Classifying Schizophrenia Cases by Artificial Neural Network Using Japanese Web-Based Survey Data: Case-Control Study

Area of Science:

• Epidemiology
• Machine Learning
• Psychiatry

Background:

• Estimating schizophrenia prevalence in Japan is challenging.
• Schizophrenia patients often experience comorbidities and poor well-being.
• Machine learning (ML) offers high-precision modeling for epidemiological studies.

Purpose of the Study:

• To construct an artificial neural network (ANN) model for accurate schizophrenia case classification.
• To validate the generalizability of the ANN model using Japanese web survey data.

Main Methods:

• Utilized a large Japanese web-based survey dataset (2021) with 223 schizophrenia cases and 1776 controls.
• Employed an ANN for classification, with logistic regression (LR) as a reference.
• Compared model performance using metrics like area under the receiver operating characteristic curve, accuracy, and specificity.

Main Results:

• The ANN model outperformed LR in area under the curve (0.86 vs. 0.78) and specificity (0.96 vs. 0.94).
• LR showed higher sensitivity (0.63 vs. 0.56).
• Key predictive variables included sleep medication use, age, income, and employment type.

Conclusions:

• An ANN model was successfully developed for classifying schizophrenia cases from web survey data.
• The model demonstrated high internal validity.
• Findings support improved prevalence estimation and future epidemiological research on schizophrenia in Japan.