the studies were found to be at high risk of bias. Such issues undermine the credibility of the prediction models and further delay their potential integration into clinical settings.
Interestingly, the review also noted that models incorporating clinical predictors tended to perform better than those that did not. This suggests that while complex multimodal models might seem promising, simpler models focusing on clinical data might actually be more effective and easier to implement.
Despite these challenges, the review underscores the growing interest in prediction science within the field of ADHD. The number of prediction models has increased rapidly in recent years, reflecting a broader trend towards precision medicine in psychiatry. However, the review’s authors caution that more rigorous research is needed to develop high-quality, externally validated models that can truly enhance clinical practice.
The review calls for a new generation of research that addresses these gaps by focusing on the development of replicable and externally validated models. It also emphasises the importance of implementation research to ensure that once reliable models are developed, they can be effectively integrated into clinical practice.
