Title: A marginal structural model for longitudinal observational data with multiple outcomes
Authors: Halima Twabi - University of Malawi (Malawi) [presenting]
Samuel Manda - University of Pretoria (South Africa)
Dylan Small - University of Pennsylvania (United States)
Hans-Peter Kohler - University of Pennsylvania (United States)
Abstract: Causal inference methods for observational studies are available for a single outcome and often under time-invariant treatment exposure and confounders. We propose a causal inference method for longitudinal observational data with multiple outcomes using a Marginal Structural Model (MSM) and derive an Inverse Probability Weighting (IPW) estimator for balancing the time-varying confounders. We illustrate the proposed methodology by estimating the causal effect of awareness of HIV-positivity on condom use and multiple sexual partners using individuals enrolled in the Malawi Longitudinal Study of Families and Health (MLSFH). Awareness of HIV-positivity was negatively associated with multiple partners but positively associated with condom use. We have demonstrated the considerable potential of marginal structured models for estimating causal effects in longitudinal observational studies with multiple outcomes.