Identifying Risk Factors for Suicide Attempts in Adolescents and Building a Model Using Neural Networks for...

Area of Science:

• Pediatric Health
• Mental Health Research
• Artificial Intelligence in Medicine

Background:

• Adolescent suicide attempts present significant public health challenges.
• Laniado Hospital observed an increase in adolescent suicide attempts, necessitating a study on risk factors.
• Established risk factors for adolescent suicide attempts require continuous evaluation.

Purpose of the Study:

• To identify and characterize key risk factors associated with adolescent suicide attempts.
• To develop a predictive model for early identification of adolescents at risk of suicide attempts.
• To leverage machine learning for enhanced early detection of at-risk youth.

Main Methods:

• Retrospective study of adolescents (11-18 years) admitted for suicide attempts between 2015-2021.
• Analysis of epidemiological, medical, and psychosocial data from 82 patients and 82 matched controls.
• Development of a predictive neural network model utilizing identified significant risk factors.

Main Results:

• Key risk factors identified: living outside the home, female gender, unmarried parents, advanced age, higher hemoglobin, higher mean corpuscular volume (MCV), and prior psychiatric diagnosis.
• The neural network model achieved a high predictive accuracy of 99.85%.
• Significant associations found between suicide attempts and factors such as unmarried parents (OR=98.51) and prior psychiatric diagnosis (OR=71.82).

Conclusions:

• Prior psychiatric history, parental marital status, female gender, and living situation are critical indicators for adolescent suicide risk.
• Elevated hemoglobin and MCV levels were observed in adolescents with suicide attempts.
• The developed neural network demonstrates exceptional accuracy in identifying adolescents who have attempted suicide.