UCSF

Psychiatric Phenotyping Using Symptom Profiles: Can Self-Report Symptoms Inform a New Psychiatric Taxonomy?

by

Jessica Ross

The Diagnostic and Statistical Manual (DSM) has served as the gold standard for psychiatric diagnosis for the past several decades in the United States, and it mirrors mental health and substance abuse diagnoses in the ICD-9 and ICD-10, which are used in numerous other countries. However, DSM diagnoses have severe limitations when used as phenotypes for studies of the pathophysiology underlying mental disorders, as well as for clinical treatment and research. This dissertation proposes a novel approach for deconstructing DSM diagnostic criteria using expert knowledge to inform feature selection for unsupervised machine learning. A multimodal dataset comprised of combat veterans, approximately one-third of whom had received a DSM-IV diagnosis of post-traumatic stress disorder (PTSD), is used in these analyses. Unsupervised learning methods are employed to identify robust groups of patients who clustered together with respect to clinical symptoms. Symptom profiles are used to stratify subjects into cohorts who have clinical and biological homogeneity, irrespective of their DSM diagnoses. Clusters identified suggest that prior contrasting biomarker findings in patients with PTSD may be due to heterogeneity that is reduced when using phenotypes derived from self-report psychiatric symptoms. Results of these analyses can be represented in rich clinical phenotypes that relay both clinical and biological markers of interest. These findings suggest that itemized self-report symptom data may be useful to inform a new taxonomy for psychiatry, enhancing the bidirectional translation of knowledge from the bench to the clinic through a common terminology.