Developing a Prediction Model for Suicidality Among COVID-19 Patients in Korea Using Timely Data From the National...

Area of Science:

• Psychiatry
• Public Health
• Data Science

Background:

• No prior models predicted suicidality using timely data from coronavirus disease 2019 (COVID-19) patients.
• This study addresses the need for early screening tools for mental health risks in COVID-19 survivors.

Purpose of the Study:

• To develop and validate a predictive model for suicidality in patients diagnosed with COVID-19 within the past month.
• To create screening tools based on timely patient data for public mental health applications.

Main Methods:

• Analysis of data from 96,694 Korean COVID-19 patients.
• Utilized classification and regression tree (CART) and random forest models.
• Included 39 features: demographics, COVID-19 factors, and psychological symptoms (depression, anxiety, PTSD).

Main Results:

• Random forest model achieved an area under the curve of 0.85, with 73.4% sensitivity and 83.9% specificity.
• Key predictors of suicidality included suicidal ideation, fatigue, anxiety, and depressive mood.
• A scorecard with cut-off scores was developed based on the best-performing model.

Conclusions:

• A high-performing model for predicting suicidality in Korean COVID-19 patients was successfully developed.
• The developed scorecard enhances the practical application of the model in public mental health settings.
• This research provides a valuable tool for early identification and intervention of mental health issues in COVID-19 patients.