J Korean Neuropsychiatr Assoc.  2022 Nov;61(4):297-307. 10.4306/jknpa.2022.61.4.297.

A Preliminary Study for the Development of Data-Driven Screening Tool for Mental Health Using Self-Report Questionnaires Reflecting General Functional Status

• Choi W
• Hwang WJ
• Kim H
• Kim M
• Kwon JS

• KMID: 2536670
• DOI: http://doi.org/10.4306/jknpa.2022.61.4.297

Abstract

Objectives
Mental health is not only closely related to daily functioning but also affects adaptability to issues in life. To date, however, there have been no simple and widely available screening tools that assess general mental health. Therefore, we sought to conduct a preliminary study for the development of a general mental health screening tool that can easily detect mental health problems at home.
Methods
We administered nine self-report questionnaires and the Global Assessment of Functioning (GAF) scale to 143 patients with psychosis, a typical mental illness, 125 subjects with a clinical high-risk for psychosis, a representative group with non-specific psychiatric symptoms such as depression or anxiety, and 118 healthy subjects as controls. A machine learning approach was used to identify a set of items that optimally predicted the GAF scores. A model was trained using the random forest algorithm, and the set limit was measured through ABC analysis to select the final set of questionnaire items.
Results
The results showed that 12 items from four questionnaires were included in the function prediction model; the Positive and Negative Affect Scale 1, 8, 17, 19, the Beck Depression Inventory 6, 14, the Symptom Checklist 3, 5, 15, 20, and the Affect Intensity Measure 39. The average mean squared error, which is representative of model performance, was 879.07.
Conclusion
These results suggest that these 12 self-report items could comprise a simple and easy mental health screening tool. We propose to conduct a follow-up study on a large general population to validate the clinical usefulness of this preliminary screening tool.

Keyword

Mental health; Screening tool; Machine learning; Self-report; Functioning