BACKGROUND
Background: Major depressive disorder (MDD) and its severity have been steadily increasing worldwide, especially in the Covid-19 era. Traditional clinical interviews coupled with questionnaires do not provide a complete clinical picture that includes the physiological process of inflammation and oxidative stress, which necessitates a more holistic approach to MDD diagnostics. Determining the level of depression can be problematic when adding a large number of factors contributing to MDD occurrence. Including physiological markers such as oxidative stress that are also indicative of Covid-19 infection may provide an adjunct basis for the classification of depression severity in addition to psychiatric/psychological reviews.
OBJECTIVE
The aim of this study is to develop a prediction model to determine the level of depression and severity using oxidative stress that is also indicative of Covid-19 infection in addition to clinical and clinical and socio-demographic features.
METHODS
Machine learning models included oxidative stress biomarkers, socio-demographic and health-related features in a Logistic Regression (LR), Random Forest (RF), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Naïve Baye (NB) and Artificial Neural Network (ANN) machine learning application. The Patient health Questionnaire (PHQ-9) score was used as a measure of depression severity with a total of 830 participants in the study.
RESULTS
The results show that oxidative stress biomarkers are included in the classification models, and the classifier can predict with an AUC greater than 90%, providing a novel insight into MDD severity and a link to possible effects due to Covid-19 pathophysiology.
CONCLUSIONS
It was possible to effectively predict the severity of depression through various machine learning algorithm techniques by using oxidative stress indicators integrated with health and demographic features.
CLINICALTRIAL
Ethics approval number: 2006/042