Physical Activity is Associated with the Risk of Diabetes Onset: A Study from the All-of-US EHR data

Abstract

Background: Type II diabetes (T2DM) is a chronic condition that affects how the body regulates blood sugar (glucose). As the prevalence of this disease continues to rise, understanding its underlying mechanisms, improving treatments, and preventing its onset has become increasingly important. T2DM is often linked to lifestyle factors, which motivates us to analyze data from the All-of-Us program. This program is launched by the NIH, and integrates clinical, environmental, and lifestyle information from a diverse population. Specifically, we focus on Fitbit data from more than 15,620 participants, which captures daily active zone minutes, through wearable devices. Objective: This study aims to describe and test the different longitudinal activity patterns among patients with and without diabetes. In addition, we also assess the impact of the activity patterns on the risk of developing T2DM. Methods: To measure the weekly active level of each individual, we use Fitbit data to calculate each individual's time of staying moderate-to-vigorous activity level in a week. The outcomes of T2DM incidents were selected from the electronic health records by a well-developed proprietary algorithm. To overcome the issue of low T2DM incident rate (2.25%) in the cohort, we conducted a nested case-control study based on basic demographic variables (age groups, sex at birth, race, ethnicity, and BMI categories) and lifestyle variables (smoking behaviors and alcohol drinking frequencies) for robust analysis. We compared the functional predictor Occupation Time Curves (OTC) between the case and control groups to explore the differences in their physical activity patterns. We also conduct t-test to test the difference statistically. Additionally, to assess the relationship between activity distributions over time and T2DM risk, we summarize the participants' cumulative activity patterns over time by calculating the area under the curve (AUC) for OTC. We fit the Cox proportional hazard model on participants' average active times using their AUC of OTC as covariates. Results: Both the OTC plot and the t-test results show that the non-diabetes group spends more time on higher weekly active levels compared to the diabetes group. (AUC_{case}-AUC_{controls} = -0.002, p = 0.08). The results from the Cox proportional hazard model show a significant relationship (p = 0.01) between an increased weekly activity level and decreased estimated risks of developing diabetes. Conclusions and discussions: Our findings suggest that individuals who engage in higher levels of moderate-to-vigorous physical activity over time have a lower risk of developing T2DM. In future research, we are also interested in the stability of the active level over a long period of time, and its impact on diabetes risk.