Fitness Trackers and Smartphones Used to Predict Behavior Changes in People with Multiple Sclerosis During the COVID-19 Stay-at-Home Mandate
Posted December 13, 2022
Mayank Goel, Ph.D., Carnegie Mellon University
Multiple sclerosis (MS) is a chronic neurological disease that results in a wide range of physical and mental symptoms that are debilitating and difficult to predict. Previous studies have focused on the analysis of physical disturbances, such as gait and balance. However, other comorbidities such as energy levels, sleep quality, and mental health disorders have not been well-examined within patient environments, and the lack of research in this domain is a significant unmet need – especially since mood disorders such as depression have been associated with increased morbidity and mortality, and MS patients regard them as a direct impact on their quality of life. Research in this field became increasingly important during the COVID-19 pandemic, as the lockdowns and social distancing associated with it had significant negative physical and mental impacts for those with MS and other chronic diseases.
The CDMRP Multiple Sclerosis Research Program (MSRP) funded a Fiscal Year 2019 (FY19) Exploration - Hypothesis Development Award (EHDA) to Dr. Mayank Goel and his team of researchers at Carnegie Mellon University. The original intent of this study was to collect data about individual behaviors by leveraging the information provided through sensors located in the participants’ smartphones and fitness trackers. Researchers describe this process of quantifying observable traits, such as behavior and sleep patterns, through moment-to-moment analysis using digital devices as creating a “digital phenotype” (Waring 2020). The data are compiled with the goal of developing a machine learning algorithm to detect the presence and severity of various MS comorbidities. As a result of the COVID-19 pandemic, the study was able to collect data from a subset of patients who were affected by stay-at-home orders (Chikersal 2022). The unique opportunity presented by the pandemic resulted in a first-of-its-kind data set to account for an emergent global health crisis.
Dr. Goel created a data set of mental health measures using the aforementioned passive sensor data collected during the COVID-19 pandemic. During the stay-at-home mandate period, Dr. Goel’s research team also succeeded in the development of a machine learning model to predict mental health outcomes as a result of behavior change. More specifically, they collected three months’ worth of data on mental health and MS comorbidities using passive sensor data in 106 patients with MS, 56 of whom had an overlap during the stay-at-home mandate. Equipped with the algorithm they developed, the researchers were able to detect depression with an accuracy of 82.5%, high global MS symptoms burden with an accuracy of 90%, severe fatigue with an accuracy of 75.5%, and, lastly, poor sleep quality with an accuracy of 84%. For each outcome, they averaged and then dichotomized the measurements using threshold scores for the presence or absence of a particular behavior. For instance, a participant was categorized as having either a higher or lower global MS burden of symptoms based on how they scored on the Multiple Sclerosis Rating Scale-Revised (MSRS-R), which examines a patient’s perceived level of disability across eight domains (walking, upper limb function, vision, speech, swallowing, cognition, sensory function, and bladder and bowel function).
Using the funding provided by the FY19 MSRP EHDA, Dr. Goel and his team developed an approach that is expected to help clinicians identify and triage patients with MS and other chronic neurological disorders. This promises to greatly impact the ability of clinicians to deliver precise medicine and aid to the appropriate people, particularly during stressful times, such as a pandemic or a natural disaster. From a societal perspective, this type of digital technology is an invaluable tool for increasing access to health care through monitoring individuals in their own environment.
Waring O and Majumder M. 2020. Introduction to Digital Phenotyping for Global Health. In: Leveraging Data Science for Global Health (Celi LA, Majumder MS, Ordonez P, Osorio JS, Paik KE, Somai M, Eds.). Chapter 15. Pp 252-253. https://doi.org/10.1007/978-3-030-47994-7_15
Chikersal P, Venkatesh S, Masown K, et al. 2022. Predicting multiple sclerosis outcomes during the COVID-19 stay-at-home period: Observational study using passively sensed behaviors and digital phenotyping. JMIR Mental Health 9(8):e38495. https://doi.org/10.2196/38495
Last updated Tuesday, December 13, 2022