Getting RID of the blues: Formulating a Risk Index for Depression (RID) using structural equation modeling

Area of Science:

• Psychiatry and Mental Health
• Computational Statistics
• Epidemiology

Background:

• Known risk factors for depression are numerous, yet a widely utilized, comprehensive depression risk index is lacking.
• Existing research highlights demographics, lifestyle, diet, biomarkers, and somatic symptoms as key determinants of depression.

Purpose of the Study:

• To develop a flexible, modular Risk Index for Depression (RID).
• To utilize structural equation models (SEMs) blending machine learning and traditional statistics to identify depression determinants.

Main Methods:

• Employed generalized SEMs with logistic regression on National Health and Nutrition Examination Study data (N=5546).
• Utilized demographic, clinical, and laboratory variables, splitting data into training and validation sets.
• Assessed model fit using goodness-of-fit statistics and Area Under the Receiver Operator Characteristic Curves (AUROC).

Main Results:

• The SEM identified diet (OR=4.09), lifestyle-environs (OR=2.15), and somatic symptoms (OR=2.10) as significant direct predictors of depression.
• Demographics (OR=1.46) and biomarkers (OR=1.39) showed a potential indirect influence via somatic symptoms and biomarkers.
• Achieved good predictive performance with AUROC values of 0.850 (training) and 0.813 (validation) for logistic regression.

Conclusions:

• The developed modular Risk Index for Depression (RID) methodology offers flexibility in incorporating diverse risk determinants.
• The RID shows promise for clinical application by indicating patient predisposition to depression based on key determinants.
• The methodology is adaptable for developing risk indices for other affective disorders.