Causal Analysis of Bipolar Disorder
Student
Hunter Chavis-Blakely
Advisor
Erich Kummerfeld
Abstract
This project takes a look at the causal effects of patient behaviors such as smoking, alcohol use, and drug use on the wellbeing of individuals with bipolar disorder. The outcomes of interest can be represented as a continuous end of follow up outcomes. Parametric G-formula was employed to estimate the causal effects of treatment(in this case treatments being patient behaviors) on the mean response for the continuous end of follow up outcome. The sample of data used in the study comes from the National Institute of Mental Health’s STEP-BD (Systematic Treatment Enhancement Program for Bipolar Disorder). Potential model treatments include patient smoking status, patient cigarette consumption, and patient current experience alcohol/drug dependency. Potential model time-varying confounders consist of medication prescriptions. Potential model time-fixed covariates include number of mood episodes patient experienced in year leading up to study, patient’s socioeconomic status, and patient’s sex. Potential model outcomes include depression symptom severity, manic symptom severity, quality of life assessment, ability to function impairment assessment, and patient’s instances of wellness in a given interval.