Wearable device for prevention of postoperative and post‐discharge hypoxemia: A randomized pilot trial

The Oxalert Enhanced Pulse Oximeter (EPO) is a wearable device that detects and alerts patients to hypoxemia. In a preplanned pilot trial, we estimated the effect of continuous saturation monitoring with patient alerts on in‐hospital and post‐discharge saturation; we further assessed the feasibility of the intervention.

devices assessing vital signs, including oxygenation, are a practical requirement. In this pilot study, a wearable pulse oximeter, with and without hypoxemia alarms, was tested for feasibility and acceptability for signal collection in postoperative cases, including at home. Results indicate that a full-scale trial is warranted to test for possible clinical benefit with this type of "wearable" where late postoperative hypoxia could be a concern. The trial was registered at ClincialTrial.gov (NCT04453722).

| INTRODUCTION
Most surgical patients experience moderate to severe pain, especially during the initial few postoperative days. 1,2 Opioids are effective analgesics but cause dose-dependent respiratory depression. 3 For example, about half of patients given self-controlled intravenous opioid analgesia experience nocturnal hypoxemia. 4 Furthermore, a third of postoperative patients have continuous SpO 2 <90% lasting at least an hour, and a fifth have ≥10 min/h with SpO 2 <90%. Disturbingly, nurses monitoring at 4-h intervals missed 90% of serious desaturation episodes. 5 Since postoperative patients are being discharged earlier than previously, 6 many presumably continue to experience hypoxemic episodes at home, especially during sleep.
Food and Drug Administration (FDA)-cleared continuous saturation monitoring systems for in-hospital use are commonly available, and unsurprisingly increase detection of respiratory depression. 7,8 Many are battery powered, untethered, and can directly be connected to electronic record systems. However, monitoring alone may not prevent ventilation complications if clinicians do not recognize desaturation quickly and intervene appropriately. Home saturation monitors are also readily available. 8 But again, monitoring alone will not prevent respiratory disasters unless caregivers intervene. Given that respiratory depression is frequently missed in hospitals, it seems apparent that it will usually be missed at home as well.
An alternative to caregiver intervention is a device that both detects hypoxemia, and alerts patients to the problem. 9,10 The Oxalert Enhanced Pulse Oximeter (EPO) from Med-botics, LLC (Colorado Springs, CO, USA) is a battery-powered wrist-worn system that continuously monitors for desaturation and, when necessary, generates spoken alerts followed by tactile stimulation. It is designed to rapidly detect desaturation and arouse patients sufficiently that they again breathe adequately.
We conducted a preplanned pilot trial of the Oxalert device with the overall expectation that the results would guide design of a future robust larger trial. Specifically, we tested the primary feasibility hypothesis that Oxalert EPO is well tolerated by patients, defined as the fraction of time during the initial six postoperative in-hospital days that patient wore the device. We also evaluated feasibility by asking both patients and nurses about their experiences with the system. There were four efficacy outcomes: (1) time-weighted average (TWA) SpO 2 below a threshold of 90% during hospitalization up to 6 days; (2) the number and duration of in-hospital postoperative desaturation events (saturation <90%) lasting at least 2 min; (3) the number and duration of hypoxemic episodes (saturation <90%) during the initial 24 postdischarge hours lasting at least 2 min; and (4) TWA SpO 2 <90% during the preoperative baseline assessment, TWA SpO 2 <90% during the post-discharge assessment, and the difference between them.

| MATERIALS AND METHODS
The Cleveland Clinic IRB approved our pilot trial (identifier 2020-055).
The trial was registered prior to patient enrollment at ClincialTrial.gov (NCT04453722, principal investigator Daniel I Sessler, date of registration June 10, 2020). Minor changes were made on July 10, 2020 before trial data were accessed. Written informed consent was obtained from all participants at least a day before surgery. All patients were enrolled at the Cleveland Clinic Main Campus. This manuscript adheres to the applicable CONSORT guidelines.
We enrolled patients at risk for respiratory insufficiency and desaturation. Specifically, we focused on patients scheduled for laparoscopic or open major abdominal or pelvic surgery who had a body mass index ≥25 kg/m 2 and anticipated primary opioid analgesia.
Participating patients were American Society of Anesthesiologists physical status 1-3, 18-85 years old, and had an expected duration of hospitalization at least 24 h after surgery. All patients were enrolled at the Cleveland Clinic Main Campus.
We excluded patients who were scheduled for epidural analgesia (although field and fascial plane blocks were permitted), had a preoperative SpO 2 <95%, severe hearing loss, peripheral neuropathy, or lack of English language fluency.

| Protocol
We tested the Oxalert EPO which is a small (9 Â 5 Â 2 cm) battery-powered apnea detection and alerting device. The device is wrist-worn and secured by a watch-like strap. It is also connected via two standard electrocardiogram pads positioned on the back of the wrist. The system continuously monitors SpO 2 and generates audible prompts ("Please take a breath and check the finger sensor") when the probe is disconnected or when saturation is <90% for 2 continuous minutes. Audible prompts continue until saturation increased to ≥90%.
If saturation remains <90% for an additional minute, very brief (pulse-width 290 μs) and mild electrical shocks (2 Hz at 35 mA) to the posterior wrist electrodes provided tactile stimulation, with voltage progressively increasing four times at 20-s intervals to a to a maximum of 35 mA which does not cause cutaneous injury. Patients could discontinue the alerts for 60 s at any point by pressing a button on the device. Although the Oxalert devices normally display saturation, those used in the trial were modified to conceal saturation.
Participating patients were given an Oxalert EPO set to Monitor-Only mode without hypoxemia alerts and asked to wear the device for at least one preoperative night at home to determine their baseline incidence and severity of hypoxemia. The device was thereafter either brought to the Clinic the morning of surgery or returned by prepaid mailer.
Enrolled patients were randomized 1 : 1 without stratification based on computer-generated codes using random-sized blocks. Allocation was concealed with a web-based system that was accessed during or after surgery. The Oxalert devices were positioned in the postanesthesia care unit, and patients were asked to wear one throughout hospitalization for up to 6 days, and for 24 h after discharge. Randomization determined whether the devices were set to Monitor-Only mode or monitoring combined with hypoxemia alerts (Monitor + Alert). Patients were paid $3 per hour of Oxalert monitoring, to a maximum $500.

| Measurements
An unblinded team of investigators (K.L. and M.B.) were responsible for providing devices set to monitoring only or monitoring with alerts, changing batteries, and downloading continuous SpO 2 data from Oxalert devices to a custom Redcap database. All other patient interactions were conducted by investigators who were blinded to device status (R.S. and R.N.), including telephone follow up after discharge and nursing queries. Patients were not told whether their devices were set to deliver alerts or not-although those who received alerts presumably recognized their treatment allocation.
Patients were asked to rate device comfort and tolerability at hospital discharge. During hospitalization, one nurse responsible for each participating patient was asked to participate in an anonymous questionnaire to evaluate the convenience of the device. Patients were called on the second post-discharge day and asked about opioid consumption during the initial 24 post-discharge hours. They were also asked to rate convenience of the Oxalert system and report adverse events.

| Statistical analysis
Randomized groups were compared on demographic and morphometric characteristics using standard descriptive statistics and assessed for balance using the absolute standardized difference. Based on the achieved sample size of N = 49, variables with ASD >0.56 would be considered imbalanced and adjusted for in all analyses. 11 Feasibility was assessed in several ways. Descriptive statistics were used to summarize enrollment rate. The extent to which the system was tolerated was estimated as the fraction of time during the initial six postoperative in-hospital days during which patients wore the device. We also summarized the fraction of time successful records were made, defined as the percent of time in each period that usable SpO 2 monitoring data were retrieved and usable, independent of whether the device was being worn or not. Measurements were made in the 24 h prehospital, the initial 6 postsurgical hospital days; and the post-discharge 24 h. Groups were compared using the Wilcoxon rank sum test.
Monitor + Alert and Monitor-Only patients were compared on 10 Likert-style questions (ranging from either 0 to 10 or 1 to 5) measuring patients' attitude toward the device, both in-hospital and 24 h after discharge using the Wilcoxon rank sum test and Hodges-Lehmann estimator of median difference, as with the primary outcomes. Groups were similarly compared on nine questions assessing nurses' attitude toward the device.
We also estimate the frequency of hypoxemic episodes detected by the Oxalert EPO that generate an alert in the hospital. Since the prototype device we used does not record whether alerts were given, we use the collected SpO 2 data to summarize for each group the frequency of desaturation events lasting 1, 2, 3, 4, …, ≥10 min numerically and using histograms. Groups were further compared on the dichotomous proportion incurring any desaturation events with a Pearson chi-square test and estimating the relative risk (95% CI). Finally, randomized groups were compared on the change from preadmission to post-discharge on the TWA SpO 2 <90% (%) and AUC SpO 2 <90% (% * min) using the Wilcoxon rank sum test; change within group was assessed using the Wilcoxon signed rank test, with Bonferroni correction within each group.
The significance level for all hypotheses was 0.05. Bonferroni correction was made when comparisons were made within-group, that is, using significance criterion of 0.05/2 = 0.025. However, in this pilot trial we focused on estimating of treatment effects instead of the need to find statistical significance. Therefore, our CIs are the key results when testing hypotheses and will be used to plan a larger trial. SAS statistical software, Carey, NC, was used for all analyses.
Sample size justification. The study was designed to be able to assess feasibility of the study intervention and to estimate the treatment effect of interest with reasonable precision. With the chosen sample size of 25 per group we would be able to estimate a 95% CI having width no larger than 0.65 SDs of the outcome variable with probability of 80%. We believe that CIs this narrow are sufficient for planning a larger trial based on the pilot results.    (Table 1).  Figure S2). F I G U R E 1 Primary results: timeweighted average SpO 2 <90% by treatment group, in-hospital (left) and during the first 24-h post-discharge (right). As expected in a small pilot trial, differences were not statistically significant in-hospital (p = .12) or postdischarge (p = .31), although the inhospital treatment effect was substantial.

| Changes from preadmission to post-discharge
Nineteen patients in each group had both preadmission and postdischarge SpO 2 data available. As reported in groups on patients' attitudes toward the device (Table S1). After discharge, satisfaction with the electrode patch (p = .023) and willingness to use the device after the study (p = .023) were a bit lower in the Monitor + Alert group than in patients assigned to Monitoring Only. There were several statistically significant differences between groups on nurses' attitudes toward the device, but none was clinically meaningful (Table S2).

| Feasibility assessments
We achieved the goal of a single-center enrollment rate of at least two patients per week at the Cleveland Clinic Main Campus. We randomized 49 patients over 4.2 months, or an average of 2.7 patients/week ( Figure S3).  Table S4 and Figure S4. Randomized groups did not differ on opioid consumption (Table S5).

| DISCUSSION
In this pilot trial of 49 patients who had major abdominal surgery, we tested the hypotheses that the use of a battery-powered, wrist-worn apnea prevention device is well tolerated by patients and reduces hypoxemia. Our feasibility goals were met in that the Oxalert was well tolerated by postoperative patients who used the system during 90% Why the periods should differ remains unclear, but is possibly consequent to the small number of patients we enrolled. There appears to be little previous quantification of post-discharge hypoxemia in surgical patients, and more is clearly indicated.
Our results are specific to the device we used which was a prototype and is not FDA cleared. Results would surely differ had the alerts started earlier and been more intrusive. But more intrusive alerts would presumably also reduce tolerability of the system which is designed to be worn continuously. At least in hospital, it would be theoretically possible for the system to alert nurses if patient alerts were insufficient to reverse hypoxemia. 8,10 The trial was not fully blinded because patients who received alerts would correctly conclude that they were in the alert group, where as those who did not would not know if they were in the Monitor-Only group or simply never became hypoxemic. Given that saturation is an objective response, it seems unlikely that lack of full blinding much influenced the observed amount of hypoxemia.
Patients were paid $3/h to wear the Oxalert because only half would benefit from the monitor. To the extent that payment encouraged system use, patient tolerance may have been overestimated. The major limitation of our trial is that we restricted enrollment in our prespecified pilot trial to about 49 patients. The results, while encouraging, are by no means definitive-and mostly not even statistically significant. They are, however, a strong basis for a full trial.
In summary, the Oxalert system was well tolerated in both groups and enrollment was strong. Patients randomized to active Oxalert systems experienced half as many postoperative desaturation events while hospitalized, although the difference was not statistically significant in this small pilot trial. In contrast, the Oxalert system did not reduce post-discharge desaturation. A full trial of in-hospital desaturation alerts thus seems warranted.

CONFLICT OF INTEREST
Lloyd Olson is the founder and CEO of Med-botics, LLC (Colorado Springs, CO, USA), who developed the Oxalert EPO device we tested.
None of the other authors has a personal financial interest in this research.