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

Abstract

Objective: While risk factors for depression are increasingly known, there is no widely utilised depression risk index. Our objective was to develop a method for a flexible, modular, Risk Index for Depression using structural equation models of key determinants identified from previous published research that blended machine-learning with traditional statistical techniques. Methods: Demographic, clinical and laboratory variables from the National Health and Nutrition Examination Study (2009–2010, N = 5546) were utilised. Data were split 50:50 into training:validation datasets. Generalised structural equation models, using logistic regression, were developed with a binary outcome depression measure (Patient Health Questionnaire-9 score ⩾ 10) and previously identified determinants of depression: demographics, lifestyle-environs, diet, biomarkers and somatic symptoms. Indicative goodness-of-fit statistics and Areas Under the Receiver Operator Characteristic Curves were calculated and probit regression checked model consistency. Results: The generalised structural equation model was built from a systematic process. Relative importance of the depression determinants were diet (odds ratio: 4.09; 95% confidence interval: [2.01, 8.35]), lifestyle-environs (odds ratio: 2.15; 95% CI: [1.57, 2.94]), somatic symptoms (odds ratio: 2.10; 95% CI: [1.58, 2.80]), demographics (odds ratio:1.46; 95% CI: [0.72, 2.95]) and biomarkers (odds ratio:1.39; 95% CI: [1.00, 1.93]). The relationships between demographics and lifestyle-environs and depression indicated a potential indirect path via somatic symptoms and biomarkers. The path from diet was direct to depression. The Areas under the Receiver Operator Characteristic Curves were good (logistic:training = 0.850, validation = 0.813; probit:training = 0.849, validation = 0.809). Conclusion: The novel Risk Index for Depression modular methodology developed has the flexibility to add/remove direct/indirect risk determinants paths to depression using a structural equation model on datasets that take account of a wide range of known risks. Risk Index for Depression shows promise for future clinical use by providing indications of main determinant(s) associated with a patient’s predisposition to depression and has the ability to be translated for the development of risk indices for other affective disorders.