On the Comparison of Adaptive Interventions Using Data Arising from a SMART, with Application to Autism Research

• Activity Number: 265
• Type: Invited
• Date/Time: Tuesday, August 5, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Mental Health Statistics Section
• Author(s): Daniel Almirall*+ and Xi Lu and Inbal Nahum-Shani and Connie Kasari and Susan Murphy
• Companies: University of Michigan and University of Michigan and University of Michigan and University of California, Los Angeles and University of Michigan
• Keywords: dynamic treatment regimens ; personalized medicine ; adaptive treatment strategies ; inverse probability-of-treatment weights ; multi-stage experiments

Abstract

An adaptive intervention is a sequence of individually-tailored decisions rules that specify whether, how or when to alter the intensity, type or dosage of treatment at critical decision points in the course of care. Adaptive interventions (often referred to as dynamic treatment regimens) provide clinicians with a guide for how to adapt and re-adapt treatment over time, in response to the changing needs or circumstances of the individual. They can be used as a guide to individualized (personalized) clinical practice. Sequential, multiple-assignment randomized trials (SMART) were developed explicitly for the purpose of constructing high-quality adaptive interventions using experimental design principles. In this talk, we report on recent work using a weighted-and-replicated (WR) regression estimator that can be used to compare the adaptive interventions embedded in a SMART. We will discuss simple ways to improve the statistical efficiency of the WR estimator. The methodology will be illustrated using data from a SMART study aimed at developing an adaptive intervention to improve spoken communication in minimally verbal children with autism.