Individual Average Treatment Effect Estimation Using Real-World Observational Data from Wearables

• Activity Number: 470 - Beyond Precision Medicine: Making It Personal with N-of-1 and Single Case Methods for Medicine, Rare Diseases, Digital Health, Behavior, and Wearables
• Type: Topic Contributed
• Date/Time: Wednesday, August 10, 2022 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Eric J. Daza* and Logan Schneider
• Companies: Evidation Health and Alphabet Inc.
• Keywords: causal inference; time series; wearable; longitudinal; digital; personalized

Abstract

Temporally dense single-person "small data" have become widely available thanks to mobile apps and wearable sensors. Many caregivers and self-trackers want to use these data to help a specific person change their behavior to achieve desired health outcomes. Ideally, this involves discerning possible causes from correlations using that person’s own observational time series data. In this paper, we estimate within-individual average treatment effects of sleep duration on physical activity, and vice-versa. We introduce the model-twin randomization (MoTR; “motor”) and propensity score twin (PSTn; “piston”) methods for analyzing one year of the authors’ sleep duration and step count Fitbit data. MoTR is a Monte Carlo implementation of the g-formula (i.e., standardization, back-door adjustment); PSTn implements propensity score inverse probability weighting. Both do so under serial interference. They estimate idiographic stable recurring effects, as done in n-of-1 trials and single case experimental designs. We compare our approaches to standard methods (with possible confounding) to show how to use causal inference to make truly personalized recommendations for health behavior change.