|
Activity Number:
|
383
- New Developments in Sensitivity Analysis for Unmeasured Confounding
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, July 31, 2018 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
|
Abstract #326629
|
Presentation
|
|
Title:
|
Assessing Sensitivity to Unmeasured Confounding with Multiple Treatments and a Binary or Survival Outcome: A Bayesian Approach
|
|
Author(s):
|
Liangyuan Hu* and Chenyang Gu and Michael Lopez
|
|
Companies:
|
Icahn School of Medicine at Mount Sinai and Harvard Medical School and Skidmore College
|
|
Keywords:
|
Unmeasured confounding ;
Sensitivity analysis ;
Bayesian ;
Multiple treatments
|
|
Abstract:
|
Current statistical sensitivity analysis (SA) approaches do not consider the complexity associated with the multiple treatment setting. We propose a flexible Bayesian SA framework for unmeasured confounding with multiple treatments and binary or survival outcomes. We first derive the biases in treatment effect estimates when the assumption of no unmeasured confounding is violated, with the magnitude and direction of the violation governed by sensitivity parameters. We then develop a flexible semi-parametric Bayesian approach that corrects the bias attributable to unmeasured confounding. Inferences about the sensitivity to unmeasured confounding will be obtained from the posterior distribution of the average treatment effects. Extensive simulations are carried out to investigate the joint impact of the prior distribution of the sensitivity parameters and various combinations of design factors for the multiple treatment setting, including confounding level, response surfaces and treatment assignment mechanism. Our approach is applied to a SEER-Medicare dataset to elucidate the effectiveness of robotic-assisted surgery compared to traditional approaches for lung cancer patients.
|
Authors who are presenting talks have a * after their name.
