Date of Award
Open Access Thesis
Master of Public Health (MPH)
School of Public Health
As a secondary analysis of the VAST-D clinical trial data, we employed a multi-layered strategy to describe the complicated clinical features of Major Depressive Disorder (MDD) and the heterogeneity among depressive symptoms, using the following analytical approaches:
(1) Cluster analysis was used to transform a large and heterogenous mix of survey questions into a small number of correlated MDD symptom clusters: Four robust and highly-interpretable MDD symptom clusters (core emotional, appetite and weight, sleep disorders, atypical) were identified within the VAST-D trial, consistent with the findings from other relevant studies.
(2) Decision tree analysis was used to identify symptom thresholds with particularly effective discriminability in identifying remitters who were being treated with the three different study medications. Classification trees built for remission using a CART algorithm, were used for each of the three treatments and for the total cohort in the VAST-D study to facilitate:
(a) Generation of practical guidance that could be used to inform decision-making in real clinical settings;
(b) Identification of features for the sub-groups of patients showing low/high responses to each of the three treatments;
(c) Identification of the most important factors for remission through the use of random forests.
Chen, Yangmin, "Symptom Clustering And Decision Tree Analysis Within The Vast-D Randomized Clinical Trial" (2019). Public Health Theses. 1821.
This Article is Open Access