B0172
Title: Addressing longitudinal missing data to develop an individualized treatment rule for the choice of antidepressant drug
Authors: Janie Coulombe - Université de Montréal (Canada) [presenting]
Erica Moodie - McGill University (Canada)
Susan Shortreed - (United States)
Abstract: It is unclear whether tailoring variables such as patient demographics, medication and comorbidities can be found for adapting the choice of antidepressant drugs. Previous work has found no significant tailoring variable in data from the United Kingdom. The medical health records data is accessed from Kaiser Permanente Washington (KPWA) in the United States. In those data, a summary of depressive symptoms called the patient health questionnaire (PHQ) is available and used as the clinical outcome of interest. The goal is to develop an individualized treatment rule for choosing between different antidepressant drugs to reduce the survival outcome time to a 50\% reduction in the PHQ. However, the PHQ is only partially assessed at certain points in time. A sequential multiple imputations approach is discussed and used to recover monthly values of the PHQ and the associated challenges. After imputations, dynamic weighted survival modelling (DWSURV) is used to develop an individualized treatment rule. The results are compared with those from previous studies.
