Model-Twin Randomization (MoTR): A Monte Carlo Method for Estimating the Within-Individual Average Treatment Effect Using Wearable Sensors

August 2, 2022
Publications

Model-Twin Randomization (MoTR): A Monte Carlo Method for Estimating the Within-Individual Average Treatment Effect Using Wearable Sensors

August 2, 2022
Publications

Model-Twin Randomization (MoTR): A Monte Carlo Method for Estimating the Within-Individual Average Treatment Effect Using Wearable Sensors

Eric J. Daza, Logan Schneider

August 2, 2022
Publications

Model-Twin Randomization (MoTR): A Monte Carlo Method for Estimating the Within-Individual Average Treatment Effect Using Wearable Sensors

August 2, 2022
Publications
Eve: Evidation's brand mark which is a yellow glowing orb

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 physical activity on sleep duration, and vice-versa. We introduce the model twin randomization (MoTR; “motor”) method for analyzing an individual’s intensive longitudinal data. Formally, MoTR is an application of the g-formula (i.e., standardization, back-door adjustment) under serial interference. It estimates stable recurring effects, as is done in n-of-1 trials and single case experimental designs. We compare our approach to standard methods (with possible confounding) to show how to use causal inference to make better personalized recommendations for health behavior change, and analyze 222 days of Fitbit sleep and steps data for one of the authors.

INTRODUCTION: The ever-increasing abundance of frequently collected, temporally dense single-person small data has fueled the desire to extract “personalized insights” from this digital individual-level information. However, standard analyses of such intensive longitudinal data merely characterize statistical associations, making conclusions of causal effects generally hard to defend.

Causal effects are needed to make these insights truly “actionable” to the extent that one actually expects to see an impact from intervening on a measured factor. Epidemiologists and econometricians have developed and used causal inference methods to better inform population-level policies and decisions to improve societal outcomes. How can behavior change scientists likewise apply these methods to within-individual observational studies—in order to better inform individual-level behavioral interventions and habit-changing practices?

One solution is to conduct within-individual experiments via the mobile phone apps and wearable or implantable sensors that enable small-data collection. These could be simple randomized or forced crossover designs as are commonly used in within-individual studies, that do not generally require causal inference methods.

Individuals with common recurring or chronic health conditions in particular could benefit from such digitally informed “self-experiments”. These conditions include migraines, chronic pain, asthma, and irritable bowel syndrome. However, substantial barriers to self-experimentation exist for these conditions, highlighting the need to make better use of dense digital data by extracting insights that are not just “correlational” (i.e., statistically associated) but also plausibly causal.

Read the full preprint here.

No items found.