A Novel Shoulder-Mounted Pulse Oximeter in Patients with Suspected Sleep Apnea: Design and Patient Perceptions

1. Introduction

Introduction

Wearable continuous monitors technologies are increasingly common in healthcare [1,2]. Continuous pulse oximetry would benefit several high risk populations including patients with chronic lung disease, sleep-disordered breathing, and patients with active opioid use [3,4,5]. Conventional pulse oximeters limit use of the hands, but alternate pulse oximeter probe sites range from the forehead to the toes sometimes with smartphone interface [6,7,8,9,10]. Motion artifacts in ambulatory patients limit accuracy, potentially resulting in delayed detection of hypoxemia [1,11]. Wearable biosensor technologies in the outpatient setting must be designed with user acceptability, device effectiveness, and hold the potential for additional features such as integrated respiratory monitoring [4,12,13].

Studies of shoulder-mounted pulse oximetry technology are limited [14,15]. Our group developed a shoulder-mounted device with a high level of acceptability in certain populations [16]. We sought to investigate the feasibility and acceptability of a shoulder-based pulse oximeter among a convenience cohort undergoing diagnostic polysomnography.

2. Materials and Methods

We conducted a quantitative and descriptive pilot of two prototype biosensor designs at the Penn Sleep Medicine Diagnostic Program, an outpatient sleep center in Philadelphia, Pennsylvania, USA. The study protocol was approved by the University of Pennsylvania IRB and was performed in compliance with relevant laws and institutional guidelines. Inclusion criteria included clinical suspicion for sleep-disordered breathing and age >21. Exclusion criteria included pregnancy.

Patients wore a standard pulse oximeter (EMAY EMO-80 Sleep Oxygen Monitor, Figure 1a) and two prototype biosensors: Prototype-ring (Figure 1b) and Prototype-shoulder (Figure 1c). Prototype-ring (Figure 1b) consisted of a ring worn on the index finger to assess oxygen satu-ration as well as a wristband containing a battery and data storage hardware. Proto-type-shoulder (Figure 1c) was an adhesive armband worn over the deltoid.

We assessed the prototypes based on FDA guidelines for pulse oximeters [18]. We trained the model using an 80-20 train-test split on a 5-fold cross validation. We then tested the R2 score, mean absolute error, and standard deviation of the mean absolute error to ensure the model was not overfit (Supplemental Methods). No adjustments were made for known confounders such as skin tone, positioning, or sleep stage. We also evaluated the comfort via a 5-point Likert scale and an open-ended qualitative questionnaire. Independent patient-level variables included demographics and participant BMI. Differences in mean results were assessed with a two-sided t-test with alpha 0.05.

3. Results

3.1. Study Participants

Participant demographic information is described in Table 1. The patients sampled were majority white (57.1%) and male (58.3%). Participants had a median BMI of 28.5 (IQR 26.5 - 35.1). Participant demographics are described in Table 1. The patients sampled were majority white (57.1%) and male (58.3%). Participants had a median BMI of 28.5 (IQR 26.5 - 35.1). A consort diagram is shown in Figure 2. Data from the shoulder was unusable for 3 patients due to corrupted data files (2) and SD card unavailability (1). The FDA cleared control device had 8 patients with unusable data. All patients with interpretable finger pulse oximetry data had interpretable shoulder-mounted data.

Agreement among the 19 patients with data from both devices is shown in Figure 3. The Prototype-shoulder pulse oximeter had a 0.72% mean absolute error from the control values of the commercial finger-based pulse oximeter (Figure 3). Participants removed the prototype 0 times, and the control device a mean 4.6 (sd 3.7) times (Figure 4a), and the time removed is shown in Figure 4b. 5 point Likert Scale feedback on comfort showed a mean rating of 4.6 for the shoulder, 4.5 for the ring and 3.1 for the finger mounted control (Figure 5). Both the shoulder and ring were rated higher than the control de-vice (p<0.01).

Themes were identified in open-ended questions regarding the comfort or fit of the study-provided devices: Many participants reported that their control devices fell off either before falling asleep or during the night. One participant noted that the finger-based “commercial device came off a few times before sleeping” which caused “restless sleep”. Several participants also noted that the control devices either fell off or were purposefully re-moved throughout the night. Participants cited irritation, fit, and using the restroom as reasons for control device removal. One person noted that the wristband portion of the Prototype-ring fell off in the middle of the night. None complained of comfort worsened by the shoulder mounted device. Several participants complained that one or multiple devices disrupted their sleep. These devices included the Prototype-ring and commercial pulse oximeter. Tightness of fit was the most commonly problematic factor among these devices. One participant specifically cited the plastic shell that encased the data hardware in Prototype-ring as causing pain. Other participants removed the control device due to discomfort of a “burning sensation” or complaints that the device “was getting hot.” There was only one participant who specifically commented on Prototype-shoulder. This person stated that they slept on their side but there was “not much movement [to the device] while sleeping.”