A1488
Title: Bayesian inference for cluster-randomized trials with multivariate outcomes with missing data
Authors: Guangyu Tong - Yale University (United States) [presenting]
Abstract: Cluster-randomized trials (CRTs) with fragile populations often face complex attrition, in which missing outcomes arise from heterogeneous causes: participants may be alive, deceased, or of unknown status, each with distinct missing-data mechanisms. Existing methods address death truncation but cannot jointly handle dropout unrelated to mortality or unknown survival. We propose a Bayesian framework to estimate survivor average causal effects in CRTs while accounting for multiple missingness types. Our approach models multivariate outcomes, producing posterior estimates that distinguish individual- and cluster-level survivor effects. Simulation studies demonstrate low bias and high coverage across varied scenarios. We illustrate the method using data from a geriatric CRT, focusing on a bivariate continuous outcome, though the framework readily extends to multiple or alternative endpoints (e.g., binary). We offer a general modeling strategy for handling complex missingness in CRTs, broadly applicable in aging and palliative care research.
