Abstract

In the presence of time-varying treatment exposures and confounders, Marginal structural models with inverse probability treatment weights (MSM-IPTWs) are used to control for confounder variable imbalances in the estimation of causal effects in longi tudinal observational studies. However, these methods have largely been developed for univariate outcomes. This Ph.D. thesis develops causal inference methods that could be used in longitudinal observational studies with multiple outcomes. The proposed methods are built around MSM-IPTWs and encompass two additional elements namely; a) re-defining the weights as a product of inverse weights at each time point and b) accounting for the possible correlation between and within (serial) the multiple outcomes. The capabilities of the proposed methods are demonstrated with an application to estimate the effect of HIV positivity awareness on condom use and multiple sexual partners using the Malawi Longitudinal Study of Families and Health (MLSFH) data. The results have shown that the causal estimate of HIV positivity awareness and their associated standard errors were slightly attenuated using our proposed methods compared to the standard bivariate (condom use and multiple sexual partners) models. We recommend the use of the proposed MSM-IPTWs when estimating the causal effects of treatment exposures on multiple outcomes in longitudinal observational studies. The methods correctly control for time-varying treatment and confounders and adjust for possible dependency within and between the outcomes.

More details

School	: School of Natural and Applied Sciences
Issued Date	: 2023

Download full document