AI Algorithm Enhances Prediction of Adolescent Self-Harm and Suicide Attempts

JJohn September 4, 2023 8:17 AM

Researchers have utilized a machine learning model to accurately predict adolescents' risk of self-harm and suicide attempts, offering a more precise tool than existing methods. The model also highlighted the significant role played by environmental factors in these risks.

Machine learning improves suicide risk predictions

Researchers have tapped into the potential of machine learning to provide more accurate predictions of adolescents' risk of self-harm or suicide attempts. The algorithm analyzes a multitude of factors beyond just past attempts, which are often the only risk factor considered in traditional approaches. The researchers believe this method could facilitate more personalized care for vulnerable patients, especially during adolescence, a critical and often challenging formative period.

The study drew on data from 2,809 adolescents, part of the Longitudinal Study of Australian Children. Using this data, the researchers were able to identify more than 4,000 potential risk factors related to mental health, physical health, relationships, and school and home environment, among others. The complexity of this data underscores the multifaceted nature of suicide risk and the necessity of a machine learning approach to fully understand it.

Random forest algorithm identifies key factors

The researchers used a random forest (RF) algorithm to analyze the data, identifying key predictive factors of suicide and self-harm attempts among the adolescents. This algorithm, a type of supervised machine learning model, combines the output of multiple decision trees to reach a single, more accurate result. By considering a wider range of factors, the model could offer a deeper understanding of the risks involved.

Machine learning model outperforms traditional methods

The machine-learning model outperformed traditional prediction methods based solely on a history of self-harm or suicide attempts. The model trained to predict self-harm showed a fair predictive performance with an area under the curve (AUC) of 0.740, while the model predicting suicide attempts achieved an AUC of 0.722. Compared to traditional methods, which achieved AUCs ranging from 0.630 to 0.647, the machine learning models show significant promise.

Environmental factors play crucial role in risk prediction

The machine learning model highlighted several key factors as top predictors of self-harm and suicide attempts. These included elements from the Short Mood and Feelings Questionnaire (SMFQ) and Strengths and Difficulties Questionnaire (SDQ), which assess depression symptoms and behavior/emotions respectively. Additional factors such as stressful life events, puberty scales, the child-parent relationship, a sense of belonging to school, and whether the child had a romantic partner were also identified. Notably, home and school environment played a significant role, suggesting that environmental factors are crucial in predicting these risks.

The research also uncovered unique predictors for suicide and self-harm. Lack of self-efficacy, where a person feels a lack of control over their environment and future, was identified as a unique predictor of suicide. Meanwhile, lack of emotional regulation was a unique predictor of self-harm. These findings challenge the stereotype that poor mental health is the only driving factor behind suicide or self-harm, and emphasize the need for individualized care and risk assessments.

Further research needed for clinical application

While the machine learning model shows potential for individualized risk assessments, more research is needed before it can be utilized in a clinical setting. The researchers plan to apply the models to real-life clinical databases to validate their effectiveness in predicting self-harm and suicide attempts. The ultimate goal is to integrate this technology into electronic medical records systems, allowing clinicians to access and use the data to enhance their assessments and patient care.

More articles

Also read

Here are some interesting articles on other sites from our network.