Background: The rising popularity of digital health trackers presents the opportunity to understand the relationship between personal behaviors and health outcomes outside of the lab or clinic. In this study, we use longitudinal weight data and minute-level sleep data from over 6,000 users of digital health trackers across the United States to examine the relationship between sleep patterns and a clinical outcome: weight gain.
Methods: We analyzed minute-level sleep and weight data collected from fitness trackers for 11,552 users of a proprietary platform that incentivize healthier lifestyles (Achievemint.com) in the period 4/1/2015-4/1/2016. We considered users with at least 5 weight measurements spanning two or more months. Users with 5 or more nights of sleep recorded in at least two calendar months spanned by their weight measurements were included. We modeled monthly percent weight change for each user using fixed-effects panel regression analysis, performing separate regressions for each gender. We included average nightly sleep start time, sleep duration, number of naps during the day, time in bed until sleeping, number of restless episodes, as well as the standard deviation of their time asleep during the month as explanatory variables. We further included variables to control for seasonality and variation in sleep/weight measurement device and frequency.
There is a significant association between frequency of restless episodes and weight gain in the population studied
Results: 1,087 male and 5,207 female users met the inclusion criteria. The strongest effect in the regression was an association between restless episodes and weight gain. Each additional restless episode per night was associated with a 0.058 percentage point monthly increase in weight for females (95% CI [0.046, 0.072]; p < .001) and a 0.052 percentage point increase for males (95% CI [0.024, 0.081]; p < .001). On average, users had 7.8 ± 3.2 restless episodes per night.
Conclusions: There is a significant association between frequency of restless episodes and weight gain in the population studied. By individually identifying those with sleep difficulties, our results suggest it is feasible to offer personalized care that focuses on preventative risk factors. Identifying poor sleep is the first step towards offering tailored sleep therapies such as insomnia CBT, sleep hygiene, psychological, and medical care. Our results also represent a scalable and accessible means of generating clinically actionable health data given that many across the world increasingly own and use trackers and wearables.