UCLA Health researchers have developed machine learning models to enhance suicide-risk prediction in children, outperforming traditional methods used by health systems
Amid a national youth mental health crisis, health providers struggle to identify children at risk of suicide or self-harm.
UCLA Health researchers discovered that conventional methods used by health systems, such as diagnostic codes and chief complaints, often fail to capture crucial information, leading to incomplete data on children with self-injurious thoughts or behaviours.
Ultimately, AI models which have recently been a success in healthcare, such as machine learning and algorithm prediction offer a strong solution to this mental health crisis.
A promising solution to suicide risk
Researchers designed three machine-learning models for suicide-risk prediction to address the data gaps. The most comprehensive model incorporated 84 data points from electronic records, enabling it to outperform traditional methods.
Additionally, another model considered all mental health diagnostic codes, while a third model assessed various indicators, such as medications and lab tests.
These machine learning models demonstrated superior performance compared to ICD codes and chief complaints, offering a promising solution to improve risk prediction without complex algorithms.
Inclusive strategies and future research in mental health
The study raised concerns about gender and racial disparities, as male children, preteens, and Black and Latino youth were more likely to be underrepresented in risk prediction models.
Nonetheless, the machine learning models were seen as a positive step forward. Dr. Juliet Edgcomb, the lead author, emphasized that a comprehensive approach rather than sophisticated models could lead to better detection. Future research by Edgcomb aims to enhance youth suicide-risk prediction, especially among elementary-school-age children, where such models have been scarce.
UCLA Health researchers’ machine learning models offer hope for improved suicide-risk prediction among children, overcoming the limitations of traditional methods and paving the way for more inclusive and sensitive screening tools.
Leveraging the potential of machine learning models
By leveraging the potential of machine learning models, they have overcome the limitations of conventional data tracking methods within health systems. The comprehensive approach involving 84 data points has proven to be a game-changer, surpassing the accuracy of traditional diagnostic codes and chief complaints.
This innovative technique offers a more inclusive and sensitive screening tool for identifying children at risk of self-harm or suicide risk prediction. With further research focusing on enhancing prediction models for all age groups, this breakthrough may help tackle the growing youth mental health crisis more effectively.
