Rationale, Design, and Development of SleepWell24: A Smartphone Application to Promote Adherence to Positive Airway Pressure Therapy among Patients with Obstructive Sleep Apnea

Abstract

Background:

Positive airway pressure (PAP) therapy is the gold standard treatment for obstructive sleep apnea (OSA), a chronic disorder that affects 6-13% of the adult population. However, adherence to PAP therapy is challenging, and current approaches to improve adherence have limited efficacy and scalability.

Methods/design:

To promote PAP adherence, we developed SleepWell24, a multicomponent, evidence-based smartphone application that delivers objective biofeedback concerning PAP use and sleep/physical activity patterns via cloud-based PAP machine and wearable sensor data, and behavior change strategies and troubleshooting of PAP therapy interface use. This randomized controlled trial will evaluate the feasibility, acceptability, and initial efficacy of SleepWell24 compared to a usual care control condition during the first 60 days of PAP therapy among patients newly diagnosed with OSA.

Discussion:

SleepWell24 is an innovative, multi-component behavior change intervention, designed as a self-management approach to addressing the psychosocial determinants of adherence to PAP therapy among new users. The results will guide lengthier future trials that assess numerous patient-centered and clinical outcomes.

1. Introduction

In American adults, the prevalence rate of obstructive sleep apnea (OSA) ranges from 6 to 13%[1], but among overweight/obese adults, the prevalence rate is 41%. [2] With the ever increasing - obesity rates, the incidence of OSA will likely increase in concert. Insufficiently treated OSA increases the risk for cardiovascular disease,[3,4] poor glucose metabolism,[5] stroke,[6] and premature death.[6] Positive airway pressure (PAP) therapy is the “gold standard” treatment for OSA and its use can reduce apnea-hypopnea index (AHI),[7] improve OSA-related symptoms (e.g., excessive daytime sleepiness and sleep disturbance),[8,9] reduce risk for subsequent disease, improve quality of life, and reduce healthcare costs.[10–13] Poor adherence to PAP therapy markedly undermines treatment effectiveness. Nearly 25% of “naïve” PAP users fully discontinue use within two weeks[14,15] and up to 50% stop use by one year.[16,17] Among those who continue PAP therapy, most are do use it throughout the entire night as prescribed.[18,19]

Existing PAP adherence programs are costly, difficult to integrate into overburdened clinical environments, and have demonstrated only moderate success.[20] Few interventions incorporate active problem-solving and self-management skills that address the psychosocial determinants of adherence. [14,21,22] A notable exception is the Sleep Apnea Self-Management Program (SASMP), an education program that has been incorporated into clinical care processes [23]. SASMP has demonstrated greater efficacy for improving PAP adherence compared with other programs but remains limited in scalability and impact on morbidity-related outcomes. [24,25]

Smartphones and wearable sensors (i.e., mHealth technologies) are emerging as important tools in clinical environments and, when used as adjuncts to chronic disease self-management programs, may enhance effectiveness.[26] Further, PAP self-management strategies may be more effective when coupled with other lifestyle interventions, particularly behaviors that occur across the 24-hour day (i.e., sleep patterns, physical activity, sedentary behavior, diet/nutrition). These behaviors may have both direct and indirect (via objective monitoring and feedback) effects on PAP adherence, OSA symptoms and outcomes, and symptom management.[27–32] Our team has developed SleepWell24, a smartphone application to improve PAP adherence. SleepWell24 links to PAP machine and wearable sensor data stored in cloud-based systems and delivers behavior change strategies found to be efficacious within SASMP for PAP adherence and behaviors across the 24 hours.[33–38]

Our objective in the development of SleepWell24 was to improve PAP adherence by using mHealth technologies to deliver an evidence-based, chronic disease self-management program that leverages behaviors across the 24 hours (hereafter, “24h”) spectrum. The purpose of this paper is to describe the development of the SleepWell24 app, along with the rationale and design of a randomized controlled trial to test its feasibility and preliminary efficacy. The aims of this trial are to (1) determine the feasibility and acceptability of SleepWell24 from patient and provider perspectives, and to (2) determine the extent to which SleepWell24 improves PAP adherence and treatment outcomes over the first 60 days of use relative to usual care.

2. Methods

2.1. Development of the SleepWell24 smartphone application

2.1.1. SleepWell24 conceptual framework.

SleepWell24 draws from both SASMP, an evidence-based program that addresses the psychosocial determinants underlying successful OSA self-management, and from the evidence that monitoring, providing feedback about, and improving 24h lifestyle behaviors (i.e., sleep, physical activity, sedentary behavior), may affect OSA symptoms and their management (Figure 1). We also include an additional mechanistic pathway that emphasizes the roles of self-monitoring and objective feedback on lifestyle behaviors in affecting OSA symptoms and treatment. Evidence suggests improvements in lifestyle behaviors (i.e., sleep, physical activity, sedentary behavior) across the 24h spectrum improve OSA symptoms. [27,28] These 24h behaviors may have a direct effect on OSA symptoms (which may, in turn, lead to enhanced symptom management), and also objective monitoring of and feedback regarding these behaviors may have an indirect impacts on OSA treatment outcomes via increased PAP adherence. Evidence suggests that clinical interventions that provide feedback from objective monitoring of health behaviors lead to improved treatment adherence. [29,33–37] Physical activity may support simultaneous and synergistic (i.e., bi-directional) changes in other health domains such as sleep,[30–32] which could contribute to a general “healthy living” schema, including PAP adherence. Adding lifestyle behavior change plus objective feedback to a PAP adherence program may provide a more robust impact on PAP adherence and outcomes than pre-existing programs.[38]

Figure 1.

Study conceptual framework. Dotted lines are secondary outcomes. Adapted from Stepnowsky, 2007.

2.1.2. Development phase.

SleepWell24 was developed using a community-embedded iterative design framework.[17,39] This process involved sharing paper and functional prototypes with Mayo Clinic interdisciplinary clinical teams through focus groups (n=6) and through interviews with OSA patients, who were either new or experienced PAP users (n=5). Clinical team focus groups discussed improving integration of SleepWell24 with current clinical practices, terminology, refining app content regarding PAP use and troubleshooting, and development of app components to enhance patient-provider communication. PAP user interviews focused on patient understanding of PAP terminology, preliminary expectations for the functionality of an app focused on enhancing PAP adherence, and app organization and refining the user interface. The Mayo Clinic Institutional Review Board approved of this phase of the study.

2.1.3. SleepWell24 components

2.1.3.1. Self-monitoring component.

Participants will track behaviors across the 24h (PAP use, sleep, physical activity, sedentary behavior, diet). Behavior tracking is automated (with the exception of diet) by establishing linkage to the ResMed (for PAP use) and Fitbit (for sleep, physical activity, sedentary behavior) application programming interfaces (APIs). Sleep, physical activity, and sedentary behavior data in near real-time were developed to be viewable through an interactive “Activity Log” that is pre-populated with Fitbit data (Figure 2). The “Activity Log” displays categorized time allocations of activity and sleep in five-minute time blocks over the previous 18 hours. Sleep periods are initially determined based upon Fitbit algorithms to identify sleep windows and can be revised by the participant using a standardized sleep diary available within the app. Activity categories are differentiated based upon metabolic equivalents (METs) provided through the Fitbit API, where sedentary is <1.5 METs, light-intensity physical activity is ≥1.5 METs and <4.0 METs, and moderate-vigorous physical activity is ≥4 METs. A simple algorithm is applied to detect periods when the Fitbit is not being worn (based upon minutes with zero step counts and undefined heart rate readings).

Figure 2.

Screenshots of SleepWell24 app, Activity Log.

2.1.3.2. PAP adherence component.

This component includes OSA and PAP education, tools and strategies to overcome barriers to PAP use, daily and weekly PAP adherence data (via ResMed API) with prompt goal setting for increasing usage, daily monitoring of benefits and problems experienced with PAP treatment (e.g., mask discomfort), encouraging problem solving and offering strategies, goal setting, and action planning with realistic behavior change. Daily PAP usage, leakage, and AHI data are provided to the participant as a home screen for the component (Figure 3, Panel C). A self-guided approach[40] to exposure therapy with a goal-setting feature is included, based on previous work,[41] in case patients indicate claustrophobic reactions. In addition, the primary focus of this component is to enhance self-efficacy and positive outcome expectations for PAP use, as potentially crucial mediators for increasing PAP adherence. This component is personalized to participant and wearable device (PAP machine; Fitbit) inputs. The purpose of these features is to facilitate self-guidance through behavioral strategies that support PAP adherence and to give individualized strategies based on objective PAP data and self-reported symptom management.

Figure 3.

Screenshots of SleepWell24 app.

2.1.3.3. Sleep.

The SleepWell24’s sleep component emphasizes fundamental and evidence-based sleep education, hygiene, and elements of stimulus control (i.e., regular sleep-wake schedule, minimizing naps, going to bed when feeling sleepy).[42,43] A standardized daily sleep diary is used to complement passively monitored sleep metrics from the Fitbit, and the app provides personalized feedback regarding daily and weekly sleep duration and quality metrics. Sleep duration and efficiency metrics are provided to the participant as a home screen for the component (Figure 3, Panel D). A full description of this component is described elsewhere. [17]

2.1.3.4. Physical activity and sedentary behaviors.

The physical activity and sedentary behavior components are based on social cognitive theory and use (a) self-regulatory, scheduling-based strategies to reduce sedentary time and (b) user-generated tips and goal-setting to gradually increase the frequency and duration of moderate-to-vigorous physical activity.[44,45] The context of sedentary time can be specified by the participant as either screen time, work, transport, socializing, or other. Weekly minutes of moderate-vigorous physical activity and sedentary time context are presented to the participant as a home screen for the component (Figure 3, Panels B and E). A full description of these components is described elsewhere.[17]

2.1.3.5. Diet.

Dietary intake strategies follow recommendations from the Diabetes Prevention Program dietary modification approach[46] with a focus on dietary quality over calorie restriction. Strategies include mindful eating, portion control, healthfulness (e.g., intake of fruit & vegetables, saturated fats, and added sugars), better choices of meal and snack items, healthy food preparation techniques, careful restaurant selection, and daily or weekly (selected by the participant) tracking of weight. Due to an increased burden of daily tracking of diet, this component provides simple tracking of dietary intake adapted from the “MyPlate” plan in the 2015-2020 Dietary Guidelines for Americans[47] (Figure 3, Panel A).

2.1.3.6. Patient-provider communication.

SleepWell24 also includes a “show my provider” screen. The dashboard-like screen displays relevant PAP (e.g., usage, AHI) and sleep (e.g., total sleep time, sleep efficiency) metrics, and most frequent problems experienced with PAP therapy since initiation (Figure 4). The precise metrics developed were driven by clinical provider focus groups conducted during the development phase. This component was developed to enhance communication between patients and providers at the PAP follow-up clinic visit (31-45 days post-PAP therapy initiation). This concise information is meant to assist the patient in describing the symptoms and behaviors related to PAP adherence, to communicate metrics relevant to the patient, and to indicate whether mask or other adjustments are necessary. We will assess the utility of this component during patient and provider post-treatment interviews.

Figure 4.

Screenshots of ‘Show My Provider” screens within SleepWell24 app.

2.2. Study design

In this randomized controlled trial, patients (N = 110), newly diagnosed with OSA and prescribed PAP therapy, will be randomized to one of two groups: (a) Usual care + a wrist-worn wearable sensor and its associated smartphone application (UC+WS); or (b) SleepWell24 + wrist-worn wearable sensor + usual care (SleepWell24). The SleepWell24 intervention and the UC+WS arm will be delivered for 60 days following the participant receiving their PAP machine from their designated durable medical equipment (DME) vendor. The primary adherence goal of SleepWell24 is to initiate and sustain PAP use each night, all night.

We hypothesize that the incorporation of evidence-based self-management strategies and real-time objective monitoring feedback built into SleepWell24 will be acceptable to patients, and foster the improvement of PAP adherence within the first 60 days of use.

The Mayo Clinic and Arizona State University Institutional Review Boards reviewed and approved of the protocol described.

2.3. Study setting and recruitment

Participants will be enrolled through the Mayo Clinic Centers for Sleep Medicine (CSM) in the greater Phoenix, AZ, USA and Rochester, MN, USA regions, and randomized to either the UC+WS or SleepWell24 groups.

Potential participants will be identified either through direct sleep medicine professional (SMP; e.g., physicians, nurses, technicians) referral, or via medical chart review of clinic patients that either completed a diagnostic polysomnography (PSG) study for OSA, started on PAP therapy during their diagnostic PSG study, or met criteria for PAP therapy prescription following their home sleep testing. Clinical research coordinators (CRC) will review patients’ AHI/respiratory disturbance index (RDI) scores and OSA diagnosis results for eligibility, and they will contact potential participants by phone or at the clinic visit. Recruitment is designed to follow common clinical practice, which includes either meeting with potentially eligible patients during their in-person PAP orientation/prescription visit, or scheduling an in-person visit with the patient after he/she receives a phone call PAP prescription from their SMP (Figure 5).

Figure 5.

Flow Diagram of SleepWell24 study.

2.4. Study eligibility

Participants will be asked to participate in the study if they have a first-time diagnosis of OSA (AHI/RDI ≥5) determined by clinical diagnostic testing via laboratory or home-based sleep study, and prescribed PAP therapy with the ResMed Airsense 10 device. They must currently own and use an appropriate Apple (≥iOS9) or Android (operating system ≥4.2) smartphone device and agree to be randomized to one of two groups, using a smartphone application and wearable wrist sensor. Other inclusion criteria include (1) ages ≥ 21 years, (2) able to read and understand English, (3) prescribed continuous or auto-titrated PAP, and (4) not have other acute or severe health, cognitive, or psychological conditions that would deter participation. Participants also will be considered ineligible if any of the following apply: (1) do not agree to be randomized; (2) currently participating in other lifestyle programs (e.g., active, formal weight loss program or research study; smoking cessation program, etc.); (3) lost ≥ 10lb over the past 4 weeks; (4) are known to be pregnant, lactating, or trying to become pregnant; (5) decide to not use a ResMed Airsense 10 device for their PAP therapy; (6) use high-dose benzodiazepines; (7) use daily opioid medication at night; (8) unwilling to discontinue use of any current wearable sensor for the duration of the trial; (9) previous PAP use; (10) documented history of treatment/referral for claustrophobia; (11) planning to travel for more than seven consecutive nights during the trial; (12) currently engaging in shiftwork defined as night shift or rotating day and night shifts; or (13) unwilling to have an in-person follow-up appointment at the CSM in either Arizona or Minnesota.

2.5. Study enrollment, randomization, and pre-trial set-up

Upon confirmation of eligibility, the CRC will arrange to meet with the patient during which time they will describe the study, obtain informed consent, complete baseline measures, and provide two wrist-worn wearable sensors: 1) Geneactiv device, for up to 7 nights of pre-trial assessment of sleep and physical activity patterns and 2) Fitbit Charge 2 band, to provide objective feedback of patient sleep and physical activity patterns during the 60-day trial. Demographic information, vital signs, anthropomorphic measures, cognitive and neurological functioning, medical, and other health-related data will be collected at the baseline assessment. In addition, after the clinic visit, all participants will be asked to complete a baseline online survey of other sleep and health information that staff will email via the Qualtrics survey platform. All information will be directly recorded or transferred (i.e., Qualtrics) to the REDCap platform for data management and storage. After completion of the baseline assessment, participants will be randomized to either UC+WS or SleepWell24 through the REDCap platform. A block randomization schedule stratified by AHI/RDI (<15 vs. ≥15) will be constructed with a block size of eight, and implemented within REDCap to result in 1:1 intervention to control ratio with a balance within these groups for AHI/RDI severity. After randomization, the appropriate app(s) (i.e, SleepWell24 + Fitbit for participants randomized to the SleepWell24 intervention group, and Fitbit for participants randomized to the UC+WS group) will be downloaded to the participant’s phone, but not activated until Day 1 of the trial when the patient receives their PAP device from their local DME vendor.

Between the time of baseline measures and Day 1 of the trial (received PAP machine and app initialization), integration with each participant’s respective DME vendor will be established. Each respective DME vendor will be contacted and provided instructions on how to add the study as an integration partner (IP) in the ResMed AirView system, a cloud-based patient management system for online patient monitoring. Once the integration is complete, we will collect PAP device data through cloud-based linkages with the ResMed API irrespective of the treatment assignment. This also allows for PAP data to be automatically populated to the SleepWell24 app for participants randomized to the active intervention as objective biofeedback to review and monitor daily.

As integration is being established, the study staff will communicate with enrolled participants regarding when they are scheduled to receive their PAP machine from their DME vendor. Once that date is ascertained, the study staff will schedule a phone call with the participant within three days of receiving their PAP machine. The Fitbit Charge 2 band and app(s) that were initially downloaded onto their smartphone and introduced during the clinic visit will be set-up, activated, and toured briefly through a phone call with an accompanying detailed document of content. All participants will receive detailed instructions to set up the Fitbit. Once the setup with the Fitbit Charge 2 band and app(s) is completed, connection will be verified for both. Participants will be instructed that Bluetooth connection is required to receive their PAP and Fitbit data.

2.6. Interventions

2.6.1. Usual care control arm (UC+WS).

Per usual care, all participants will receive instructions/education on PAP use, multiple mask fittings, and encouragement to use PAP every night, and staff is readily available in the event of problems. Control participants will also receive a Fitbit Charge 2 band to use during the study and will be guided through downloading the Fitbit application on the participant’s smartphone. UC+WS participants will observe their activity patterns via the features and capabilities of the Fitbit application. We are including the wearable monitor to isolate the effect of SleepWell24 on PAP adherence from potential novelty effects due to receiving a generic health behavior change application and other non-specific intervention effects. The Fitbit platform does provide some interactive and individualized features focused on sleep and physical activity, but only a limited set of evidence-based behavior change strategies,[48] and none focused on PAP adherence. In contrast, SleepWell24 is built solely upon evidence-based behavior change strategies (see 2.5.1) and provide PAP-related feedback and adherence components. UC+WS participants also will authorize access to their Fitbit as well as their PAP device data through cloud-based linkages to the Fitbit and PAP device manufacturer (i.e., ResMed) APIs. We will monitor API data for UC+WS patients from both devices. However, they will not receive the objective feedback of their PAP usage unlike the participants randomized to the SleepWell24 intervention.

2.6.2. SleepWell24 intervention arm.

In addition to usual care and the Fitbit Charge 2 band, participants randomized to the treatment arm will receive the SleepWell24 application. Upon randomization, participants will be instructed to download the Fitbit application onto their smartphones to allow for continuous syncing of Fitbit data to the Fitbit API. However, participants will be asked to not use the Fitbit app directly for the remainder of the trial as all Fitbit data will be viewable within SleepWell24. SleepWell24 is programmed for Android operating systems ≥4.2 and Apple iOS ≥ 9. SleepWell24 meets privacy and HIPAA standards. It never collects location or other contextual data. Because our server system requires only ID entry to track users, all data captured are de-identified. Privacy and technology officers have reviewed and approved the data transfer and storage security procedures.

2.7. Measures

Data will be collected primarily at baseline, 30, and 60-days after the participant’s first night of PAP use. In-person assessment will occur at baseline before randomization. All other assessments will be administered via Qualtrics except for a few measures (e.g., cognitive functioning, anthropometric measures, and vital signs) captured during a regularly scheduled clinic follow-up visit that occurs 31 to 45 days following the first night of PAP use per Medicare/Medicaid compliance requirements. Continuous monitoring of other key mediator and outcome variables (e.g., PAP use, sleep and physical activity data from the Fitbit sensor) will be conducted throughout the intervention period. Table 1 shows assessments taken at each time point and period.

Table 1.

SleepWell24 trial measures.

Measure	Baseline	Day 30	Day 60†
Demographics	◊	•
Primary outcome: treatment credibility and acceptability
Recruitment/retention	•	•	•
App usage*	•	•	•
Treatment acceptability (participant)	◊	◊	◊
Patient and provider interviews			◊
Secondary outcome: pap adherence
PAP use*	•	•	•
Treatment outcomes (exploratory)
Height, weight, BMI, and waist circumference	•		•
Self-reported weight*		◊	◊
Epworth sleepiness scale*	◊	◊	◊
Insomnia severity index*	◊	◊	◊
Functional outcomes of sleep questionnaire	◊		◊
Cognitive and neurological impairment	•		•
Hospital anxiety and depression scale	◊		◊
PROMIS global health scale	◊	◊	◊
Mediators (exploratory)
PAP self-efficacy, outcome expectations, processes of change, social support	◊	◊	◊
Sleep, sedentary, and physical activity patterns*	•	•	•
Nutrition/diet consumption and habits	◊	◊	◊

• Objectively measured, ◊ Survey- or interviewer-administered.

^*Refers to measures that will be tracked continuously throughout the intervention period.

^†CMS requires patients to be seen 30–60 days following PAP initiation, assessment will coincide with this clinical visit; however, PAP use will be monitored the full 60 days.

2.7.1. Primary outcomes: feasibility and acceptability

2.7.1.1. Recruitment/retention

will be assessed by tracking screening, reasons for refusals, and dropout rates.

2.7.1.2. SleepWell24 usage.

Participants randomized to the SleepWell24 arm will have their interactions with SleepWell24 recorded using built-in capabilities of the application. A variety of detailed usage statistics will be recorded, such as the number of times SleepWell24 was launched, self-monitoring use, and use of specific functions.

2.7.1.3. Treatment acceptability and credibility.

All participants will complete an 8-item adapted version of the Therapy Evaluation Questionnaire (TEQ)[49], a measure originally designed for use with behavioral sleep interventions. This measure contributes to the primary outcomes of feasibility, acceptability, and credibility. At baseline, after initial exposure to SleepWell24 through the introductory “tour”, the SleepWell24 participants will, using 5-point Likert-type response scales, provide their perceptions regarding the 8 attributes of treatment acceptability: logical, appropriate, easy to use, helpful, perceived effectiveness, confidence in continued use after treatment, and likelihood and importance of recommending to others. The TEQ will be administered again at post-treatment.

We will also conduct post-intervention semi-structured interviews with participants randomized to SleepWell24 to query patient-provider communication and satisfaction with the user interface of the app. Interviews will be conducted over the phone by trained staff from the Mayo Clinic Office of Patient Education using an interview guide. All interviews will be audio-recorded and transcribed for analysis. In addition to obtaining opinions of participants, we also will conduct interviews with Mayo Clinic health providers and clinical staff (outpatient providers and sleep medicine physicians, nurses, and technicians) asking them to describe their interactions with SleepWell24, and to identify barriers and facilitators to integrating the application into clinical practice.

2.7.2. Secondary outcomes

2.7.2.1. PAP adherence.

Minutes spent using PAP will be measured objectively each night and extracted via the ResMed API. We will also collect other relevant nightly PAP data, including AHI and leak rate.

2.7.3. Exploratory outcomes

2.7.3.1. Body mass index (BMI).

BMI (weight [kg]/ height [m²]) will be calculated using weight (nearest 0.1kg) and height measured without shoes (nearest 1mm) at baseline and at the participants’ regularly scheduled clinic follow-up visit occurring 31-45 days post-PAP therapy initiation.

2.7.3.2. Perceived sleep outcomes.

All sleep-related outcomes will be assessed at baseline, 30-day follow-up, and 60-day follow-up. In addition, participants randomized to SleepWell24 will complete weekly measures of daytime sleepiness and probable insomnia with a survey feature within the application. Daytime sleepiness will be measured with the Epworth Sleepiness Scale, within which participants rate the likelihood of dozing off or falling asleep during eight different activities. [50] We will assess probable insomnia and its severity with the Insomnia Severity Index (ISI), a 7-item well-validated questionnaire that assesses insomnia symptoms and their impact on daytime functioning[51]. To assess the impact of sleep apnea on functional status, we will administer the short, 10-item version of the Functional Outcomes of Sleep Questionnaire (FOSQ)[52].

2.7.3.3. Cognitive and neurological performance.

All neurological and cognitive tests will be administered at baseline and during the participants’ regularly scheduled clinic follow-up visit occurring 31-45 days post-PAP therapy initiation. Neurological functioning will be measured with the King-Devick test,[53,54] a sensitive indicator of hypoxia-related changes and sleep restriction. [53,55] Executive functioning will be assessed via the Trail Making A and B test. Working memory performance will be measured with the Digit Span Forward and Backward tests.

2.7.3.4. Health-related quality of life.

Perceived global health will be measured at baseline, 30-day follow-up, and 60-day follow-up with the PROMIS Global Health Scale[56], a 10-item measure that assesses both physical and mental health.

2.6.3.5. Anxiety and depressive symptoms.

Symptoms of anxiety and depression will be assessed at baseline, 30-day follow-up, and 60-day follow-up with the Hospital Anxiety and Depression Scale (HADS)[57], a 14-item questionnaire inquiring of a variety of depressive and anxiety symptoms over the past week.

2.7.4. Exploratory behavioral and social-cognitive mediators

2.7.4.1. Self-efficacy and outcome expectations.

Self-efficacy for managing OSA and PAP therapy and expectations of outcomes related to PAP therapy will be measured with the Perceived Self-Efficacy Measure for Sleep Apnea (SEMSA)[58], a measure previously validated among newly diagnosed OSA patients. In addition, self-efficacy related to operating the PAP device regularly and expectations for receiving support from health providers will be assessed with six items scaled in 10% increments ranging from “not at all confident: 0%” to “completely confident: 100%.” The items will include the following: “I can use CPAP nightly”; “I can use CPAP nightly even if I do not feel like it”; “I can use CPAP nightly even if I experience uncomfortable side effects”; “I can operate the CPAP machine to make it more comfortable for me”; “I will get the help I need from others to use CPAP nightly”; and “I will get the help I need from the healthcare staff to use CPAP nightly.”

2.7.4.2. PAP-specific social support.

Perceived social support in using PAP therapy will be assessed with an adapted version of a scale of social support developed by Stepnowsky and colleagues. The six-item scale assesses the degree of agreement in 10% increments from “not at all agree: 0%” to “completely agree: 100%” on if there are people in the participant’s life that are available to provide support and encouragement to use PAP therapy. Social support will be assessed at baseline, 30-day follow-up, and 60-day follow-up.

2.7.4.3. PAP-specific processes of change.

PAP-specific processes of change in terms of socio-emotional facilitators and barriers to PAP use will be evaluated using an 18-item, adapted version of a measure developed by Stepnowsky et al.[59] Examples of items include “I do something nice for myself for making efforts to use PAP nightly,” “I look for information related to using PAP,” and “warnings about the health hazards of sleep apnea move me emotionally.” Response options are on a five-point scale ranging from “never” to “always.” PAP-specific processes of change will be assessed at 30-day and 60-day follow-up.

2.7.4.4. Sleep, sedentary behavior, and physical activity patterns will be monitored using two devices.

For up to seven days and nights each at baseline (between enrollment and PAP therapy initiation), days 23 through 30, and days 53 through 60 of the trial, participants will wear Geneactiv wearable sensors on their non-dominant wrist. The GENEactiv is a wave-form wrist accelerometer that is waterproof, allowing the device to be worn continuously, 24h a day. The GENEactiv accelerometer has been validated for sleep,[60] sedentary behavior, and postural allocation (i.e., sitting vs. standing), and moderate-vigorous physical activity.[60–62] In addition, throughout the trial, all participants will wear the Fitbit Charge 2 wearable sensor continuously as a component of their assigned intervention.

2.8. Sample size calculation

We have conducted power calculations based on the secondary PAP adherence outcome. We estimated necessary power for a medium effect (d=0.61) based on the average effect sizes found in previous studies examining the effects of similar interventions on PAP adherence.[24,63] Power calculations using G*power 3.1.2[64] indicated that with N=94 we will have sufficient power (i.e., 80%) to detect a medium effect of a single predictor with a two-sided Type 1 error of 0.05 for the repeated measures ANOVA analyses that form the primary methods of analysis for assessing the PAP adherence outcome.

2.9. Statistical analysis plan

2.9.1. Preliminary analyses

We will use univariate and bivariate descriptive statistics (e.g., means, standard deviations, correlations, cross-tabulations) and plots (e.g., histograms, scatterplots) to examine distributions of and associations among study variables, to screen for potential data entry errors and biologically implausible values, and to characterize the degree and patterns of missingness. Where appropriate, internal consistency reliabilities (e.g., Cronbach’s α, MacDonald’s ω) of composite measures will be estimated. For variables collected daily, intraclass correlation coefficients will be computed to characterize the degree of non-independence among observations. Preliminary analyses and plots will be generated using R version 3.6.0 or above. [65]

2.9.2. Analyses for primary outcomes.

Feasibility will be assessed using descriptive summaries including percent patients eligible among those recruited, total attrition, attrition over time, and app usage statistics. We will compare group differences on scores from the TEQ to determine treatment acceptability and credibility.

To analyze the patient interviews, two investigators with extensive experience in qualitative data analysis, will collect and analyze participant usability and opinion data through individual interviews. These investigators are not involved in the creation of the SleepWell24 application, the overall study design, or the conduct of the trial. All SleepWell24 study arm participants will be asked to complete the individual interviews. Interviews will follow a structured guide, which will include open-ended and scaling questions. The interviews will be approximately 30 minutes in length, audio-recorded, transcribed verbatim, and de-identified. Investigators will develop a coding strategy and independently code interviews using methods of content analysis (i.e., systematic process of sorting and coding information based on themes).[66] Predominant themes (i.e., personal opinions, experiences, or concerns repeated across multiple participants) will be identified and coding compared until consensus is reached. Because more than one technology is mentioned in the interviews, investigators will repeatedly return to the original transcripts to verify that the coded opinion data attributed to the SleepWell24 application was specific to that technology (e.g., versus PAP therapy or Fitbit). QSR’s NVivo 11 (QSR International, Doncaster, Victoria, Australia; NVivo 2010) qualitative data software analysis program will be used to aid in data organization.

2.9.3. Secondary outcomes analysis

To examine intervention effects on PAP usage (minutes used per night), we will first estimate parametric generalized linear mixed models (GLMMs) that test between-group (SleepWell24 vs. usual care) differences in trajectories of usage from Baseline to Day 60. These differences will be captured by Group (SleepWell24 vs. usual care) x Time (days since Baseline) interaction terms in the models. In addition to main effect terms for Group and Time, and the Group x (linear) Time interaction, these models will include main effect and interaction terms for quadratic (i.e., Time²) and cubic (Time³) temporal trends. To minimize collinearity among Time terms, a deviation score (centered) version of the Time variable will be created by subtracting the midpoint study day value (30) from the raw Time variable. Centered Time will be used in creating polynomial and interaction terms involving Time. Models will initially be fit with random effects for (person-level) intercepts and Time (linear, quadratic, and cubic) terms. If needed to achieve model convergence, random effects for Time terms will be dropped to reduce model complexity. Temporal autocorrelation among occasion-level residuals will be accounted for using a variance-covariance structure (e.g., AR [1]) selected based on tests of relative model fit. Selection of model link function (e.g., identity, log) and error distribution (e.g., normal, Poisson, negative binomial) will be guided by inspection of the distributions of values for PAP usage and model residuals and via tests of relative model fit. Parametric GLMMs will be estimated using an appropriate mixed model package (e.g., nlme, glmmTMB) under R version 3.6.0 or above.

Further exploration of between-group differences in trajectories of usage will be conducted using GLMMs that allow for estimation of time-varying effects of the intervention effect. Time-varying effect models[67] (TVEMs) can estimate the intervention effect as a smooth, flexible function of time rather than as differences in fixed polynomial (e.g., linear, quadratic) functions of time. TVEMs are semi-parametric models that allow for visualization and description of the timing of changes in the magnitude and/or sign of the between-group difference in the outcome variable. For example, a TVEM estimating the intervention effect as a function of time could reveal when and for how long PAP usage in the SleepWell24 group is significantly higher than usage in the usual care group. TVEMs will be estimated using specialized, publicly available SAS macros[68] under SAS version 9.4.

2.9.4. Exploration of intervention effects on treatment outcomes and hypothesized mediators

To examine potential intervention effects on measures of body composition (BMI, waist circumference), sleep-related outcomes (sleepiness, insomnia severity, FOSQ subscale scores), cognitive and neurological impairment, and psychological and physical well-being (anxiety, depression, global health), we will estimate ANCOVA-type generalized linear models (GLMs) in which the Day 30 and/or Day 60 value of each outcome is predicted from Group (SleepWell24 vs. usual care), adjusting for the outcome’s baseline value. Should preliminary analyses reveal significant Group x Baseline interactions in predicting Day 30 (or Day 60) outcomes, we will instead estimate GLMMs with Group x Time (e.g., Baseline vs. Day 30) interactions to characterize the magnitudes of intervention effects. A parallel modeling strategy will be followed for assessing potential intervention effects on hypothesized psychosocial (PAP self-efficacy, outcome expectancies, perceived barriers and benefits of PAP usage, goal setting, and action planning) and behavioral (dietary habits) mediators. As with the models of PAP usage, model link functions and error distributions will be guided by inspection of the distributions of outcome scores and model residuals and tests of relative model fit. GLMs and GLMMs will be estimated using an appropriate GLM or mixed model procedure under R.

2.9.5. Exploration of potential indirect effects of the intervention on outcomes

To assess potential indirect effects of SleepWell24 on PAP usage as mediated by changes in psychosocial variables and dietary behaviors, we will estimate multilevel structural equation path models (MSEMs) in which Group (SleepWell24) and a specific psychosocial or dietary behavior variable as measured at Baseline are exogenous variables, the focal psychosocial or dietary behavior variable as measured at Day 30 is the mediator, and PAP usage (measured daily) is the outcome. MSEMs will be estimated in Mplus 8.2, [69] which affords straightforward estimation of models from data with multilevel data structures and discrete outcomes (e.g. counts of minutes of PAP usage per night). Mplus does not currently estimate bootstrap confidence intervals for MSEMs, but it can produce Bayesian credible intervals, which can be used to evaluate the significance of indirect effects in these models.

3. Discussion

PAP therapy for OSA can be burdensome. The Center for Medicare & Medicaid Services requires minimum PAP device use of ≥ 4 hours per night on 70% of nights during a consecutive 30-day period within the first 3 months of use to maintain coverage.[70] Yet 46-83% of new PAP users do not meet this requirement.[71] Common issues encountered that are associated with early discontinuation are lack of active problem-solving skills regarding PAP device side effects (e.g., mask fit and comfort, claustrophobia), self-efficacy, and low outcome expectations.[12,32,33] Current PAP adherence programs are not suited for the scope of the problem nor for the primary issues that increase risk of early discontinuation because they do not address the self-management skills and psychosocial determinants that are driving poor adherence.

Our randomized controlled trial aims to test the feasibility, acceptability, and initial efficacy of SleepWell24, a smartphone application, that addresses the psychosocial determinants of adherence to PAP therapy among newly diagnosed patients with OSA. SleepWell24 uses a self-management approach that includes objective biofeedback on PAP use and lifestyle behavior engagement and evidence-based health behavior strategies for promoting adoption of PAP and lifestyle behaviors that support this therapy. We expect SleepWell24 to be feasible and useful to patients and providers and to demonstrate improved PAP adherence rates relative to usual care.

Our study design and SleepWell24 have numerous strengths that address PAP adherence issues. SleepWell24 is a comprehensive multicomponent lifestyle intervention that targets the full 24h spectrum of OSA-related lifestyle behaviors and symptoms, and their synergistic role in OSA treatment and PAP adherence. It offers integration of PAP adherence and objective behavioral data (e.g., sleep patterns, physical activity, sedentary behavior) that provides conveniently accessible feedback. The protocol and SleepWell24 itself are built to be seamlessly integrated into clinical procedures and flow to optimize its potential for clinical adoption. The educational content and behavior change strategies within SleepWell24 are built upon an evidence-based PAP adherence program (i.e., SASMP) and social cognitive theory. Lastly, the protocol meets high levels of scientific rigor that include utilizing a single-blind RCT design, numerous objective measures of study predictors and outcomes, intention-to-treat analysis, and a comprehensive set of patient-centered secondary outcomes using validated instruments. With these strengths, the study may provide valuable results that advance clinical OSA management.

Current PAP adherence programs rely heavily on face-to-face or telephone-based interactions with clinical staff. These approaches are not scalable given multiple demands on clinical staff and high cost. Given that SleepWell24 was built to be embedded into the regular flow of clinical practice, we expect it will have greater potential to be disseminated than prior programs. We expect SleepWell24 to act as an extension of the regular clinical visit, and a conduit of rapid, meaningful communication between the patient and provider of the most actionable clinical metrics within any given clinical encounter (in-person or telephone follow-up). If feasibility, acceptability, and initial efficacy are established, we plan to disseminate SleepWell24 via the Open mHealth platform,[72] which will allow other app and device developers to adapt our strategies for use in other clinical settings. This will more rapidly build the scientific basis for our collective work, and enhance dissemination and implementation.

This trial includes noteworthy study design features. We opted for an atypical usual care condition that utilizes a consumer-based wearable sensor of activity patterns. Usual care in this context typically provides no additional support beyond technical support for PAP use. This feature was necessary to control for potential non-specific intervention effects of introducing a novel smartphone application directed generally toward health improvement. Further, participants must own a smartphone to be included in the study. Most OSA patients are older adults, a demographic for whom smartphone use is rapidly growing yet ownership and general technology literacy is lower than in other age brackets. Thus, the results of the trial may be generalizable to only OSA patients who own smartphones and feel reasonably comfortable with technology to participate in a smartphone intervention. However, we have found that apps designed for older adults with their unique interests, needs, and preferences in mind, both intrigue and empower them.[39]

This study is not without its potential limitations and challenges. The design includes a short follow-up time of 60 days from PAP therapy initiation. Thus, we will not be able to assess long-term outcomes. However, prior evidence suggests that whether a patient will adopt and sustain PAP therapy is often determined within the first week of treatment.[14,15] In addition, because SleepWell24 is a multicomponent behavioral intervention we will not be able to isolate the effects of the individual components on PAP adherence. Finally, we project that we will be collaborating with multiple DME vendors when establishing access to patient PAP therapy data in both groups. DME vendors likely will vary in business model and timeliness of providing the PAP machine after prescription, thus enhancing complexity of establishing linkages between the PAP machine and the SleepWell24 intervention. This will be a key feasibility finding that will inform how future iterations of the SleepWell24 application or similar trials may be designed.

4. Conclusion

This clinical trial seeks to add to the evidence base on efficacious interventions to improve adherence to positive airway pressure, a highly effective yet often underutilized treatment for obstructive sleep apnea. The long-term goal of this work, if shown to be feasible, acceptable, and efficacious, is conduct a larger and more definitive trial to determine the longer-term efficacy (>12 months) across a more robust set of outcomes including Medicare-defined adherence (≥4h per night on 70% of nights), apnea-hypopnea and respiratory disturbance indices, cognitive health, and cardiovascular disease risk.