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

Abstract

Respiration rate, heart rate, and heart rate variability (HRV) are some health metrics that are easily measured by consumer devices, which can potentially provide early signs of illness. Furthermore, mobile applications that 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. Data on 2745 subjects diagnosed with COVID-19 (active infection, PCR test) were collected from May 21 to September 11, 2020, consisting of PCR positive tests conducted between February 16 and September 9. Considering male (female) participants, 11.9% (11.2%) of the participants were asymptomatic, 48.3% (47.8%) recovered at home by themselves, 29.7% (33.7%) recovered at home with the help of someone else, 9.3% (6.6%) required hospitalization without ventilation, and 0.5% (0.4%) required ventilation. There were a total of 21 symptoms reported, and the prevalence of symptoms varies by sex. Fever was present in 59.4% of male subjects and in 52% of female subjects. Based on self-reported symptoms alone, we obtained an AUC of 0.82 ± 0.017 for the prediction of the need for hospitalization. Based on physiological signs, we obtained an AUC of 0.77 ± 0.018 for the prediction of illness on a specific day. Respiration rate and heart rate are typically elevated by illness, while HRV is decreased. Measuring these metrics, taken in conjunction with molecular-based diagnostics, may lead to better early detection and monitoring of COVID-19.

Characterizing COVID-19 and Influenza Illnesses in the Real World via Person-Generated Health Data

“…Several recent works have explored use of wearable data, including RHR, to detect symptoms of COVID before they appear toward applications that can prompt users to intervene in pre-symptomatic disease phases and curb the spread of infection (e.g., self-quarantine while waiting for a confirmatory test). 22 , 23 , 24 , 35 While these systems have shown moderate discriminative ability between COVID-19 patients versus healthy persons in retrospective cohorts, our findings suggest that the specificity of those systems should also be measured as compared with flu patients, as they will be the overwhelming majority as flu season starts. If specificity versus flu and other respiratory viruses cannot be demonstrated, early-warning systems triggered on wearable data should be considered as more non-specific “infection screening,” and therefore be coupled with appropriate confirmatory testing mechanisms that can help to quickly relieve self-imposed quarantine of non-COVID-19 infections.…”

“…In the specific context of COVID-19, our findings support the case made by recent work that data from wearable sensors may provide low-sensitivity testing capability with daily frequency. 22 , 23 , 24 , 35 Low-sensitivity/high-frequency testing when combined with a low-delay confirmatory testing strategy has been shown by computation models to significantly reduce prevalence of spreading with minimal burden on pre-emptive quarantine for false positives. 14 , 16 Therefore, wearables could potentially support use cases, such as return to work and college reopening, 18 where most of the cohort can be asked to wear the sensors frequently.…”

“…Along these lines, several efforts are currently underway to explore the potential of using wearable technology to detect COVID-19 onset, 19 , 20 , 21 and preliminary results have shown that wearables may be able to predict COVID-19 symptoms before onset, 22 , 23 , 24 with potential application to large-scale, low-sensitivity/high-frequency testing to enable reopening in the wait for a vaccine. 14 , 15 , 16 …”

“…However, the lack of a canonical COVID-19 symptom presentation, including how symptoms progress over time, 24 , 25 undermines our ability to track, predict, and control disease progression and manage critical care. In addition, to evaluate performances of any detection system, being that for individual-level early warnings or population-level hotspot detection, it is crucial to compare and contrast symptoms, behavioral, and physiological manifestation with other ILIs, especially flu.…”

Wearable devices as a valid support for diagnostic excellence: lessons from a pandemic going forward

“…It shows that a combination of increased resting heart rate (HR), decreased physical activity, and increased amount of sleep hours (so, both physiological and behavioral data) may be a crisp sign of the onset of a viral illness, like (in particular) COVID-19 but also other infections. A project like this (which may be considered a sort of digital clinical trial [ 7 ]) – also together with other studies collecting physiological data from thousands of citizens [ 8 , 9 ] – allows scientists to merge data from worldwide populations, thus properly feeding artificial intelligence (AI) algorithms [ 10 ] to identify possible patterns able to discriminate between healthy and sick people. The project potential impact could be significant in terms of the early detection of possible risk, hence limiting the number of contacts and hindering the virus diffusion.…”