|Saturday, February 17|
|PS3 Poster Session 3 and Continental Breakfast||
Sat, Feb 17, 8:00 AM - 9:15 AM
Data Modeling to Mitigate the Impact of Missing Data in a Longitudinal Study of Injecting Drug Users (303658)
*Tania Amanda Patrao, University of Queensland, Australia
Robert Ware, University of Queensland, Australia
Keywords: Longitudinal Data, Statistical Modelling, Drug Users, Attrition, Lost to follow up.
Missing data is a challenge faced by most statistical practitioners. This presentation is relevant to applied statisticians working with datasets that are plagued by missingness. It presents examples of practical modelling techniques in the analysis of longitudinal data. Retention and attrition in longitudinal studies of marginalised populations is challenging and impacts on generalizability and reliability of study outcomes. Data from a prospective cohort study of 668 young injecting drug users from Victoria, Australia, who were interviewed at four time-points was analyzed to identify characteristics associated with significant patterns of loss to follow up. The majority of this population is socially disengaged, hard to access, mobile and vulnerable to homelessness and imprisonment. We analyzed loss to follow-up in this cohort using multiple methods including multinomial regression models and generalised linear models using a complementary log-log link to assess the association between individual characteristics and study drop-out. The performance of different models in analyzing attrition will be presented as illustrative results applicable to other disadvantaged populations.