Augmenting management of obstructive sleep apnea with consumer wearable devices to increase use of positive airway pressure therapy: a pilot randomized trial

Abstract

Study Objectives:

This pilot randomized trial measured the effects of a consumer grade wearable-augmented program that provides nightly pulse oximetry data to patients in a weekly report, on positive airway pressure (PAP) use in adults with obstructive sleep apnea (OSA) recruited from an outpatient sleep center.

Methods:

This single-blind, parallel, 2-group waitlist-controlled, single site randomized trial recruited individuals with OSA with insufficient PAP use from a Department of Veterans Affairs medical center and randomized them to immediate participation in the consumer wearable-augmented self-monitoring program versus waitlist. The primary outcome was the change in modem-measured PAP use (mean hours or minutes per night over a 7-day period) from baseline (T1) to T3 assessment, which was 35 days following randomization (7 days post-intervention for the immediate group and prior to intervention for the waitlist group).

Findings:

We randomized 50 participants (age_M 47.9 years [SD 11.4]; 88% male). The change in PAP use was +1.33 hours (+79.8 minutes) for the immediate group and -0.17 hours (-10.2 minutes) for the waitlist control group, with a between-group difference of 1.50 hours (90 minutes) (p=.0097). The change in number of days PAP was used for ≥ 4 hours over a 7-day period was +1.32 days for the immediate and 0.12 days for the waitlist control group, with a between-group difference of 1.20 (p=.0401).

Conclusions:

This pilot trial demonstrated that the consumer grade wearable-augmented program is a promising intervention for increasing PAP use. A future full-scale trial that follows participants for a longer period is needed.

INTRODUCTION

Obstructive sleep apnea (OSA) is one of the most common sleep disorders in the United States, occurring in 21% of veterans and up to 9% of nonveterans.¹ Although positive airway pressure (PAP) therapy is typically prescribed for OSA, most patients struggle to use their PAP device regularly.² Supportive interventions (e.g., telemonitoring and communication of usage) increase PAP device use by approximately 42 minutes per night.³ Behavioral interventions (e.g., cognitive behavioral therapy, motivational enhancement) are typically delivered by a healthcare provider, and studies show that they increase PAP device use by approximately 79 minutes.³ With the high prevalence of OSA and insufficient PAP use, novel programs that are highly scalable are needed.

Programs to address insufficient PAP use that leverage technological innovations are a promising approach to addressing gaps in access to programs for promoting more regular use of PAP. For example, some mobile app-based interventions aim to enhance knowledge about OSA, to provide feedback about PAP use, and to remind patients to use their PAP.^4–7 Consumer-grade wearable devices (e.g., Garmin, Fitbit) that generate patient-generated health data (PGHD) offer a new avenue for using technology to improve PAP adherence. These devices are increasingly used by individuals to gain insight into their health and lifestyle. One recently added feature, minute-to-minute measurements of oxygen saturation (SpO2) during sleep periods, has the potential to provide patients with OSA insight into the health implications of not using PAP therapy regularly.⁸

The objective of this pilot study was to measure the effect of the consumer grade wearable-augmented program on PAP use. Our primary hypothesis was that the participants in the immediate intervention group would have greater improvement in PAP use (primary outcome: change in mean hours or minutes per night; secondary outcome change in number of days [percentage] PAP was used ≥ 4 hours) from baseline to posttreatment assessment compared to a waitlist control group. We also explored changes in patient-reported sleep and daytime symptoms from baseline to posttreatment.

METHODS

Trial design

This was a single-blind, parallel, 2-group waitlist-controlled, single-site randomized trial (1:1 allocation) of a 28-day consumer wearable-augmented self-monitoring program. The intervention was designed to be 28 days so that participants would be less likely to discern whether they were in the immediate versus waitlist group. Study activities were performed at VA Greater Los Angeles Healthcare System in Los Angeles, California. The local institutional review board approved all study activities (1754210). The trial was registered (clinicaltrials.gov NCT06039865), and the protocol has been published.⁹

Participants

Participants were recruited between October 2023 and January 2025 from the VA Greater Los Angeles Healthcare System’s outpatient sleep center and other outpatient clinics by posting recruitment flyers, through clinician referrals, and by sending opt-in/out letters to sleep center patients with insufficient PAP use. Inclusion criteria were: 1) prescribed PAP therapy for 7 or more days, 2) active profile in our Center’s ResMed AirView or Respironics Care Orchestrator accounts for 7 or more days, 3) willing to continue using current PAP device for 28 days continuously (i.e., not planning to use a different type of treatment for OSA), 4) not using PAP for 4 or more hours per night on 70% of nights during past 7 days (this was a change from the protocol which previously stated 30 days—the change was made to align the timeframe with inclusion criterion #1), and 5) has a smartphone. Exclusion criteria were: nocturnal hypoxemia unrelated to OSA, home oxygen use, chronic lung disease, heart failure, conditions or activities that would prohibit or interfere with study activities (e.g., active illicit substance use disorder, dementia, planned hospitalization that would limit use of personal PAP device). Recruitment ended when the prespecified enrollment number (50 participants) was reached.

Procedures

After written informed consent was obtained, participants underwent a baseline assessment (T1), which included standard questionnaires, and chart review including abstraction of PAP use data (modem data) during past 7 and 30 days. The assessment (PAP modem data, a subset of the baseline questionnaires) was repeated 35 days after randomization (T3), which was 1 week after the immediate group completed the intervention program and the week prior to the waitlist group’s participation in the intervention. T4 marked the start of intervention for the waitlist group, which was followed by T5 when additional assessments, including inquiries about adverse events, occurred. Figure 1 summarizes the study time points. The PAP modem data were abstracted by team members blinded to study arm assignment.

Figure 1. Study Timepoints.

This figure summarizes the study activities at each timepoint. PROMIS=Patient-Reported Outcomes Measurement Information System; ESS=Epworth Sleepiness Scale

During the trial, all participants continued to receive standard care from the sleep center, such as clinic follow-up visits with sleep medicine providers, access to replacement PAP equipment such as masks every 3 months, and walk-in mask fitting with respiratory therapists.

Intervention

Immediate group

As previously described in our published protocol paper, the “mPAP” intervention consisted of providing four weekly reports to participants that displayed summaries of their nightly oxygen saturation distribution (e.g., % time spent <88%), objective PAP use (abstracted by research team from AirView [ResMed, San Diego, CA] or Care Orchestrator [Philips Respironics, Murrysville, PA]), and a summary of each participant’s own responses to weekly questions about OSA symptoms that were collected through Qualtrics survey software (Qualtrics, Provo, UT). The oxygen saturation distribution data were collected from Fitbit Charge 5 (Google LLC, Mountain View, CA) devices that were provided by the research team to the participants at the beginning of the intervention program. SpO2 data were transmitted (with the participant’s consent) to the research team’s cloud-based Fitabase account (Fitabase, San Diego, CA). The research team then used the SpO2 data to prepare the customized report provided to the participants weekly. The report visualized daily SpO2 levels while sleeping (e.g., percentage of time SpO2 >92%, 88%-92%, etc.) to illustrate the proportion of minutes in normal and abnormal SpO2 levels, juxtaposed with daily PAP use (yes/no based on modem data) and sleep-related symptoms (e.g., daytime sleepiness). See Figure 2.

Figure 2. Example of mPAP Report.

The research team produced a weekly report during the intervention, compiling oxygen saturation data from the consumer wearable, positive airway pressure modem data, and self-reported questionnaire data about obstructive sleep apnea-related symptoms.

A one-time “check-in” meeting with research staff (non-clinical) was conducted with participants during the first week of the program. Using a standard procedure, staff identified any issues setting up the device, reviewed the process of accessing the report, assessed current understanding of the content of the report (including a statement encouraging participants to note any patterns between use of their PAP therapy and any low oxygen levels since lack of PAP use can directly impact oxygen saturation levels noted in the report’s chart), reiterated the goal of the weekly reports, and answered questions about the program.

Automated reminder emails were also sent when participants did not respond to the weekly sleep symptoms survey.

Waitlist group

Participants randomized to the waitlist group were offered to begin the mPAP intervention after their T3 timepoint.

Outcomes

The primary outcome measure was the change in modem-measured PAP use (mean hours or minutes per night over a 7-day period) from baseline (T1) to T3, which was 35 days following randomization (7 days post intervention for the immediate group and prior to intervention for the waitlist group).¹⁰

Secondary outcome measures included the number of days PAP was used for ≥ 4 hours, number of days of PAP use out of 7 days, and mean apnea-hypopnea index on therapy (based on PAP modem data).¹⁰

Exploratory measures include the 8-item Patient-Reported Outcomes Measurement Information System (PROMIS) Sleep Disturbance and 8-item Sleep-Related Impairment ¹¹, and 8-item Epworth Sleepiness Scale.¹²

Sample size

As a pilot trial, our sample size calculation was based on a balance between an anticipated large effect size related to the intervention, a sample large enough to test the protocol and uncover feasibility issues, and statistical power. We estimated that a sample size of 25 participants per study group would provide at least 80% power to detect a standardized effect size of 0.72, assuming .05 two-sided significance.

Randomization and masking

Participants were randomized to immediate versus waitlist groups (1:1 ratio), using a random permuted block design with varying block size generated by the study statistician, who was not involved in day-to-day study activities. Randomization was implemented using opaque sequentially numbered envelopes that were prepared by a researcher who was not part of the study.

Research staff who were blinded to study arm assignment abstracted the modem data (primary and secondary outcomes). The participants were not provided with details about the study timeline activities or their group assignment, to reduce their ability to determine whether they were in the immediate or waitlist control groups.

Data Analysis

The primary and secondary outcomes were analyzed using an intention-to-treat approach. Descriptive statistics and frequencies were used to summarize the continuous and categorical demographic and clinical characteristics at baseline (T1), respectively. A linear mixed-effects model with a patient-level random effect, accounting for repeated measurements within individual patients, was used to estimate changes in outcome from T1 to T3 within and between groups. Each model included group (immediate vs. waitlist control), visit (T1 and T3), and a two-way group-by-visit interaction term. The difference-in-difference (DID) approach was used to compare the changes in outcome from T1 to T3 through model contrasts. The estimated changes from T1 to T3 and their corresponding 95% confidence intervals (CIs) were graphically displayed. Statistical significance was assessed using a two-sided p-value < 0.05. All analyses were conducted using SAS 9.4 software (Cary, NC, SAS Institute), and all graphs were generated using the publicly available statistical software R (R Core Team, 2024).

Role of the funding source

The funders did not have any role in the study design, data collection, analysis, interpretation, drafting of the manuscript, or decision-making regarding manuscript submission.

RESULTS

Participants

A total of fifty participants were randomized to the immediate and waitlist control groups (Figure 3 CONSORT). Baseline characteristics are shown in Table 1. The mean age was 47.9 years (SD 11.4), with an average age that was slightly older (50.4 years) in the immediate versus waitlist control group (45.4 years). Most participants in both groups were male (44 [88%]), white (21 [42%]), and non-Hispanic (32 [64%]).

Figure 3. CONSORT Flow Chart.

This figure provides a summary of the participants at each stage of the study.

Table 1.

Sample Characteristics

Variable	Total	Group Assignment
	(N=50)	Immediate (N=25)	Waitlist (N=25)
Age, Mean (SD)	47.9 (11.4)	50.4 (12.4)	45.4 (9.9)
Age, Min - Max	24 – 73	25 – 73	24 – 68
Female	6 (12.0)	4 (16.0)	2 (8.0)
Male	44 (88.0)	21 (84.0)	23 (92.0)
Race
White	21 (42.0)	10 (40.0)	11 (44.0)
Black	15 (30.0)	10 (40.0)	5 (20.0)
Asian	3 (6.0)	0	3 (12.0)
Native American	1 (2.0)	1 (4.0)	0
More than one race	3 (6.0)	1 (4.0)	2 (8.0)
Other	2 (4.0)	1 (4.0)	1 (4.0)
Missing	5 (10.0)	2 ( 8.0)	3 (12.0)
Hispanic/Latino	18 (36.0)	10 (40.0)	8 (32.0)
Not Hispanic/Latino	32 (64.0)	15 (60.0)	17 (68.0)
Heart Disease	1 (2.0)	1 (4.0)	0
Hypertension	15 (30.0)	9 (36.0)	6 (24.0)
Depression	19 (38.0)	11 (44.0)	8 (32.0)
Diabetes	7 (14.0)	2 ( 8.0)	5 (20.0)
Back pain	34 (68.0)	16 (64.0)	18 (72.0)
Osteoarthritis	11 (22.0)	5 (20.0)	6 (24.0)

SD=standard deviation; Min=minimum value; Max=maximum value

Primary and Secondary Outcomes

PAP use at T1 and T3 for the immediate (n=25) and waitlist control groups (n=25) are presented in Figure 4. The estimated average hours per night over a 7-day period (primary outcome) increased by 1.33 hours from T1 to T3 (+79.8 minutes; 95% CI: 0.54, 2.12 hours; p = .0014) for the immediate group but decreased by 0.17 hours (-10.2 minutes; 95% CI: -0.96, 0.62 hours; p = .67) for the waitlist control group, with a between-group difference in change of 1.50 hours (90 minutes, 95% CI: 0.38, 2.62 hours; p=.0097).

Figure 4. Positive airway pressure device usage at T1 (baseline) and T3 (35 days following randomization) timepoints for immediate and waitlist groups.

PAP=positive airway pressure. Panel (A) displays the primary outcome, the estimated mean hours of device use per night over a 7-day period, with standard errors (SE). The estimated change in mean hours of use per night was greater for the immediate compared to the waitlist group (1.50 hours, 95% CI: 0.38, 2.62; p<.01 ). Panel (B) displays estimated mean number of days (with SE) the device was used for 4 hours or more over a 7-day period, which also shows a significant improvement in number of days the device was used for 4 hours or more over a 7-day period (1.20 days, 95% CI: 0.06, 2.34); p=.04). These estimates are from linear mixed-effects models.

The estimated average number of days with PAP used ≥ 4 hours over a 7-day period (secondary outcome) increased by 1.32 days (95% CI: 0.51, 2.13; p = .0019) for the immediate group and remained around 0.12 days (95% CI: -0.69, 0.93; p = .77) for the waitlist control group, with a between-group difference in change of 1.20 days (95% CI: 0.06, 2.34; p = .0401). This represents an increase of 19% days ≥ 4 hours out of 7 days (i.e., 1.32/7 days = 18.9%) for the immediate and 1.7% (0.12/7 days = 1.7%) for the waitlist control, which is a between-group difference of 17.2% of days ≥ 4 hours.

The estimated mean number of days with PAP used over a 7-day period (secondary outcome) increased by 1.36 days (95% CI: 0.44, 2.28; p = .0045) from T1 to T3 in the immediate group but decreased by 0.52 days (95% CI: -1.44, 0.40; p = .26) in the waitlist control group. The between-group difference in change was 1.88 days (95% CI: 0.58, 3.18; p = .0054), favoring the immediate group.

Exploratory Outcomes and Measures

No significant differences in change from T1 to T3 were observed for PROMIS Sleep Disturbance (estimated mean change: -0.74 vs. -0.04 for immediate vs. waitlist, respectively; p = .57), for PROMIS Sleep-Related Impairment (-1.21 vs. 1.58, respectively; p = .09), or for ESS (-1.70 vs. -0.44, respectively; p = .35). All participants (25 in the immediate and 25 in the waitlist) attended the check-in sessions, which ranged in length from approximately 10–30 minutes (depending upon loquaciousness of participant).

Blinding

The PAP modem data are automated and the research staff who downloaded the reports and were blinded to group assignment were not able to change the data on the modem reports.

Adverse events

The most common adverse events reported were skin irritation from the wearable device (11 reports from T1 to T5, which was the follow-up time point for the waitlist control group), which was an anticipated adverse event noted on the informed consent document. One individual reported a bruise from the wearable. One individual was hospitalized for an unrelated event (leg injury and infection).

DISCUSSION

This pilot trial demonstrated that the mPAP program is a promising intervention for increasing PAP use, with results showing an increase in objective PAP use that is comparable in its effects (+79 minutes) to behavioral interventions for PAP adherence³ and larger than supportive interventions. This “low touch” program that has the potential to be largely automated employs a weekly patient-facing report containing pulse oximetry data collected by consumer wearables as well as information on the participant’s PAP current use (from objective modem data) and OSA symptoms (reported by the patient). Although no changes in sleep or daytime symptoms were observed, the improvements in PAP use in this study with a modest sample size are promising.

The possible mechanisms for the observed improvements in PAP use may be explained by the Health Beliefs Model.¹³ An increased awareness of perceived susceptibility to low oxygen levels, especially very low oxygen levels, when PAP is not used and perceived benefits (normal or near-normal oxygen levels as well as fewer OSA-related symptoms) when PAP is used could explain the changes in PAP use observed with the mPAP program. The reports—with the oxygen data juxtaposed to PAP use and OSA symptoms—may have served as cues to use the PAP device more frequently and for longer periods.

Many behavioral programs (e.g., focused on reducing claustrophobia) have been developed to increase PAP use,³ and PAP manufacturer apps that feedback PAP use data have been available for many years. However, leveraging the features of the increasingly popular consumer wearables to improve PAP use, to our knowledge, is a new approach to improving PAP use. The more recent availability of minute-to-minute pulse oximetry data from these devices provides an opportunity to show patients changes in their oxygen levels, without the need for a finger probe and by using a wrist-worn device that may be more comfortable and conventional than a device that covers the fingertip.

The pilot trial identified several issues that would need to be addressed prior to launching a full-scale trial testing the mPAP program. For instance, one surprising finding from our pilot trial was the number of reports of skin irritation from the wearable device. Some but not all devices provide minute-to-minute oxygen saturation data and have been independently validated for measuring oxygen saturation levels.¹⁴ A study of the Fitbit device in patients with chronic lung disease (our pilot trial excluded patients with lung conditions other than OSA), reported that Fitbit data were higher than the Nonin WristOX2 3150 oximeter device (+3.03 +/-1.67).¹⁵ Another potential concern is the accuracy of the pulse oximetry readings derived from both reflectance and transmittance photoplethysmography for patients with higher melanin concentration in their skin, an issue that could potentially be addressed by narrowing the spectral bandwidth to include less red light, which has greater impact on oxygen saturation calculations.^14,16,17 The Food and Drug Administration (FDA) recently approved an over-the-counter pulse oximetry device that the manufacturer (Nonin) states is accurate across all skin tones.¹⁸ Our program did not attempt to use the pulse oximetry feature of wearables to diagnose or treat patients; the purpose of providing the pulse oximetry readings is to stimulate use of PAP devices, which ideally should be worn for the entire sleep period on a regular basis. Nevertheless, a patient could be falsely reassured that their oxygen levels are normal without using their PAP device, if the wearable consistently provided inaccurate information.

The mPAP program in its current form requires the research team to assemble the weekly reports by combining three data streams (consumer wearable pulse oximetry data, PAP modem data, and participant’s survey data). The reports were developed so that they could be automatically produced in a future project. Delivering the reports to the participants would ideally be through a patient app (such as the Department of Veteran Affairs Share My Health Data app) where the participant would both provide patient-reported outcomes and receive the weekly report containing the different streams of data. PAP manufacturers, who already provide patients with information about PAP use, leak, and pressure, are well-positioned to display oxygen saturation alongside PAP use on their patient-facing PAP apps.

Limitations of our pilot trial include the modest sample size, lack of diversity in the sample, and single site nature of the trial, which limit generalizability of the findings. Attempts to blind participants to their group assignment were made (e.g., not providing details about the timeline), but participants may have deduced their group assignment. In a future full-scale trial, attempts should be made to ask participants to guess their group assignment. We found that the waitlist control design required a long commitment from patients in order to follow them through the delayed posttreatment period (T5), which limited enrollment of patients who may have been candidates for the trial. Future studies should also assess digital health literacy to support future implementation, as patients with lower familiarity with smart devices may have more difficulty with the intervention. To address the accuracy of the wearable device, particularly in individuals with higher concentrations of melanin in the skin, a future study could display error margins in the reports and ensure that the device provides accurate information across a wide range of skin colors. Finally, the measurement timepoint was 7 days after the end of intervention, which is a short follow-up time-period, and the intervention was only 28 days, which is relatively brief considering that PAP therapy is generally lifelong.

CONCLUSION

This pilot randomized, waitlist control trial of a novel program involving consumer-grade wearable devices to improve PAP use shows that the weekly reports that include pulse oximetry data are a promising intervention for increasing PAP use. A future full-scale trial that follows participants for a longer period is needed that addresses the limitations described above and uses a device that is accurate and acceptable to patients for long-term use. Additional studies are also needed to assess the potential for healthcare providers to incorporate pulse oximetry data collected by consumer-grade wearable devices into their sleep practices.

Statement of Significance.

This pilot trial tested a novel use of the pulse oximetry feature available on some consumer-grade wearable devices to increase use of positive airway pressure (PAP) devices among adults with obstructive sleep apnea (OSA) and insufficient PAP use. The widespread and growing use of consumer-grade wearable devices offers an opportunity to leverage the features on these devices to tackle a common, problematic issue among individuals with OSA—insufficient use of PAP therapy. The randomized controlled design provides support for future research involving these devices.

Data availability

The data for this study cannot be shared publicly because the institutional review board did not approve sharing of data with others outside the study.

Institution where work was performed: VA Greater Los Angeles Healthcare System

All authors have seen and approved the manuscript