An Index of Sensitivity to Nonexchangeability

• Activity Number: 423 - Topics in Health Services Research: Meta-Analysis, Health Care Disparities, and Observational Data
• Type: Contributed
• Date/Time: Wednesday, August 10, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Md Rashedul Hoque* and Yi Qian and Lawrence McCandless and J Antonio Aviña-Zubieta and Mary A De Vera and Hui Xie
• Companies: Simon Fraser University and University of British Columbia and Simon Fraser University and University of British Columbia and Simon Fraser University and Simon Fraser University
• Keywords: local sensitivity measures; non-exchangeability; causal treatment effect; unobserved confounders; reverse causality; administrative data

Abstract

Standard statistical methods assuming exchangeability of units between treatment groups can yield biased treatment effect estimates if the assumption does not hold. Existing methods evaluate sensitivity of treatment effect estimates to non-exchangeability due to unobserved confounders only. In practice, non-exchangeability can occur for either unobserved confounders or reverse causality. We propose an index of sensitivity to non-exchangeability (ISENSE) to measure the impact of non-exchangeability on treatment effect estimates. Unlike many existing methods, ISENSE does not require imposing any assumptions regarding the distribution or number of unmeasured confounders, and it can handle both unmeasured confounders and reverse causality. ISENSE is a local sensitivity method based on a Taylor-series approximation to the non-exchangeability likelihood, evaluated at the parameter estimates under exchangeability. One can interpret ISENSE intuitively through the ‘MinNE’ statistic values that capture the minimum non-exchangeability needed to cause important sensitivity. We evaluate ISENSE using simulation studies and illustrate its use with an example using administrative data.