Primary Submission Category: Dynamic Treatment Regimes
Efficient estimation of optimal treatment rules with fused randomized trials and missing covariates
Authors: Nicholas Williams, Kara Rudolph, Iván Díaz,
Presenting Author: Nicholas Williams*
A fundamental principle of clinical medicine is that a treatment should only be administered to those patients who would benefit from it. Treatment strategies that assign treatment to patients as a function of their individual characteristics are known as dynamic treatment rules. Randomized clinical trials are considered the gold standard for estimating the marginal causal effect of a treatment on an outcome; they are often not powered to detect heterogenous treatment effects. The availability of multiple trials presents an opportunity for combining data from multiple randomized trials, often called data-fusion, to better estimate dynamic treatment rules using the combined data-set than either data-set alone. However, there may be a mismatch in the set of patient covariates measured between trials. We address this problem here; namely, we propose a nonparametric estimator for the optimal dynamic treatment rule that leverages information from the set of randomized trials with missing covariates to estimate the conditional average treatment effect. We show, under certain conditions, that the proposed estimator is more efficient than an estimator for the conditional average treatment effect that only uses the set of trials that measure all covariates. We apply the estimator to fused randomized trials of medications for the treatment of opioid use disorder to estimate a treatment rule that would match patient subgroups with the medication that would minimize risk of relapse.