Monitoring sleep using smartphone data in a population of college students

Carsten Langholm; Andrew Jin Soo Byun; Janet Mullington; John Torous

doi:10.1038/s44184-023-00023-0

Monitoring sleep using smartphone data in a population of college students

Npj Ment Health Res. 2023 Mar 17;2(1):3. doi: 10.1038/s44184-023-00023-0.

Authors

Carsten Langholm¹, Andrew Jin Soo Byun^{1

2}, Janet Mullington³, John Torous⁴

Affiliations

¹ Division of Digital Psychiatry, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, 02215, USA.
² John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, 02138, USA.
³ Department of Neurology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, 02215, USA.
⁴ Division of Digital Psychiatry, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, 02215, USA. [email protected].

Abstract

Sleep is fundamental to all health, especially mental health. Monitoring sleep is thus critical to delivering effective healthcare. However, measuring sleep in a scalable way remains a clinical challenge because wearable sleep-monitoring devices are not affordable or accessible to the majority of the population. However, as consumer devices like smartphones become increasingly powerful and accessible in the United States, monitoring sleep using smartphone patterns offers a feasible and scalable alternative to wearable devices. In this study, we analyze the sleep behavior of 67 college students with elevated levels of stress over 28 days. While using the open-source mindLAMP smartphone app to complete daily and weekly sleep and mental health surveys, these participants also passively collected phone sensor data. We used these passive sensor data streams to estimate sleep duration. These sensor-based sleep duration estimates, when averaged for each participant, were correlated with self-reported sleep duration (r = 0.83). We later constructed a simple predictive model using both sensor-based sleep duration estimates and surveys as predictor variables. This model demonstrated the ability to predict survey-reported Pittsburgh Sleep Quality Index (PSQI) scores within 1 point. Overall, our results suggest that smartphone-derived sleep duration estimates offer practical results for estimating sleep duration and can also serve useful functions in the process of digital phenotyping.