Review 2: "Prediction of deterioration from COVID-19 in patients in skilled nursing facilities using wearable and contact-free devices: a feasibility study"

Panayiotis Kouis

doi:doi:10.1162/2e3983f5.80a20b5d

This preprint aims to use wearable devices to predict deterioration from COVID-19 in skilled nursing facility patients. Reviewers suggested improvements in determining statistical significance, clarifying the data collection period, and discussing viral transmission implications.

by Panayiotis Kouis

Published onMay 24, 2022

Prediction of deterioration from COVID-19 in patients in skilled nursing facilities using wearable and contact-free devices: a feasibility study

by Sabine von Preyss-Friedman, Linda Emmet, Dree Deckert, Dennis A.B. David, Heikki Raisanen, Kevin Longoria, Willem Gielen, and Martin G. Frasch

Show Description
dx.doi.org

Description

AbstractBackground and RationaleApproximately 35% of all COVID-19 deaths occurred in Skilled Nursing Facilities (SNFs). In a healthy general population, wearables have shown promise in providing early alerts for actionable interventions during the pandemic. We tested this promise in a cohort of SNFs patients diagnosed with COVID-19 and admitted for post-acute care under quarantine. We tested if 1) deployment of wearables and contact-free biosensors is feasible in the setting of SNFs and 2) they can provide early and actionable insights into deterioration.MethodsThis prospective clinical trial has been IRB-approved (NCT04548895). We deployed two commercially available devices detecting continuously every 2-3 minutes heart rate (HR), respiratory rate (RR) and uniquely providing the following biometrics: 1) the wrist-worn bracelet by Biostrap yielded continuous oxygen saturation (O2Sat), 2) the under-mattress ballistocardiography sensor by Emfit tracked in-bed activity, tossing, and sleep disturbances. Patients also underwent routine monitoring by staff every 2-4 h. For death outcomes, cases are reported due to the small sample size. For palliative care versus at-home discharges, we report mean±SD at p<0.05.ResultsFrom 12/2020 - 03/2021, we approached 26 PCR-confirmed SarsCoV2-positive patients at two SNFs: 5 declined, 21 were enrolled into monitoring by both sensors (female=13, male=8; age 77.2±9.1). We recorded outcomes as discharged to home (8, 38%), palliative care (9, 43%) or death (4, 19%). The O2Sat threshold of 91% alerted for intervention. Biostrap captured hypoxic events below 91% nine times as often as the routine intermittent pulse oximetry. In the patient deceased, two weeks prior we observed a wide range of O2Sat values (65-95%) captured by the Biostrap device and not noticeable with the routine vital sign spot checks. In this patient, the Emfit sensor yielded a markedly reduced RR (7/min) in contrast to 18/min from two routine spot checks performed in the same period of observation as well as compared to the seven patients discharged home over a total of 86 days of monitoring (RR 19± 4). Among the patients discharged to palliative care, a total of 76 days were monitored, HR did not differ compared to the patients discharged home (68±8 vs 70±7 bpm). However, we observed a statistically significant reduction of RR at 16±4/min as well as the variances in RR (10±6 vs 19±4/min vs16±13) and activity of palliative care patients vs. patients discharged home.Conclusion/DiscussionWe demonstrate that wearables and under-mattress sensors can be integrated successfully into the SNF workflows and are well tolerated by the patients. Moreover, specific early changes of oxygen saturation fluctuations and other biometrics herald deterioration from COVID-19 two weeks in advance and evaded detection without the devices. Wearable devices and under-mattress sensors in SNFs hold significant potential for early disease detection.

RR:C19 Evidence Scale rating by reviewer:

Potentially informative. The main claims made are not strongly justified by the methods and data, but may yield some insight. The results and conclusions of the study may resemble those from the hypothetical ideal study, but there is substantial room for doubt. Decision-makers should consider this evidence only with a thorough understanding of its weaknesses, alongside other evidence and theory. Decision-makers should not consider this actionable, unless the weaknesses are clearly understood and there is other theory and evidence to further support it.

***************************************

Review:

The study by Sabine von Preyss-Friedman and co-authors focuses on the feasibility of deploying wearables and other biosensors in nursing facilities and their impact on predicting COVID-19 deterioration. This topic has received considerable attention and the use of wearables sensors in monitoring COVID-19 patients has been suggested by many experts. Several studies are now emerging with the results of such applications and this manuscript also adds to this literature. More particularly, the manuscript focuses on a population that is most vulnerable (residents of nursing homes) and, usually, least familiar with wearable devices. Overall, the study is well conducted, and the manuscript is well-written. However, some considerations need to be taken into account:

Major comments:

1. In the study design section, the expected data collection period for each participant is reported to be 60 days from the time of the first biometric measurement. In the result, the authors do not report whether this data collection period was adhered to and to what extent. The days of monitoring are only reported for patients’ discharges to long-term care (76 days) and not for the other outcome categories. This parameter is important to demonstrate the feasibility (and acceptability by the patients) of the methods and was important to be more clearly reported.

2. The authors mention that statistical differences between movements are reported between two patient groups but these are not visible in the manuscript.

3. Results of the symptom questionnaires are also not presented in the results section.

4. In the discussion section, the authors talk about the additional predictive insights of monitoring sleep quality. However, based on the results, this was not measured, or at least not reported. The only activity was measured, and this does not provide any indication of whether monitoring sleep quality would have additional predictive insights.

5. The discussion of reducing viral transmission and especially the statement on validating the author’s own algorithms in the pre-COVID stage by future studies is problematic. Firstly the authors do not really provide any such algorithms per se and even the approaches they suggest are not applicable in the pre-symptomatic or pre-COVID stage. I can only assume that by the time oxygen saturation falls below a certain level of activity falls below a certain level to be observed by wearable sensors, other symptoms would have already alerted the caregivers to proceed with COVID testing and subsequent isolation of the patient to limit virus transmission. Overall, this part of the discussion is not supported by the results outlined.

Minor comments:

1. Figure 2 could be combined as one multifigure with bars from manual and wristband measurements appearing side by side for easier comparison.

2. In Figure 4, there is an afternoon spot check that could also be circled to be more visible

Review 2: "Prediction of deterioration from COVID-19 in patients in skilled nursing facilities using wearable and contact-free devices: a feasibility study"

Connections