Assessment of physiological signs associated with COVID-19 measured using wearable devices

Abstract

Respiration rate, heart rate, and heart rate variability are some health metrics that are easily measured by consumer devices and which can potentially provide early signs of illness. Furthermore, mobile applications which accompany wearable devices can be used to collect relevant self-reported symptoms and demographic data. This makes consumer devices a valuable tool in the fight against the COVID-19 pandemic. We considered two approaches to assessing COVID-19 - a symptom-based approach, and a physiological signs based technique. Firstly, we trained a Logistic Regression classifier to predict the need for hospitalization of COVID-19 patients given the symptoms experienced, age, sex, and BMI. Secondly, we trained a neural network classifier to predict whether a person is sick on any specific day given respiration rate, heart rate, and heart rate variability data for that day and and for the four preceding days. Data on 1,181 subjects diagnosed with COVID-19 (active infection, PCR test) were collected from May 21 - July 14, 2020. 11.0% of COVID-19 subjects were asymptomatic, 47.2% of subjects recovered at home by themselves, 33.2% recovered at home with the help of someone else, 8.16% of subjects required hospitalization without ventilation support, and 0.448% required ventilation. Fever was present in 54.8% of subjects. Based on self-reported symptoms alone, we obtained an AUC of 0.77 +/- 0.05 for the prediction of the need for hospitalization. Based on physiological signs, we obtained an AUC of 0.77 +/- 0.03 for the prediction of illness on a specific day with 4 previous days of history. Respiration rate and heart rate are typically elevated by illness, while heart rate variability is decreased. Measuring these metrics can help in early diagnosis, and in monitoring the progress of the disease.

Wearable sensor data and self-reported symptoms for COVID-19 detection

“…This work builds on our earlier retrospective analysis demonstrating the potential for consumer sensors to identify individuals with influenza-like illness, which has subsequently been replicated in a similar analysis of over 1.3 million wearable users in China for predicting COVID-19 8,9 . In response to the COVID-19 pandemic, a number of prospective studies, led by device manufacturers and/or academic institutions, including DETECT, have accelerated deployment to allow interested individuals to voluntarily share their sensor and clinical data to help address the global crisis [10][11][12][13][14] . The largest of these efforts, Corona-Datenspende, was developed by the Robert Koch Institut in Germany and has enrolled over 500,000 volunteers 15 .…”

Deep longitudinal phenotyping of wearable sensor data reveals independent markers of longevity, stress, and resilience

“…On the contrary, the number of the relevant variables is expected to increase if we turn to characterization of the organism state variation at a higher sampling rate. This might be the case for situation involving response to an acute illness on time scales of days or a few weeks [30], such as increased RHR during fever [31, 32] or change in sleep patterns as potentially a COVID-19 specific signal [33, 34].…”

Wearables in the SARS-CoV-2 Pandemic: What Are They Good for?

“…With the SARS-CoV-2 pandemic, the desire to use consumer-targeted BioMeTs to monitor potential symptoms of infection is understandable, because monitoring for signs of infection at home may reduce anxiety and increase one’s confidence in their health status (ie, healthy or sick). The danger however is that many people are improperly using consumer-targeted BioMeTs to monitor for signs of illness and relying too heavily on data that does not have a sufficient evidence base, as there is no oversight by medical professionals [ 10 , 11 ]. Furthermore, it has become increasingly common for a single product to have differentially regulated features, which only adds to the confusion.…”