Epidemiology-of-1 at SBM 2021

Dr. Eric J. Daza will be presenting preliminary findings on the individualized effects of sleep duration on physical activity at the Society of Behavioral Medicine (SBM) 42nd Annual Meeting, a virtual conference happening April 12-16, 2021. This is work he’s been doing with Dr. Logan Schneider, as we first reported in Sleepless in Silico(n Valley).

This project is based on Daza’s preprint Person as population: A longitudinal view of single-subject causal inference for analyzing self-tracked health data that uses the theory he developed in Causal analysis of self-tracked time series data using a counterfactual framework for N-of-1 trials. See the Resources page for related publications.

This is a fantastic opportunity for Stats-of-1 to showcase the utility of the n-of-1 design for discovering possible recurring causal relationships using wearable device activity data. From the SBM 2021 Annual Meeting homepage:

SBM is a vibrant, multidisciplinary society focused on the role of behavior in improving health. Most of our country’s most daunting health challenges have behavioral origins, and achieving SBM’s vision of “better health through behavior change” has never been more important. SBM members include the nation’s leading scientists and practitioners, and SBM annual meetings are the premier forum for the most influential behavioral medicine research. Learn more about SBM and who we are.

SBM’s meeting is the only truly multidisciplinary behavioral medicine conference. SBM’s membership includes approximately 2,400 behavioral and biomedical researchers and clinicians from more than 20 disciplines and across academic and industry settings.

Here are the Stats-of-1 talk details:

• Title: Epidemiology-of-1: Causal Inference via Single-case Observational Design for Sleep and Physical Activity Wearables Data
• Authors: Eric J. Daza, DrPH, MPS (Evidation Health); Logan Schneider, MD (Stanford Medicine, Alphabet)
• Abstract: Temporally rich single-subject health data have become increasingly available thanks to wearable devices, mobile apps, sensors, and implants. Many health caregivers and “self-trackers” want to use such information to help a specific person figure out how to change their behavior to achieve desired health outcomes. However, this requires an approach for discerning possible causes from correlations using that person’s own observational time series data. In this paper, we posit and estimate some plausible idiographic average treatment effects of sleep duration on physical activity. We use a recently developed causal inference framework based on n-of-1 randomized trials to analyze one year of the lead author’s Fitbit sleep duration and step count data. We then compare our findings to those of standard methods that do not account for confounding to show that causal inference is needed to make realistic recommendations for personal behavior change.
• Presentation Format: pre-recorded 2-minute “Research Spotlight”

We’ll try to post a copy of the slides and talk here on Stats-of-1 after the conference. Hope to see you there!