Identifying suicide ideation in mental health application posts: A random forest algorithm

Area of Science:

• Digital mental health
• Machine learning applications
• Suicide risk assessment

Background:

• Digital mental health applications are increasingly used.
• Reliable early screening tools are needed to identify suicide risk among users.
• Existing methods for suicide risk detection in digital platforms require enhancement.

Purpose of the Study:

• To develop and validate a machine learning algorithm for predicting suicide ideation scores.
• To assess the accuracy and reliability of the algorithm in identifying high-risk suicide ideation posts.
• To explore a novel method for early suicide risk detection in users of digital mental health services.

Main Methods:

• A lexicon-based random forest machine learning algorithm was employed.
• The algorithm analyzed 714 online community text posts from December 2019 to April 2020.
• Predicted suicide ideation scores were validated against expert-rated scores.

Main Results:

• The algorithm demonstrated high validity in predicting suicide ideation scores.
• A low error rate was observed in the algorithm's predictions.
• The model correctly identified 95% of expert-rated high-risk suicide ideation posts.

Conclusions:

• The developed machine learning algorithm shows significant potential as an early screening tool for suicide risk.
• This method offers a reliable approach to detect suicidal ideation among users of digital mental health applications.
• Findings suggest a new avenue for improving user safety and support within digital mental health platforms.