An App for Classifying Personal Mental Illness at Workplace Using Fit Statistics and Convolutional Neural Networks:...

Area of Science:

• Occupational health
• Psychiatry
• Artificial intelligence in healthcare

Background:

• Mental illness (MI) is prevalent among healthcare professionals.
• The relationship between workplace mental status and MI requires further investigation.
• A mobile application is proposed for early MI detection.

Purpose of the Study:

• To develop a predictive model for automatic detection and classification of workplace MI.
• To utilize convolutional neural networks (CNNs) and fit statistics for model building.
• To create a mobile app for staff to self-assess mental status using the Emotional Labor and Mental Health (ELMH) questionnaire.

Main Methods:

• Recruited 352 respiratory therapists (RTs) in Taiwan.
• Administered the 44-item ELMH questionnaire.
• Employed exploratory factor analysis (EFA), Rasch analysis, and CNN for classification and model development.

Main Results:

• Identified 8 key domains in the ELMH questionnaire via EFA.
• Classified RTs into 4 mental health categories using Rasch analysis.
• Achieved a high accuracy rate of 0.92 with the 44-item CNN model.
• Successfully developed and demonstrated a web-based MI detection app for RTs.

Conclusions:

• The 44-item CNN model (108 parameters) enhances mental health assessment accuracy for RTs.
• The developed MI app facilitates early self-detection of work-related mental illness.
• Further efforts are needed to increase the availability and viability of the MI app.