An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Carl Stepnowsky, PhD, Elisha Malanga, Sergio Martinez, Jamie Sullivan, MPH, Sean Deering, Cara Pasquale, MPH, and Lin Liu, PhD.
Author Information and AffiliationsChronic obstructive pulmonary disease (COPD) and obstructive sleep apnea (OSA) are 2 major chronic conditions that affect millions of Americans. When OSA and COPD coexist, they are collectively referred to as “overlap syndrome.” OSA is prevalent in 10% to 15% of the 15 million patients diagnosed with COPD. Patients with both conditions are commonly prescribed medical devices for management, continuous positive airway pressure (CPAP) for OSA, and oxygen therapy for COPD. Because our early work found that most patients with OS use CPAP but a relatively low percentage of them use oxygen therapy, this study focused on improving the use of CPAP. Although CPAP is the most efficacious treatment for OSA, its use by patients is low relative to the prescription to use CPAP whenever asleep. Suboptimally treated OSA impairs next-day functioning.
The project comprised 3 separate but related phases: (1) patient and stakeholder engagement activities; (2) focus groups to learn more from our patient community about the proposed comparative effectiveness research (CER) study; and (3) finalizing and carrying out the O2VERLAP study, which was a large-scale, CER study examining 2 interventional approaches to improve treatment device adherence and outcomes. The main aims of this randomized controlled trial were as follows:
Participants were primarily recruited from 3 communities (COPD, OSA, and PCORnet) through electronic methods. They were given access to the study website to learn more and, if interested, could choose to provide e-consent and complete a self-report eligibility questionnaire. Inclusion criteria included being aged ≥40 years, having diagnoses of both COPD and OSA, and having been prescribed CPAP therapy. Participants were then randomly assigned to 1 of the 2 groups, with outcomes assessed at baseline, after an intervention period of 6 weeks, and after a follow-up at 12 weeks. Baseline CPAP adherence data were also collected for the 30-day period before randomization. The study primary outcome was CPAP adherence, defined as the amount of time that CPAP was worn each day at the prescribed pressure; the secondary outcomes were sleep quality, daytime functioning, and daytime sleepiness.
The study enrolled 332 participants and randomly assigned 294. The mean (SD) CPAP adherence levels for the PC and RC groups were 6.1 (3.1) and 7.3 (2.4) hours/night (baseline), 6.3 (2.7) and 7.4 (2.2) hours/night (6 weeks), and 5.9 (3.0) and 7.2 (2.5) hours/night (12 weeks), respectively. The groups significantly differed in CPAP adherence at baseline (P < .001). There was no significant difference in change in CPAP adherence between the 2 study groups at either 6 weeks (difference = 0.18; 95% CI, −0.16 to 0.52; P = .29) or 12 weeks (difference = −0.05; 95% CI, −0.39 to 0.29; P = .78). There were also no significant differences in the change in patient-reported outcomes (ie, daytime functioning, sleep quality, and daytime sleepiness) at 6 weeks or 12 weeks.
In a group of patients with both COPD and OSA who used CPAP therapy, no difference was found between the provision of PC and RC. We found an unexpectedly high baseline CPAP adherence level, which meant that any improvement due to the intervention would have been very small and difficult to detect. The study was potentially underpowered to find a very small effect size, given the sample size. The high baseline CPAP adherence level may have been related to selection bias, due to the reliance on electronic recruitment methods (ie, email, social media, newsletters). Participants in both study groups were very satisfied with the care provided.
The study was designed as a large, national, electronic recruitment-only study of patients diagnosed with both COPD and OSA. Because it relied on electronic recruitment, the study was limited to patients who had access to those electronic methods of outreach. Future studies of this kind would benefit from more stringent inclusion and exclusion criteria to ensure the studies are limited to patients who are having some difficulty with CPAP use or to new users.
In the past, PCORI supported patient-powered research networks (PPRNs), which are communities of patients interested in participating in the clinical research process as part of PCORnet (National Patient-Centered Clinical Research Network: https://pcornet.org/). In 2015, PCORI launched the PPRN Research Demonstration Project initiative within PCORnet to support PPRNs in conducting comparative clinical effectiveness research on questions that are important to, and inclusive of, patients and other stakeholders. The initiative provides the foundation for the present project.
This research project comprised 3 separate but related phases: (1) patient and stakeholder engagement activities across the entirety of the project; (2) focus groups to learn more from our patient community about the proposed comparative effectiveness research (CER) study; and (3) finalizing and carrying out the O2VERLAP study, which was an ambitious, large-scale CER study examining 2 interventional approaches to improve treatment device adherence and outcomes. Figure 1 illustrates how the phases were related to each other during the O2VERLAP project period. Patient and stakeholder engagement (phase 1) was foundational for our project and took place from grant application through study dissemination. The focus groups (phase 2) were conducted before commencing the main study (phase 3).
Three Phases of the O2VERLAP Project.
Chronic obstructive pulmonary disease (COPD) is a group of progressive and debilitating respiratory conditions that affect 15 million to 25 million Americans1 and more than 300 million people worldwide.2 COPD is the third leading cause of death and the second leading cause of disability in the United States.3 Each year, COPD results in as many as 800 000 hospital admissions and 1.5 million emergency department visits.4 Obstructive sleep apnea (OSA) is a prevalent chronic medical condition characterized by repeated stops (apneas) and near stops (hypopneas) of breathing during sleep, due to collapse of the tissues in the upper airway.5 These breathing disturbances last 10 seconds or longer and cause repeated sleep disruptions and oxygen desaturations that lead to important consequences, including daytime sleepiness and increased risk of cardiovascular problems. OSA affects 17% of adults and more than 25% of older adults,6 with rates increasing in association with the obesity epidemic.7 Sleep apnea aggregates in families,8 affects all age groups, and disproportionately affects minority populations9 and those from poor neighborhoods.10 OSA requires immediate and ongoing therapy because it lowers blood oxygen levels, disrupts sleep, and is associated with hypertension (including pulmonary hypertension), myocardial infarction, stroke, atrial fibrillation, cor pulmonale, and early death. OSA also results in increased risk of depression, anxiety, cognitive issues, erectile dysfunction, irritability, daytime sleepiness, and motor vehicle crashes.11-18
Separately, COPD and OSA contribute to the morbidity and mortality of hundreds of thousands of Americans every year. However, when OSA coexists with COPD, it is referred to as overlap syndrome.19 OSA is prevalent in at least 10% to 15% of patients diagnosed with COPD.20 Although the prevalence of OSA is similar in patients with COPD as in those in the general population, individuals with both of these conditions, but who do not use continuous positive airway pressure (CPAP) therapy at night during sleep, have an increased risk of death and more hospitalizations from acute exacerbations of COPD, demonstrating the importance of OSA treatment.21
It is thought that overlap syndrome is clinically distinct from either condition alone and that patients with this syndrome have a worse prognosis compared with patients who have only COPD or only OSA, for several reasons that have important implications for diagnosis, treatment, and outcome.22 Studies that have examined the efficacy of CPAP therapy for overlap syndrome have shown that CPAP use is associated with improved walking capacity23 and longer survival in patients with COPD who are hypercapnic,24 and that higher levels of CPAP adherence are associated with better outcomes.21 However, of the approximately 80% of patients who initially accept CPAP therapy, most patients fall into a partial use pattern of 3 to 5 hours/night. Appendix A provides CPAP adherence data from studies that focused on improving adherence in new CPAP users. The overall mean adherence of the control groups of 4.0 hours/night supports the notion that CPAP is not used to the extent prescribed in the typical clinical population. The Appendix A table also shows that adherence levels in the United States tend to be lower than those outside the United States. Adherence with long-term oxygen use has a parallel story; it is beneficial the more it is used, but adherence is less than optimal, ranging from 45% to 70%.25 This evidence highlights the importance of providing the overlap syndrome patient population with the tools necessary to improve adherence to CPAP therapy.
Web-based interventions and remote data telemonitoring are now empirically supported as effective interventions for providing chronic illness care. Web-based interventions helped increase patient activation in a study of patients with multiple comorbidities.26 In a review of web-based interventions for diabetes management, successful intervention components included goal setting, personalized coaching, interactive feedback, and online peer support.27 A review of home telemonitoring interventions in patients with heart failure showed that the interventions helped reduce all-cause mortality and hospitalizations.28 The literature is evolving relatively quickly. Within the field of sleep medicine, Carl Stepnowsky, PhD, has shown that both remote telemonitoring by providers29 and an interactive web portal accessed and used by patients with OSA can improve CPAP adherence.30 More recent evidence has shown that simply providing patients with access to their CPAP data improves adherence.31 Patients who use CPAP have always been able to read limited summary data on their machines, but new technologies now allow patients to view more of their CPAP adherence and efficacy data online, along with education about how to understand this information. The O2VERLAP study's intervention was developed with this evidence in mind and therefore combined a proactive, patient-centered, peer-coaching system together with an online educational curriculum that enabled easy access to the CPAP adherence and efficacy data by both coaches and participants. The intervention was designed to address common adherence barriers using CPAP therapy, as well as adherence facilitators.
The 2 intervention groups were designed to help answer the question of whether an organization should “staff up” to provide a dedicated, multifaceted intervention and support for patients, or “staff down” and provide intervention and support only on an as-needed basis. Although the main medical decisions for CPAP therapy (eg, regarding change in pressure level or in CPAP mode) need to be made by a qualified, licensed medical professional, the interventional approach for this study was meant to be a supportive and informational adjunct to medical care provided by a licensed provider. Evidence suggests that the health care system as designed can only provide limited support for patients who are prescribed CPAP therapy, with evidence showing that CPAP adherence rates have been level for 2 decades.32 Ideally, our intervention should be deployed within a health care system; however, an argument can be made that it could also be deployed by home medical equipment (HME) companies, patient advocacy organizations, or other similar kinds of organizations that play supportive roles for patients.
The goals of this project were to carry out a large, national CER project and to learn more about doing so within PCORnet via the PPRN Research Demonstration Project initiative. The main aims of the O2VERLAP study were as follows:
As such, the study team was tasked not just with carrying out an ambitious CER project but doing so in the PCORI spirit of “doing research differently” through an extensive patient and stakeholder engagement plan (phase 1). The engagement plan started at funding application outset when the core study team went out to their communities to learn more about the interest level in this project, and the community communicated that they were very interested in this topic. From that point, a team with key patient, research, clinician, and advocacy stakeholders designed the funding application and then carried out the projects with a relatively large study team and stakeholder advisory board (SAB). In addition, we reached out to more than 40 different organizations for study promotion. This effort included significant interactions with a specific stakeholder group (HME organizations) that has traditionally not engaged with research in the past. The following section provides more information and details about project engagement activities.
Patient and stakeholder engagement (phase 1) was fundamental in this project and started by having a central role when the team came together to work on the original funding application. In the “Patient Engagement” section, we first describe the patient-community focus groups that were held as part of phase 2 and then provide descriptions of some specific patient engagement activities that were accomplished during the project period. The “Stakeholder Engagement” section focuses on important feedback provided by the SAB.
In phase 2, our team carried out a series of focus groups to learn more about the outcomes and interventions important to our patient community. The focus groups were designed to help inform the final plans for our main scientific study. A total of 17 participants (70% women, 30% men) were included; their mean age was 65.1 years (SD, 11.2; range, 47-84 years). Because the team was unsure which focus group method would be the most informative, we planned 3 types of groups: (1) teleconference (audio only; n = 6); (2) in-person (n = 4); and (3) web-based platform (n = 7). The 3 types of focus groups differed in delivery only: (1) teleconference was conducted via telephone; (2) in-person was conducted by a moderator in the same room with participants; and (3) the web-based platform was conducted using the COPD Foundation's COPD360Social platform both in real time and asynchronously.33 There was a great deal of discussion and time spent deciding how to optimally run the web-based platform focus group. The team decided to first run a 2-hour synchronous group in which study team members engaged participants in real time, which was then immediately followed by 72 hours of asynchronous time. During this time period, users could return to the platform and answer questions and provide follow-up responses at their leisure. The same questions were asked of each group.
The study team found that the transcript from the telephone and web-based focus groups produced the most coded phrases (n = 40 and 32, respectively), whereas the in-person transcript produced the fewest coded phrases (n = 16). The study team interpreted this difference as likely primarily due to the extra informal interactions that participants had in person that they were not expected to have by phone or online. In other words, participants connected more interpersonally when they were physically present with each other than when they were communicating by phone or online.
The focus groups had 2 sets of findings, 1 related to the intervention and 1 related to the planned outcome measures. In terms of the intervention, the participants were most concerned about mask fit and comfort, as a CPAP use barrier and facilitator, respectively. Other factors identified as affecting CPAP use included nasal dryness and issues concerning insurance coverage of the device. In terms of patient-centered outcomes, it was very clear that participants who experienced a disturbed night's sleep had an impaired ability to function during the following day. Some used the expression “not being able to get out of bed.” Others used the phrases “not being able to do the kinds of activities that I want to do” and “not having the energy I need during the day.” Many mentioned the need to take naps the following day to manage their fatigue, which cut into time available to do typical day-to-day activities. It was clear from the focus groups that the most important outcome for patients with overlap syndrome was their ability to function during the day. Because of this finding, our study team elevated the patient-reported outcome (PRO) of daytime functioning in the main scientific study. More details can be found in the focus group manuscript in Appendix B.
The study included numerous opportunities for patient engagement activities given that it was funded as a PPRN Research Demonstration Project. During the process of pulling together the funding application, 2 patient advocacy organizations had patient members who were part of the team, so patients had the opportunity to provide input from as early as the project planning stages. Other key areas of patient engagement included (1) incorporating the patient perspective throughout the project and (2) paying special attention to recruitment efforts. The project's first activity was to assemble the aforementioned series of focus groups, which were designed to obtain the patient's perspective on several key study components. Although not specifically covered in the previous section, it was in no small part patients finding patients via word of mouth that allowed the project team to successfully carry out those focus groups.
The patient stakeholders also provided support for the study's recruitment efforts. Not only did they assist with the preparation of all participant-facing materials (including emails and social media posts) to ensure they were readable and understandable, they also were instrumental in O2VERLAP study promotion efforts:
SAB meetings were held every 2 to 3 months throughout the project. A total of 10 meetings were convened from October 2016 to November 2019. During each of the meetings, the study team would present SAB members with the study's progress and challenges and ask them to share ideas and additional efforts that could be implemented to overcome those challenges. At the last SAB meeting, held on November 19, 2019, the study team presented a study overview with results and preliminary data analysis, because study recruitment had recently ended. That final meeting was also used to thank all SAB members for their engagement with the project.
Stakeholder members maintained their involvement with the project over time. Our first meeting on October 31, 2016, started with 19 of our 20 members in attendance. The attendance at each meeting after that ranged from 14 to 16 members. The following sections describe several of the broader topics discussed with the SAB and how the study and its methods were improved as a result.
The group brainstormed ideas related to how the study team might plan for national electronic recruitment. Ideas ranged from message awareness to how to identify specific groups of patients. The group also discussed how the team might obtain a representative sample of patients with overlap syndrome. Importantly, early feedback from the SAB informed the eligibility criteria and planned focus of the study to include patients diagnosed with both medical conditions (ie, COPD and OSA) and who were prescribed and are currently using, to some extent, both oxygen therapy and CPAP therapy. Feedback, particularly from our clinician stakeholders, suggested that this would result in a limited sample of only the most ill. The recommendation was to loosen the criteria for oxygen therapy use. All stakeholders agreed this would be an important decision for the study. The study team adopted this recommendation.
The findings from our focus groups were presented in detail to the SAB regarding raising the importance of daytime functioning. After much discussion, the SAB members agreed to use the Patient-Reported Outcomes Measurement Information System (PROMIS) instrument, which measure health outcomes from the patient perspective and has the benefit of providing an additional non-sleep-specific measure of daytime functioning. In other words, the SAB and study team agreed that the Functional Outcomes of Sleep Questionnaire (FOSQ) was the predominant measure within the sleep field but that having an additional measure from PROMIS would benefit the study. The group decided to use the Sleep-Related Impairment scale from PROMIS. Because of the advantages that the PROMIS scales afforded, the group ultimately decided to use additional PROMIS measures for the study. More details can be found in the Methods section.
Participant reimbursement, from the patient perspective, was an ongoing topic of conversation in the SAB meetings before study start. In 1 of the webinars listed in Table 1, participant reimbursement was a significant component. The SAB discussed the potential levels of reimbursement for a patient's time as a study participant and compared with other similar studies. The study team and SAB members together decided on the participant reimbursement of $25 for this study, in the form of an online gift card, on completion of each of the 3 surveys, conducted at baseline, 6-week follow-up, and 12-week follow-up.
Dates and Names of Completed O2VERLAP Project Webinars.
The topic of study promotions was likely the most discussed topic across all SAB meetings. The reason for this was clear: study promotion was 1 of the study's most significant concerns. In addition, many SAB members also had access to patients who were diagnosed with 1 or both of the study's medical conditions (ie, COPD and/or OSA). So, as the study team developed the methods of study promotion, we would discuss the ideas and approaches with the SAB for feedback. Once a study promotion toolkit was finally developed for the study, SAB members were asked for assistance. Several of the most important topics the SAB provided feedback on included (1) type of messaging for study promotion and (2) recommendations concerning study promotion details, including how to work with external organizations and respect the existing communications schedules of the organizations with their communities.
Because of this critical feedback, we structured our external calls and presentations around these ideas. For example, we would first find out from an external organization how often they communicated with their community and by what methods. By finding common ground, we could then best collaboratively explore and negotiate what that organization might be able to do for study promotion. Some partners were able to go above and beyond initial expectations of what they could do, whereas other partners were unable to engage in study promotion. We generally found that if an organization or group had experience communicating with their communities about research opportunities, they were able to help with study promotion; those external groups who had little or no experience generally found it difficult to take this next step. is HMEs were an example of this latter group. In the end, HMEs had limited communications with their communities (ie, consumers), and very few had communicated research opportunities. That said, the HMEs were all very cordial and wanted to help; in the end, however, promoting research studies did not align with their business priorities.
One additional point to make on study promotion is that 1 SAB member emphasized the potential yield the study could get by promoting it to the Veterans Health Administration (VHA), which is the largest integrated health care system in the United States, with more than 9 million patients. In addition, OSA incidence is known to be high in patients receiving care from the VHA. At the time, the study had already enrolled approximately 5 veterans. The SAB and study teams discussed this possibility at length, but the key issue was that most VHA medical centers do not allow non-VHA studies to be advertised. Dr Stepnowsky had checked with his local VHA medical center to confirm this was the case.
The topic of CPAP data sharing was the second most challenging aspect of the study. The original funding application had planned to use a simple proxy measure of CPAP use, but the planned device had limitations. Once the decision was made by the study team, with SAB input, to focus on CPAP therapy, it was also decided that the study should obtain the CPAP adherence and efficacy data from the manufacturer's servers. This resulted in a significant challenge to the study. Although the machines and data may be owned by the patients, the HME companies are, in fact, the data stewards. What this meant was that our study team needed to help our participants facilitate the HME company's data sharing with the study. The SAB was critical in listening to the study team's plans and providing feedback on what might and might not work. In the end, we made several key changes to our methods to maximize our opportunity to have patient data shared with the study. This feedback and change in our methods resulted in an improvement in data sharing from 25% at the beginning of recruitment to 93% by study end. Feedback from the SAB suggested this could be an important patient-centered topic for a “lessons learned” or “road map” (ie, instructional) type of manuscript focused on this issue. It could also be related to how to perform real-world data research, given the increase in the use of wearable devices as an example of this data-access need for research.
The SAB provided important feedback and discussion of both the potential webinar topics and their content. The list of our 5 webinars can be found in Table 1. The webinar topics determined the primary audience of interest and all were inclusive of the patient perspective. In addition, there was much discussion about whether the webinars should be specific to the O2VERLAP study or be made more general. Because the project was funded as a PCORI PPRN Research Demonstration Project, the study team and SAB all thought that for the webinars to have the most value on PCORnet Commons, they should be kept general. The webinars were provided to PCORI for upload to PCORnet Commons.
The inclusion of the variety and number of stakeholders was important to the success of this project. Patient stakeholders from both patient communities (ie, those with COPD and those with OSA) were included from project outset and helped establish the patient focus and orientation of the team. We took advantage of opportunities for patient involvement and feedback throughout the project, with an emphasis on patient-facing materials, recruitment efforts, and webinars. The SAB provided feedback in several important study areas, including measure selection, inclusion and exclusion criteria, and CPAP data sharing. The core study team greatly benefited from the stakeholder engagement.
The O2VERLAP study was a PPRN Research Demonstration Project within PCORnet to support the PPRNs in conducting comparative clinical effectiveness research on questions that are important to patients and other stakeholders. As such, the main scientific study of the project was informed by what we learned from the focus groups, as discussed in the Patient and Stakeholder Engagement section.
The main aims of the O2VERLAP study were as follows:
The O2VERLAP study was designed to be national in scope and did not take place within any defined health care system. Primary study offices were located within the COPD Foundation and the University of California, San Diego (UCSD). The study was carried out via a web portal, which was hosted by DatStat, Inc (Seattle, WA). As a PPRN Research Demonstration Project initiative, an overarching goal of the project was to determine how a research study might be best carried out within PCORnet in collaboration with PCORnet partners and collaborators.
At the start of this project, the PCORI-funded PPRNs and Clinical Data Research Networks (CDRNs) were all fully operational. The PPRNs that agreed to join the COPD PPRN and be a part of this project included PRIDEnet (San Francisco, CA); PI Connect (Towson, MD); Health eHeart Alliance (San Francisco, CA); and the ABOUT Network (Tampa, FL). This project was unique in that it was 1 of the few PPRN Research Demonstration Projects that included a CDRN (pSCANNER; principal investigator [PI]: Lucila Ohno-Machado; UCSD, La Jolla, CA), which comprised 9 large health care systems, including the 5 University of California (UC) medical centers (UCSD; UC Los Angeles [UCLA]; UC San Francisco [UCSF]; UC Irvine; and UC Davis). The PPRNs and CDRN assisted with study outreach and recruitment.
Figure 2 shows the study home page. The study's online portal was used to educate potential participants about the study design, obligations, and study inclusion criteria. Home-page content and FAQs were carefully thought out by the study team for interested potential participants to use. The study portal also housed digital e-consent and HIPAA forms, as well as the PI's and project coordinator's contact information, so participants could reach out if they had any questions or concerns.
Screenshot of O2VERLAP Study Home Page.
The O2VERLAP study relied almost entirely on electronic recruitment methods, including emails, social media posts, electronic newsletters, website home-page banners, and interactive platforms or forums; some supplemental nonelectronic methods (ie, in-person study promotional activities) also were used, including presenting at conferences, exhibiting at health fairs, and via word of mouth. The study promotion methods used the following definitions:
Based on these definitions, a campaign consisted of sending a message via a defined method to a defined audience. A short way to express our approach is “campaign = audience + method.” Appendix C provides an extensive list of the study promotional efforts and describes the community, audience, method, number of contacts, and total reach over the duration of the project, which took place from February 2018 to July 2019. Eighteen of the 46 campaigns were deployed multiple times over the 18-month recruitment period. Appendix C also provides more details on the recruitment campaigns, including (1) description of the 3 different message types that were used; (2) description of how a reminder email sent 3 to 5 days after the initial email was considered standard practice; and (3) additional information about Facebook and Twitter posts.
Each campaign comprised an audience and a method for reaching that audience. One primary recruitment metric was the size of the audience. For most campaigns, audience size was either known or could be estimated. For example, if we had a list of email addresses, then the number of individuals we emailed was a known quantity. On the other hand, newsletters or newspapers might report an estimated circulation number. Other metrics included the number and percentage of enrolled participants. The number of enrolled participants is simply a count of the number of enrolled participants for that specific campaign. The percentage of enrolled participants refers to the number of enrolled participants divided by the audience size of that campaign.
The study web vendor (DatStat, Inc, Seattle, WA) provided 2 ways for tracking participant interaction data. The first was through their URL key pairing functionality, commonly described as “referral URLs.” With this tool, we generated URLs for participant recruitment that contain a specific piece of information, such as a recruiting site or campaign. When participants clicked the URL in an invitation email or manually entered the URL from a mailer or site recruitment poster, they were taken to the DatStat portal and the referring ID was stamped into their session. Participants who registered had their participant record updated automatically using the data stamped in the session, which effectively linked that participant to the recruitment campaign used.
DatStat Connect platform was also tied to Google Analytics and user flow was tracked with a Google Analytics Tracking ID. This gave more generic information on the types of sites (eg, social media, Google, direct URL) from which patients were being referred. Google Analytics is a product feature; thus, it is tied to all major site functions, such as registration, log-in events, and pagination, but it is not customizable to specific implementations. For specific recruitment details based on an individual study, the referral URL functionality was recommended by DatStat.
The inclusion criteria were as follows: being aged ≥40 years; being able to speak and read English; having diagnoses of both COPD and OSA; having a prescription of CPAP therapy; and having access to the internet and a personal computer, tablet, or smartphone (to complete the online study activities). In addition, the CPAP device needed to have wireless connectivity (via an internal or external modem). Exclusion criteria for the O2VERLAP study were being a non-English speaker and having a life expectancy of ≤6 months.
Signing up for the O2VERLAP study was a 2-step process: (1) registration and (2) consent. Because the consenting process was done via the study platform, hereafter it is referred to as an e-consent. The e-consent encouraged potential participants to discuss participation with their family and friends, should they want additional support. The last page of the e-consent form included a checklist for the participant to review and interactively check off. The list included confirmation that the participant understood the study design, described the low risk for participating, and confirmed that the participant comprehended and met the eligibility criteria. Those who did not sign consent were contacted by study staff via phone and email to confirm they did in fact intend to stop and not continue with consenting into the study. Recontacting this group was considered another lesson learned for the project, because during this process, we found a group of people who did want to continue to step 2 but did not because they either had technical issues moving on to the next step or were unsure of how to continue to the consent portion.
After an individual digitally signed the e-consent form, they were then prompted to take a first survey which relied on self-reported confirmation that they met the study's eligibility criteria. If an individual responded to a question in a way that indicated they did not meet the eligibility criteria, the study team was notified by email to schedule a call and confirm that the individual was truly ineligible to participate.
When an e-consent was signed, it also triggered an email to the study team to notify them and to prompt the team to reach out to the newly enrolled individual for their first study phone call with the study coordinator. The purpose of the Confirmation of Eligibility (CoE) phone call was to verbally confirm that the individual met the study's eligibility criteria. If the study coordinator confirmed that the consented individual was, in fact, eligible, the outcome was documented in the study portal through the corresponding CoE survey; the study coordinator then proceeded to complete additional forms that provided evidence of the participant's eligibility (ie, CPAP device information, medical history survey, and demographics survey). In addition to the portal tracking all participants as they signed up, the study team also kept a study screening log (ie, a password-protected Excel spreadsheet) that captured all individuals who registered and e-consented and also documented the following scenarios: those who subsequently failed to meet inclusion criteria on the initial self-reported CoE survey, those who subsequently failed to meet inclusion criteria on the CoE phone call with the study coordinator, and all who were eligible for the study and would continue on to the next tasks.
After confirming someone's eligibility, study staff had to contact the participant's HME company to request access to the participant's CPAP data. This data sharing process involved the HME company adding the O2VERLAP study's sleep lab to their participant's integrator and physicians list on the EncoreAnywhere (Philips Respironics data platform; Koninklijke Philips N.V.) for Philips devices or AirView (ResMed data platform; ResMed, Inc) for ResMed devices. This technique would allow a wireless flow of their CPAP data from the data platform to the O2VERLAP study portal, which would display the participants' nightly CPAP data metrics (ie, hours used, Apnea-Hypopnea Index [AHI], and mask leak) displayed in user-friendly graphs on the portals for the coordinator and participant to view.
Potential participants who either had an older CPAP model, did not have a device from 1 of the major CPAP manufacturers, or no longer had their device were given an opportunity to obtain a machine with a newer model either on their own or through the ASAA CPAP Assistance Program. Approximately 35 individuals received a CPAP device or accessory replacements so they could participate in the study.
The study design was a comparison of 2 intervention groups: PC and RC. Once participants met all eligibility criteria and completed the RT introductory call, they would be assigned the next available participant identification number (PIN) from a preset randomization scheme spreadsheet that was tied to a specific randomized group assignment. This process was carefully handled only by the study coordinator and allocation was tracked in the password-protected Excel randomization scheme spreadsheet. Simple randomization to the 2 groups was based on the use of a random number generator (https://www.rand.org/).
In both groups, participants first received their introductory phone call from their assigned research study RT (ie, not their personal RT). On this introductory call, RTs would follow a scripted questionnaire to review the participant's baseline CPAP adherence data and work with the participant to set 3 SMART (Specific, Measurable, Attainable, Relevant, and Timely) goals for improvement. The set goals were reviewed with the PC group at the end of the intervention. Once the RT introductory call was completed, the PIN was assigned from the preset randomization scheme spreadsheet, and the baseline survey would become available to the participant via the online portal. For PC group participants, the online curriculum opened next.
PC is considered the study intervention. If an individual was randomly assigned to the PC group, their involvement included the following:
An introductory call from an RT and a COPD Information Line coach who acted as peer coaches in health topics covered in the curriculum
Access to module 1 of the online curriculum
Weekly dyadic peer coaching calls by COPD Information-Line coaches
Access to modules 2 through 6 of the online educational curriculum, covering topics on COPD, OSA, and overlap syndrome
Access to module 7, the final module in the curriculum
COPD Information Line coach call and RT follow-up call on completion of module 7
Participants in the PC group also had online access to their CPAP adherence monitoring data to track their progress as they advanced through the study program, as well as access to a chat function in the portal to ask questions or contact the study team throughout the intervention period.
The RC group of the study was given access to an RT who would make an introductory call during week 1. Participants in the RC group were given the phone number of the COPD Information Line that they could contact to seek advice about any aspect of CPAP therapy or information about general health topics related to overlap syndrome. The RC group participants also had online access to their CPAP adherence monitoring data. They had access to general informational COPD and OSA materials via the website. The primary characteristic of the RC group was that they had access to the same educational materials as the PC group but were not required to go through them; they were also provided with the contact information for support. When the team was pulling the grant proposal together, the perspective was that of a patient advocacy nonprofit organization providing this education and support outside of the care system in an adjunctive way. From that perspective, the patient advocacy team was proactive in providing education and support, and any study participants would be reactive if they took the initiative to seek this information and support.
A previous online COPD educational curriculum was used as the model for the O2VERLAP educational curriculum. The study team met to develop the specific outline. Two sleep education specialists were identified by the ASAA to write the content under the supervision of Dr Stepnowsky, who has developed several OSA- and CPAP-specific curricula for previous CPAP adherence studies. Table 2 provides the module titles and the lessons, divided by chapters, that make up each module. There was a total of 22 chapters, 1 of which was an introduction. Of the 21 topical chapters, 11 were focused on OSA/CPAP, 7 on COPD/oxygen, and 3 on both content areas.
Content of the Online Educational Curriculum.
Of the 153 participants randomly assigned to the PC group, 120 (78%) completed all 7 modules in the online curriculum. Participants needed to complete all of the pages and activities in a chapter before advancing to the next chapter. Similarly, participants needed to complete all of the chapters within a module before advancing to the next module. There were 33 (22%) participants who did not complete all 7 modules. The following number of participants completed each number of modules: 6 modules (n = 3); 5 modules (n = 2); 4 modules (n = 2); 3 modules (n = 11); 2 modules (n = 8); 1 module (n = 2); and 0 modules (n = 4). One participant who withdrew after randomization never started the curriculum.
Both PC and RC group participants received the same scripted RT introductory coach call before randomization. Of the 153 participants, 151 (99%) who were randomly assigned to the PC group and 134 of the 141 (95%) who were randomly assigned to the RC group received the RT introductory coach call. Some participants did not receive the RT introductory call because, in the first few weeks of the study, our process was to randomly assign participants first and then complete the RT introductory coach call; as a result, 9 participants (n = 2 in the PC group; n = 7 in the RC group) advanced too far along into the intervention before training of the RT coaches was completed. The study team decided to change the workflow to complete the RT introductory coach call before randomization to prevent this from occurring again. A second scheduled RT call was made only to PC group participants on completion of the online curriculum.
Only participants who were randomly assigned to the PC group of the intervention and completed the baseline assessment received their first COPD Information Line coach phone call. Four other consecutive weekly calls were scheduled with the participant and the COPD Information Line coach, totaling 5 calls. There were no scheduled calls for the RC group, but the participants were given the contact details to reach out to the Information Line coaches for support, which represented the “reactive” component of the RC intervention. Table 3 provides descriptive data on the number of calls made by the RTs and COPD Information Line coaches by group.
Number of Information Line Coach and RT Coach Calls by Group.
The primary study outcome was CPAP adherence, which was objectively measured by the CPAP device, with data sent wirelessly to the study team. All participants were assigned a baseline survey package. Together the surveys measure several important patient-centered outcomes, including daytime sleepiness, daytime functioning, and COPD functioning. Based on our focus group findings, consultation with the SAB, and discussions with the study team, the patient-reported study outcomes included (1) daytime functioning (as measured by the FOSQ and the PROMIS Sleep-Related Impairment scale); (2) sleep quality (as measured by the Pittsburgh Sleep Quality Inventory [PSQI] and the PROMIS Sleep Disturbance tool); and (3) symptoms (as measured by the COPD Assessment Test [CAT], Epworth Sleepiness Scale [ESS], and PSQI Daytime Dysfunction subscale).
CPAP adherence was operationally defined as the number of hours that CPAP was used at the prescribed pressure per day (ie, over a defined 24-hour period). The sources of CPAP data were the CPAP manufacturer websites (AirView and EncoreAnywhere). Because these 2 manufacturers represent approximately 85% of the CPAP marketplace in the United States, only participants with 1 of these CPAP devices were included in the study. In addition, the study included 2 measures of CPAP efficacy: mask leak, which was defined as the amount of air that escaped (liters per minute) and the AHI, which is a measure of the number of apneas and hypopneas per hour of CPAP use. These metrics were available to the participants via the portal, and the coaches used these metrics to provide feedback on how well CPAP was working to control OSA. Suggestions to improve CPAP use were based on these metrics. For example, if mask leak was moderate to high, suggestions to improve the mask fit were discussed. These 2 metrics were intended for interventional purposes, but there were no plans to analyze them, given that CPAP adherence was the study's primary outcome.
In our qualitative work with the community of patients with overlap syndrome, we discovered that the most important outcome to patients is daytime functioning. The FOSQ measures impact of sleepiness on activities of daily living (ADLs).34,35 The FOSQ-10 consists of 10 questions on a scale of 1 to 4 (1 = extreme difficulty, 4 = no difficulty). A lower score indicates more difficulty with ADLs due to lack of sleep. The FOSQ total score is the mean of subscale scores (ie, vigilance, productivity, social outcome, intimacy, activity) multiplied by 5. The scores range from 5 (maximum difficulty) to 20 (no difficulty). Change in FOSQ total score is calculated from baseline to end point, with higher (positive) values representing improvement. The worst possible change value would be −15 and the best would be +15.
Clinical measures are important but may not reflect the day-to-day functioning and well-being of patients with chronic diseases. The PROMIS initiative of the National Institutes of Health was developed to advance the methodology and application of PROs among patients with chronic diseases for use in research and clinical practice.36,37 The study used 2 related PROMIS 8-item sleep scales (sleep-related impairment and sleep disturbance), as well as the following additional PROMIS measures: global health (2-item measure); physical functioning (4-item measure); ability to participate in social roles and activities (4-item measure); anxiety (4-item measure); depression (4-item measure); pain interference and intensity (4-item measure); and cognitive functioning (4-item measure). All PROMIS measures are scored in the following way: (1) sum the total (follow instructions for a prorated score if any items are missing for a measure) and (2) translate the total score (or prorated score) to a T-score per PROMIS instructions. A T-score is a standardized score with a mean of 50 and SD of 10. PROMIS scores are interpreted with higher scores representing more of the concept being measured.
The PSQI is a self-rated, 19-item questionnaire used to assess sleep quality and disturbances over the previous 1 month.38 The PSQI measures 7 areas of sleep: (1) subjective sleep quality, (2) sleep latency, (3) sleep duration, (4) habitual sleep efficiency, (5) sleep disturbances, (6) use of sleep medication, and (7) daytime dysfunction. Items are scored on a Likert scale, with 0 being indicative of better sleep and the maximum value of 3 being indicative of poor sleep. PSQI scores can range from 0 to 21, with higher scores indicating worse sleep quality. The PSQI total score was used in our study unless otherwise specified.
The CAT is a simple, 8-item health status instrument for patients with COPD, which is highly practical,39 has good psychometric properties, and has been shown to be responsive to pulmonary rehabilitation and recovery from exacerbation.40-43 CAT scores range from 0 to 40, with higher scores representing a more severe impact of COPD on a patient's life. The minimally important clinical difference score has been shown to be 2 points.44,45 The CAT includes a sleep item and an energy item, which is relevant to those patients with overlap syndrome.
The ESS is an 8-item validated measure of daytime sleepiness and is the most widely used subjective measure of excessive daytime sleepiness in research and clinical settings.46,47 The questions on the ESS ask respondents to estimate how likely they are to doze in a variety of different situations, with 0 meaning they would never doze and 3 meaning they would have a high chance of dozing. The range of ESS scores is 0 to 24, with higher scores indicating a higher level of sleepiness. The ESS can be used to discriminate the sleepiness level of patients with OSA from that of healthy controls.48
The Functional Comorbidity Index (FCI) is a validated measure of comorbidity with functional level as the outcome of interest.49 The FCI is composed of a list of 18 comorbid medical conditions that the study respondents self-reported having or not having. The conditions are simply summed, such that a higher number represents higher comorbidity.
The sociodemographic information we collected included age, sex, race, ethnicity, sexual orientation, and income. Additional participant characteristics included smoking status, geographic location, years since diagnosis, and comorbidities.
Oxygen therapy adherence was assessed by self-report. Several items asked about whether oxygen therapy was administered, as well as type and timing of oxygen therapy.
Participant satisfaction was assessed by self-report for each communication with the study staff (coach, RT, other), by method (phone or online). Participants were asked to provide a rating based on a scale of 1 (dissatisfied) to 10 (satisfied).
The power analysis was based on the primary hypothesis that CPAP adherence (ie, the number of hours that CPAP was used in a 24-hour period) would be improved in the PC group in the first 6 weeks compared with the RC group. A sensitivity analysis was conducted by considering a range of sample sizes from 100 to 180 participants per group. Assuming a 2-sided type I error of α =.05, we could detect a standardized effect size (for the difference in CPAP adherence between the PC and RC groups) ranging from 0.296 to 0.398 with 80% power. These calculations indicate we would have sufficient power to detect a small to medium standardized effect size (0.325) in adherence between the PC and RC groups if enrolling 150 participants per group. By definition, standardized effect sizes are unitless and their advantage is that the size of the effect can be compared across studies.
Study recruitment started in January 2018 and ended in July 2019. Data collection occurred at baseline, 6 weeks, and 12 weeks.
The study team collected all questionnaire/survey data electronically or by phone, using the study's Coordinator and Participant portals. Both portals contained questionnaires for all users to complete that would become available according to a previously established time-sensitive workflow that started with a Self-Report Eligibility questionnaire, completed by the participant in the Participant portal after signing consent. Completion of the Self-Report Eligibility questionnaire would then trigger a consequent form for the coordinator to complete in the Coordinator portal (eg, the CoE questionnaire, Demographics, Medical History).
The randomly assigned participants then completed 3 main questionnaires that were available to them via the Participant portal. Those time-sensitive questionnaires were the baseline, 6-week follow-up, and 12-week follow-up surveys. Each time point had a 2-week window during which the questionnaire was available online to the participant. Participants were offered an incentive of a $25 online gift card that was emailed to them on completion of each survey. Reminder phone calls were made and email reminders sent to those participants who had a survey due, to inform them of the approaching follow-up window due date. Any surveys that were not completed by their due date were considered missing.
CPAP data were included in the study in 2 ways. First, a data workflow integration was established such that data calls were made 2 times each week (on Monday and Wednesday) to populate the O2VERLAP study portal. These data were used by both participants and interventionists to monitor progress and intervene as necessary. An intermediary, Corepoint Health (Frisco, TX), was contracted to provide data integration services and provide middleware between the CPAP manufacturer servers and the O2VERLAP study portal. More than 90% of the data were transmitted successfully. Second, to use a comprehensive and accurate CPAP adherence and efficacy data set, our research team engaged in a double check of every CPAP data point in our data set to be sure it was consistent with the data at the source, namely, the manufacturer's servers. Given the unexpected and novel findings of the very high CPAP use levels found in this study, we are confident in our conclusions because of our extensive CPAP data quality-assurance efforts. See Appendix D for more details.
Preliminary analyses began by examining the distribution of study variables and providing descriptive statistics (ie, mean, median, SD, quartiles for continuous variable, frequency, and percentage for categorical variables) about the study population. Patient characteristics were compared between study groups by Wilcoxon rank-sum tests (or Fisher exact tests, as appropriate). Variables on which the groups differ initially were explored as covariates in subsequent analyses. The primary analyses were intent to treat (including all enrolled participants), and all analyses were performed using 2-sided tests with α = .05. Mean differences and 95% CIs were reported with the P values. Summary metrics were reported by mean (SD), unless otherwise specified. Analyses were conducted using R statistical software.50 Analysis plans that addressed each hypothesis are described in the following sections.
We hypothesized that CPAP adherence at the 6-week time point would be improved in the PC group compared with the RC group. A random-effects model was used to compare the mean CPAP adherence over 6 weeks between the PC and RC groups. Daily CPAP adherence data were used for analysis. A random intercept was included in the model to account for the correlation between repeated measures of CPAP adherence over each assessment period as well as the correlation among 3 assessment periods (baseline, week 6 during the intervention, and 6 weeks after the intervention [week 12]) within each patient. A multivariable random-effects model was used to assess the difference in CPAP adherence between the PC and RC groups, with adjustment for potential covariates. Adjustments were made to correct for baseline imbalances across study groups and to adjust for variables known to influence the outcome. Baseline demographics and other clinically important characteristics were assessed for imbalance among the study groups, using Wilcoxon rank-sum test, chi-square, or Fisher exact test, and their association with the outcome was assessed using a simple random-effects model. These variables were included as covariates in the multivariable model if found to be moderately associated with the outcome or unbalanced (P < .15) across groups. All covariates significant at P < .10 were kept in the final model.
We hypothesized that improvement in patient-centered outcomes at 6 weeks and 12 weeks would be larger in the PC group than in the RC group. The change in patient-centered outcomes from baseline to week 6 and week 12 was compared between the PC and RC groups. The difference in change of each outcome was assessed using analysis of covariance (ANCOVA), with intervention group as a main effect and baseline score as a covariate. A linear random-effects model was fit to assess the change from baseline to week 6 and the change from baseline to week 12 by considering the correlation of measurements (at baseline, week 6, and week 12) within each participant. A multivariable linear random-effects model was used to assess the difference in change scores between groups, with adjustment for baseline characteristics using approaches similar to those described for the primary aim.
The PRO measurements are divided into 3 categories: daytime functioning, sleep quality, and daytime symptoms. Daytime functioning was deemed the most important PRO per our focus groups. Daytime functioning was measured by the FOSQ. Sleep quality was measured by the PSQI. Daytime symptoms were measured by the ESS.
We conducted additional exploratory analyses that were unanticipated at study outset: examination of CPAP data use levels. CPAP adherence data, measured in duration of use per night, is most meaningful when compared relative to total sleep time (TST) and/or total sleep period (TSP). Because we did not have an objective measure of TST in this study, we opted to use TSP from the PSQI. Descriptive statistics and Pearson correlation coefficient were used.
To evaluate the heterogeneity of treatment effects, we assessed the interaction between treatment group and several baseline patient characteristics including age, sex, education, and socioeconomic status. If an interaction was significant, the treatment effect was estimated separately for each study subgroup.
We included all available data for the analyses using a random-effects model. The mixed-effects method allows the inclusion of participants with missing data or those who were terminated early in the study, without relying on data imputation procedures. We also performed ANCOVA, which only included participants with complete data to compare the change score with adjustment for baseline score. Because the results from the random-effects model and ANCOVA were consistent, we only provide results of the former.
There were 4 key changes to the original study protocol: (1) medical device adherence (ie, oxygen therapy), (2) focus on CPAP adherence, (3) measure selection, and (4) study time frame.
In the original study protocol, we planned to study adherence to the 2 medical interventions that are prescribed for patients with both COPD and OSA: oxygen therapy and CPAP, respectively. It became clear during our qualitative focus-group work that less than half of patients with overlap syndrome used oxygen therapy at night, whereas all of them reported using CPAP therapy. Because of this observation, we changed the study protocol to primarily focus on the use of CPAP therapy, with a limited focus on oxygen therapy, given the relatively low base rates of its use found in this specific patient group.
The original study protocol acknowledged the potential difficulty of enrolling patients with OSA using CPAP therapy and being able to obtain their CPAP data. When a study takes place within a single health care system, typically the patients with OSA in that health care system only use a few HME provider companies. Even though most CPAP devices are owned by the patient, the HME companies are considered the CPAP data stewards, those who play an oversight or data governance role. Practically, this means that for us to obtain access to the CPAP data for our study, we needed the permission of both the patient and the HME company. Because the study was designed as a national study that was enrolling existing CPAP users, we could not limit the potential number of HME companies with whom we would have to work to obtain the CPAP data.
For this reason, the original study protocol was designed to use an innovative device (Evermind, Nashville, TN) designed to obtain a proxy measure of CPAP use. This small device plugs into the wall, and the power cord of the CPAP device plugs into it. The measurement concept was simple: whenever the CPAP blower drew electrical power, the device would measure the power surge. When power was being used, so was the CPAP. In this way, we could easily obtain a proxy measure of CPAP use for any CPAP make or model, regardless of age or interconnectivity of device.
However, once the study became solely focused on CPAP adherence, the study team decided, based on feedback from our patient community, that the project should use the most rigorous CPAP data (ie, the data obtained by the CPAP device itself), which were sent to the manufacturers' servers and made available to the research team. The study team decided to use the most commonly used CPAP devices (ResMed and Philips Respironics). More details on CPAP data sharing methods are provided in the Discussion section.
The qualitative focus groups were instrumental in helping the study team finalize its measure selection. The focus groups determined that by far the most important outcome to patients was daytime functioning. Originally, daytime functioning was considered a tertiary outcome, but its importance was elevated to the most important secondary measure (CPAP adherence remained the primary outcome) because of this important work in listening to the patient community.
The only change that was made to the original milestones was a 2-month, no-cost extension request in May 2019 to give us more time to recruit and reach our 100% recruitment milestone. For the project to meet our original date of May 31, 2019, for the 100% recruitment milestone, the study needed to have each of our planned campaigns perform well. Although we were close to meeting our milestone, in the end several campaigns either did not do as well as they had in the past (eg, social media posts on Facebook); were unable to be implemented (namely, PRIDEnet; Health eHeart Alliance; and 2 UC pSCANNER sites: UCLA and UCSF); or resulted in fewer-than-expected patients with COPD and OSA (UC Irvine). The no-cost extension allowed the main research period to be changed from May 31 to July 31, 2019, and consequently pushed back each subsequent milestone by 2 months.
An O2VERLAP study campaign was defined as sending a message via a defined method to a defined audience. Table 4 shows the 4 primary communities used in this study; the recruitment method type for each; the actual or estimated size of each audience; and the total number and percentage of participants who were enrolled in the study from each community and audience.
O2VERLAP Study Campaigns: Number and Percentage of Participants Enrolled by Campaign.
The only audience for which we could not determine a total size was the “Web browsing and word of mouth” audience, because it was difficult to measure accurately. At study outset, we knew that we needed to carefully track the yield of our campaigns because our project was staffed to accommodate approximately 20 to 30 new participants per month in combination with the current participants who needed to be followed each month as well. This meant that our campaigns needed to meet a range of new participants each month, without going too far above or below the required number. The more we understood the projected yield of a campaign, the better we could plan for which type of campaign and when it should begin. So, for each campaign that allowed it, we counted the number of individuals for each campaign method. For the word-of-mouth and web-browsing audiences, we could not accurately keep track of how many people engaged in each. For this reason, audience size is missing for these 2 campaign types.
Table 5 provides a summary of the study home-page web analytics results that were a function of the study campaigns. Table 5 is a summary of the full table in Appendix E. It also provides a summary of the analytics by community. The term “sessions” here is defined as a period during which a user is actively engaged or viewing the website or post. Note the highest number of sessions for the COPD and OSA communities occurred with 2 campaigns that were based on “boosted” posts (ie, paying for additional posts). The 2035 sessions resulting from the highest-yielding COPD campaign were more than double that of the next highest-yielding COPD campaign. Likewise, the 13 193 sessions resulting from the highest-yielding OSA campaign were >13 times more than the next highest-yielding OSA campaign. Appendix E lists the number of trackable URLs that were assigned to each campaign. This number is nearly equivalent to the number of times that a campaign was promoted. We point this out because the number of sessions appears to be primarily a function of the size of the community and the number of times a campaign was promoted. We believe an additional factor may have come into play, namely, that the campaign text and images (eg, see Figure 2) primarily emphasized sleep apnea and CPAP, which would have appealed to the OSA community more than to the other communities.
O2VERLAP Study Home-Page Web Analytics.
Figure 3 shows the study CONSORT diagram. All study recruitment efforts in aggregate resulted in 1315 individuals registering for the O2VERLAP study on the study home page. Of those who registered, 657 individuals (50%) proceeded to sign consent, whereas 658 did not. Of the 657 who consented, a total of 541 participants (82%) completed the CoE phone call, and 116 (18%) were either not reached or were reached but decided they did not want to proceed with the study.
O2VERLAP Study CONSORT Diagram.
The O2VERLAP study goal was to enroll 330 participants and randomly assign 300 of them, which factored in a 10% prerandomization attrition rate. On July 31, 2019, the study reached its enrollment goal with 2 additional participants when 332 (61%) of the 541 who completed the CoE phone call were found to be eligible to participate. The remaining 209 (39%) were deemed ineligible for the study. Table 6 provides a breakdown of the reasons for ineligibility and then categorizes them by diagnosis, device, and miscellaneous reasons. The diagnosis category had the highest number (n = 134), followed by device (n = 69) and miscellaneous (6).
Number of Persons Ineligible to Participate and Reason for Study Ineligibility (n = 209).
“Did not have a COPD diagnosis” was the most common reason for ineligibility because of 2 related factors: (1) the high volume of the ASAA campaign's study promotional efforts to their OSA community, and (2) OSA is more common in people diagnosed with COPD than the reverse.
Figure 4 and Figure 5 show study enrollment over the 18-month recruitment period. Figure 4 shows the cumulative enrollment, which appears to be relatively smooth, with participant enrollment ahead of goal for the first half of the study and then slightly behind the goal. Figure 5 shows the monthly enrollment figures and demonstrates the monthly variability of enrollment that the project experienced due to the relatively unknown campaign yields. The study team needed to estimate campaign yields to attempt to enroll the goal of 20 to 30 new participants per month. In other words, the team tried to avoid over- or under-enrolling in any 1 month, given considerations about coach staffing and time effort.
Cumulative Enrollment Over 18-Month Recruitment Period (February 2018-July 2019).
Monthly Enrollment Over 18-Month Recruitment Period (February 2018-July 2019).
Once participants were considered eligible and enrolled in the study, we required 1 additional step before they could be randomly assigned: getting permission for CPAP data sharing. We successfully obtained permission to share the CPAP data of 310 (93%) of 332 enrolled participants. We were concerned that data sharing would be 1 of the most significant challenges to the study. In the end, only 22 participants (7%) withheld permission for CPAP data sharing. Of the 22 who did not share CPAP data, 7 were on hold or never provided their device serial number or HME name and contact information; 6 withdrew during this time; 6 had data transmission issues; 2 were nonresponsive; and 1 died during this time. Two reasons we think participants withdrew or declined at this point were that (1) for some of these individuals, it was likely the result of the data sharing process taking a long time, and the initial interest in participating in the study naturally waning over time; and (2) they changed their mind about participating and wanted to leave the study before being randomly assigned.
To successfully obtain CPAP data sharing permissions from CPAP providers, the study team had to work with each participant's HME provider. At study outset, 1 concern with using this methodology was that potentially we would have to work with 330 different HME companies. However, in the end, we found that we “only” had to work with 132 different HMEs. We found that our study had multiple participants in some of the large national HMEs. Ultimately, the study had >10 participants in 5 HMEs, 3 to 10 participants in 9 HMEs, 2 participants in 11 HMEs, and 1 participant in 94 different HME companies. Importantly, those 5 HMEs with which >10 of the study participants were associated accounted in total for 151 participants, or 49% of the total. Those 5 HMEs were as follows: O2VERLAP HME account (n = 69), which included the participants who were not associated with an HME at the time of enrollment; Lincare (Clearwater, FL; n = 37); Apria Healthcare (Lake Forest, CA; n = 28); VHA (Washington, DC; n = 14); and Sleep Data, Inc (San Diego, CA; n = 11). The study team was able to establish efficient working relationships with these large national HME organizations, without which the study may not have been able to achieve its recruitment goals. One of those HME accounts was not associated with an existing HME company, which we discuss next.
The CPAP data sharing process would have caused this study to fail had it not been for the study team's creation of its own HME account for participants who were no longer being followed by an HME company or who had opted out of being followed by an HME. This O2VERLAP study HME account included 69 participants. In addition, the support of the ASAA CPAP Assistance Program was instrumental in helping some participants receive updated CPAP devices with wireless capabilities. The 69 participants represented 22% of the 310 from whom we successfully obtained CPAP data. Given the study's ambitious timeline in needing the full 18 months for study recruitment, it was clear that without doing this, the study would have been unable to meet its objectives. The Discussion section provides more coverage of this issue, with an emphasis on the challenges that patients face in obtaining their own medical data.
Of the 310 participants who reached the CPAP data sharing point of the study workflow, there were an additional 16 participants who did not move forward to the randomization phase for 2 reasons: they either “declined” to move forward with the study (n = 9) or the study team was unable to contact them despite multiple attempts by phone and email to reach them for randomization (n = 7). We put the word “decline” in quotation marks because some told us they no longer wished to participate in the study. Others simply said they were no longer interested in the study and did not elaborate.
A total of 294 (89%) of the 332 enrolled participants were randomly assigned to 1 of the 2 intervention groups: 153 were randomly assigned to the PC group and 141 to the RC group. In comparing the 38 participants (who signed informed consent but were not randomly assigned) with the 294 participants (who met all study criteria and were randomly assigned), there were no significant differences in any demographic characteristic.
The main study was conducted on a sample size of 294 participants. This section summarizes the findings by category: sample characteristics, primary and secondary aims, and exploratory analyses.
Women and men comprised 47.3% and 52.7% of the sample, respectively. The mean age of the sample was 64.0 (SD, 9.6) years and ranged from 41 to 89 years. Table 7 shows the sample characteristics.
Sample Characteristics.
Figure 6 provides a map of the United States color-coded by region (West, Midwest, South, Northeast). The map is based on US Census regions and divisions.51 The map shows the geographic distribution of the O2VERLAP study sample size of 294 participants. The study included 3 participants from Canada with dual citizenship who had residences in both the United States and Canada but who were living in Canada during the time of the study. The study IRB advised that this was allowed per its policies. The study team attempted to ensure geographic diversity during the study.
Geographic Distribution of O2VERLAP Study Participants by US Region and Canada.
Appendix F provides 2 supplemental data tables that show the decision-making by the team at the 11-month time point and then the final geographic distribution. The final rates for each of the 4 regions, from highest to lowest, were as follows (in number of participants per 10 million): West, 12.6; Midwest, 9.9; Northeast, 6.7; and South, 6.3. Rates were calculated by dividing the number of participants enrolled in a region by the population of that region.
Table 8 provides the number (percentage) of participants with the lengths of time since diagnosis for both COPD and OSA. Note that 54% and 58% of the sample was diagnosed with OSA and COPD ≥6 years ago, respectively. Relatively few participants were diagnosed <2 years ago (19% OSA and 10% COPD).
Years Since COPD and OSA Diagnoses (N = 293).
The FCI was used; the mean number of medical conditions was 6.4 (SD, 2.7; range, 2-17). When the optional write-in medical conditions were included, the mean was 8.4 (SD, 2.9; range, 2-17). The top 5 endorsed medical comorbidities in this sample were visual impairment (ie, cataracts, glaucoma): 182 (54.8%); obesity (body mass index [BMI] ≥ 30): 180 (54.2%); arthritis: 177 (53.3%); peripheral vascular disease: 168 (50.6%); and upper gastrointestinal disease: 148 (44.6%).
The study design resulted in 100% of study participants having and using CPAP therapy. We also found that 44% (n = 144) of all participants were using oxygen therapy to some degree, meaning that 56% (n = 184) were not using oxygen therapy.
In terms of smoking status, 233 participants (70.2%) reported being past smokers, 27 (8.1%) current smokers, 71 (21.4%) never having smoked, and 1 (0.3%) refused to answer. The past smokers reported smoking for 31.4 years (SD, 12.3; range, 1-60) and 9.7 packs/week (SD, 5.4; range, 1-30). The current smokers reported smoking for 35.1 years (SD, 12.3; range, 15-59) and 5.6 packs/week (SD, 4.5; range, 1-24).
Participants in both groups were presented with an automated online satisfaction survey within the Participant portal after a communication task with an RT and/or Information Line coach was completed. The questions on the satisfaction survey were as follows: (1) Who did you have a study communication with? Response options: Information Line coach, RT, or other. (2) Was your communication by phone or portal messaging? Response options: video, phone, or online (chat). (3) Communication Satisfaction score (1-10 scale). The purpose of the last question was to rate the participant's satisfaction regarding their perceptions of the quality of communications with either an RT or a COPD Information line coach. For the PC group, the satisfaction survey also appeared after communication via a chat function in the study portal. Table 9 provides a summary of the satisfaction surveys completed and scores based on a range from 1 to 10, with higher scores indicating greater satisfaction.
Satisfaction Survey Scores for Coach's Communications.
The primary aim of the study was to examine the effect of the intervention (ie, PC or RC) on CPAP adherence. Table 10 provides the adherence values by group and time point. Time point refers to the 3 assessment time points (baseline, 6 weeks, and 12 weeks). The groups differed at baseline, with the RC group (7.3 hours/night) using CPAP slightly more than the PC group (6.1 hours/night; P < .001) during the 30 days before study start.
CPAP Adherence by Group and Study Time Point.
In an unadjusted linear random-effects model, the interaction between time point and intervention group was not significant, which indicated no significant difference in change of CPAP adherence between the 2 study groups in either week 6 (difference = 0.18; 95% CI, −0.16 to 0.52; P = .29) or week 12 (difference = −0.05; 95% CI, −0.39 to 0.29; P = 0.78). Removing the interaction term from the model, we found that overall, the week-12 CPAP adherence level was significantly lower than at baseline (difference = −0.17; 95% CI, −0.34 to −0.002; P = .047) while controlling for the group, and the PC group had lower CPAP adherence compared with the RC group while controlling for the time point (difference = −1.16; 95% CI, −1.75 to −0.58; P < .001).
CPAP adherence was significantly related (P < .15) to race, ethnicity, income, education, and smoking status, but not related to age, sex, or marital status. Adding the identified covariates to a multivariable model resulted in similar findings as found in the analysis that was not adjusted for these additional baseline characteristics.
The secondary aim of the study was to examine the relationships between group assignment and the following PROs: daytime functioning, sleep quality, and daytime symptoms. Table 11 provides a summary of these measures by group and time point.
PROs by Group and Study Time Point.
Baseline scores on the FOSQ 10-item tool (FOSQ-10) did not differ between the 2 groups (P = 0.16), with a mean score of 14.8 for the RC group and 14.1 for the PC group. In an unadjusted linear random-effects model, the interaction between time point and intervention group was not significant, which indicated no significant difference in change in FOSQ-10 score between the 2 study groups in either week 6 (difference = 0.12; 95% CI, −0.53 to 0.77; P = .72) or week 12 (difference = 0.16; 95% CI, −0.51 to 0.83; P = .64). Removing interaction from the model, we found that the week-6 FOSQ-10 score was marginally significantly higher than at baseline (difference = 0.32; 95% CI, −0.01 to 0.64; P = .06) while controlling for the group, and the PC group had a marginally significantly lower FOSQ-10 score compared with the RC group while controlling for the time point (difference = −0.64; 95% CI, −1.39 to 0.12; P < .10). The results from a multivariable random-effects model with adjustment for baseline covariates were similar to those of the unadjusted analysis.
Baseline scores on the PSQI were significantly different between the 2 groups (P = .01), with a mean score of 8.1 for the RC group and 9.4 for the PC group. In an unadjusted linear random-effects model, the interaction between time point and intervention group was not significant, which indicated no significant difference in change in PSQI score between the 2 study groups in either week 6 (difference = −0.26; 95% CI, −0.93 to 0.42; P = .46) or week 12 (difference = −0.07; 95% CI, −0.77 to 0.63; P =.85). Removing interaction from the model, we found that the week-12 PSQI score was significantly lower than at baseline (difference = −0.59; 95% CI, −0.94 to −0.24; P =.001) while controlling for the group, and the PC group had a significantly higher PSQI score compared with the RC group while controlling for the time point (difference = 1.19; 95% CI, 0.29-2.09; P < .01). The results from a multivariable random-effects model with adjustment for baseline covariates were similar to those of the unadjusted analysis.
Baseline scores on the ESS were not significantly different between the 2 groups (P = .16), with a mean score of 8.5 for the RC group and 9.5 for the PC group. In an unadjusted linear random-effects model, the interaction between time point and intervention group was not significant, which indicated no significant difference in change in ESS score between the 2 study groups in either week 6 (difference = −0.06; 95% CI, −0.84 to 0.73; P = .89) or week 12 (difference = −0.15; 95% CI, −0.96 to 0.66; P = .72). Removing interaction from the model, we found that the week-12 ESS score was significantly lower than at baseline (difference = −0.66; 95% CI, −1.06 to −0.25; P = .002) while controlling for the group, and the PC group had a marginally significantly higher ESS score than that of the RC group while controlling for the time point (difference = 0.92; 95% CI, −0.14 to 1.98; P = .09). The results from a multivariable random-effects model with adjustment for baseline covariates were similar to those of the unadjusted analysis.
CPAP is prescribed for use during sleep, and nearly all patients use it for some portion of their TSP. Patients seldom use CPAP for longer than their TSP. However, per anecdotal reports, some patients may use CPAP during nonsleep periods because they like how it helps with their breathing.
TSP was calculated as uptime minus bedtime, and its units are in hours. The source of uptime and bedtime data was the PSQI. Table 12 provides the TSP by group and time point. Note that the average TSP for the entire group at each time point was quite high, at 8.1 hours/night, and that it ranged quite substantially from 2 hours on the low side to 14 hours on the high side.
TSP by Group and Study Time Point.
Figure 7 shows a time graph of CPAP use over the course of an approximately 90-day period for 1 nonidentified participant. The green bars indicate the times when CPAP was used during each 24-hour period. Breaks in the green bar indicate when the CPAP mask was removed. The single red bar on March 11 indicates a day when CPAP was not used. The blue box indicates when a normal, approximately 8-hour TSP typically occurs (ie, 10 pm to 6 am). The green bars outside of the blue box show those times when CPAP was used outside of the normal sleep period.
Time Graph of CPAP Use by Day.
The percentage of CPAP use during TST was calculated using the following ratio: CPAP use (hours) divided by TSP (hours), which we here refer to as the CPAP to TSP ratio. A CPAP to TSP ratio of 1.0 means that a CPAP user who slept 6 hours used CPAP for the full 6 hours. A ratio of 2.0 means that a CPAP user who slept 6 hours used CPAP for 12 hours. The mean CPAP to TSP ratio at baseline was 83% (SD, 33%; range, 0%-223%) and at 12 weeks was 87% (SD, 32%; range, 0%-297%). Table 13 provides the CPAP to TSP ratio by group and time point.
CPAP to TSP Ratio by Group and Time Point.
Our SAB suggested there might be a relationship between severity of COPD and CPAP use. The team examined the relationship between COPD severity as measured by the CAT score and CPAP adherence. For the entire group, there was a nonsignificant relationship between the CAT score and CPAP adherence at the 12-week time point. However, when the subgroup of high CPAP users (defined as CPAP to TSP ratio >1.0) was analyzed separately, the correlation coefficient was 0.250 (P = .04). Figure 8 shows the scatterplot for this subgroup.
Scatterplot of CAT Score by CPAP Adherence.
In the O2VERLAP study, we did not find a difference between the 2 intervention groups (PC and RC) in CPAP adherence or PROs. The baseline CPAP adherence level for the entire sample was 6.7 hours/night (SD, 2.8; range, 0-17.3 hours/night), and the mean CPAP to TSP ratio at baseline was 83% (SD, 33%; range, 0%-223%). This very high CPAP level at baseline (ie, before the start of the intervention) represents the goal of interventional studies (see Appendix A) and is therefore considered an unexpected and novel finding of this study. Possible reasons for the high baseline CPAP use level may be due, in part, to the shortened study time frame and use of electronic recruitment methods that resulted in a sample of very consistent CPAP users at baseline. High adherence rate at baseline appears to have resulted in a ceiling effect, meaning that there was little room for improvement for both groups.
In addition, the baseline difference between the 2 groups (RC: 7.3 hours/night; PC: 6.1 hours/night) was also a surprising finding. Most interventional studies show an effect of increasing CPAP use levels by 1.0 to 1.5 hours/night, but those studies are of new CPAP users, not existing users, as in our study.32 The baseline use level represented a ceiling above which improvement can be difficult. To our knowledge, this high level of CPAP use in patients diagnosed with COPD and OSA is a finding that has not been previously reported in the medical literature.
Several reasons may account for the uniquely high CPAP use level found in this study relative to past studies reported in the literature. Appendix A provides the weighted mean adherence level for the overall group (3.48 hours/night across 18 studies and 4000 participants) and for US-based studies (3.11 hours/night) as reference values to put into context the adherence levels found in the present study. The mean age group in this study was of retirement age, so the participants likely were free in the daytime hours to use CPAP. It appeared that more use, especially for daytime CPAP users, was associated with worse COPD severity. It may be that other published studies did not have samples with the degree of COPD severity that our sample had. Finally, the present study was unique in its national, electronic recruitment method conducted via the COPD and OSA communities. It may be that patients who are actively involved in monitoring social media channels and who are willing to respond to research opportunities are in some ways different than those who are not.
The study as designed can help answer the question, Should a clinic or patient advocacy organization be more proactive in setting up online and personnel support for their communities? In the end, because we found patients who were already using CPAP at a high level, the findings of the present study cannot help answer this contextual question. In fact, because there was a downward trend in the PC group at 12 weeks (after an initial slight increase), it may be that providing structured support to active, consistent users has a slightly negative effect. The recommendation from the study team to a clinic or organization that is considering staffing up or providing minimal resources would be to do the latter and build up only if the demand can be quantified. This is not to say that support should not be provided. Support should continue to be provided, just on an as-needed basis based on documentation of poor adherence. Good care for chronic illness is providing the right support at the right time to the right person.52
In addition to carrying out this large, national CER study, the study team learned a great deal about (1) the pros and cons of using electronic methods (primarily direct emails and social media posts) to carry out study promotional efforts; and (2) the complex issue of the health care system providing the sharing of medical device data to patients, specifically with CPAP devices. We discuss these 2 areas next.
We learned several important lessons concerning conducting primarily electronic recruitment in a large-scale, national study. We organize our findings by lessons learned about social media expertise, messaging, working with PCORnet partners, and using social media platforms.
First and foremost, the earlier a project can bring a social media expert on board, the better. Our team developed the experience and skill in social media campaigns, but it would have been far more effective from study outset had we brought this person onboard at the start. They could have helped write the initial email/social media post text and headers and engaged in A/B (or similar) testing to determine which text and headers performed the best. A/B testing refers to creating 2 separate versions of a social media post (“A” vs “B”) to evaluate which post is more effective. A social media expert could have helped identify and design graphics that were catchy and met the specific requirements for research on Facebook and other social media outlets. They also could have helped us understand the fine details of social media posts and their measurement. For a defined research network such as PCORnet, it is highly recommended that the central coordinating personnel search for and identify a highly qualified individual who is available on a recharge basis for individual projects. Alternatively, identifying a group of qualified individuals to hire as part-time staff or consultants is another possibility.
Many lessons were learned regarding message content and timing. As described, our team primarily used 3 different types of messaging: (1) initial general content; (2) text geared toward the individual recipient, but with an additional request for that individual to promote the study to their family and friends; and (3) “last chance,” which communicated an end-of-project urgency. In addition to the content of the messaging, the timing is important as well. We found that mid-week campaigns (eg, Tuesday-Thursday) tended to work better than weekend campaigns (eg, Friday-Monday) because of the potential for getting overlooked in weekend activities. Finally, 18 (51%) of 35 campaigns were only implemented once. Given that we found objective evidence for the continued very good yield of implementing ≥2 campaigns to a single audience, this in hindsight appears to be a lost opportunity. The lesson learned here is, if possible, to screen potential study-promotion partners for their willingness to engage in >1 outreach to their community. The benefit to the study is the possibility of needing to work with fewer partners to achieve study goals.
We first discuss this topic generally and then divide it into PCORnet partners and non-PCORnet partners. Generally, the key point we found was to be respectful of the existing communication methods and practices of a partner organization. We ended up creating an internal framework for our calls, such that each call (1) first discussed the current community communication methods of the organization (eg, timing, methods, size, inclusion of research); (2) assessed interest in research in general and O2VERLAP specifically; and (3) then discussed the possibility of the partner engaging in O2VERLAP communications and how that might look. We would then provide an email summary of the call and schedule a follow-up call to finalize details.
Our study team found that the willingness of PCORnet partners to help with study recruitment was in no small part related to the PCORI funding cycle. All the PCORnet CDRNs and PPRNs were nearing the end of the funding cycle when the O2VERLAP project had <6 months to go. In addition, future funding from PCORI to remain a PCORnet member was in doubt because the People-Centered Research Foundation was established, particularly for the PPRNs. We strategically decided to recruit from the COPD and OSA communities first, given the higher likelihood of having both medical conditions, which was a prerequisite for study participation. Despite initial enthusiasm and verbal agreement from many PCORnet partners, when it came time to engage in study recruitment messaging, many either opted not to participate or to send a single communication. In no small part because of this issue, our study team had to request a no-cost extension to the project to send additional emails to UCSD and UC Irvine, to achieve our recruitment milestone.
The study team found a variety of issues with non-PCORnet partners. These types of partners were quite diverse. We attempted to go beyond the medical nonprofit organizations and health care systems that were part of the broad network by going to the medical device suppliers themselves. This included a wide of range durable medical equipment or HME providers, from umbrella or membership organizations, to large national companies, to smaller regional companies, and to online-only companies. We asked them if they would consider communicating with their customers to promote a research study. Although we received a fair amount of interest in communicating with patients who used their services, in the end we found that very few had formal, planned communications with their customers, and that for those that did, the HME providers were either unwilling or reluctant to help with the study. In the end, we do not believe that O2VERLAP was promoted by an HME company via a single campaign. However, we did find 2 HME companies that were willing to post about O2VERLAP. We also found an online CPAP HME company that was willing to post about the study in their online community. Finally, we identified 1 online forum geared toward medical conditions that asked for $25,000 for a single post, despite language in their Terms of Conditions that promised their community that 1 of their stated goals was to inform them about research opportunities (ie, they would inform their community about research opportunities but at a very high cost, one that is prohibitive for all but industry research).
For this study, we used Facebook a great deal, with ultimately very poor returns. The primary issue with Facebook is that the fate of a post directed to a defined community cannot be tracked. For example, if an HME organization has a Facebook community with 20 000 individuals, Facebook was unwilling to tell us whether that post would reach all 20 000 individuals or a subset. Based on our analytics, it appeared that it reached a subset of that defined group. We therefore engaged in “paid boosts” to increase coverage. After using a couple of metrics to define the demographic we were interested in (but not having the ability to target people with known or suspected COPD or OSA), we found that paid boosts were a very poor use of limited study resources. Only if a study is recruiting a sample that is consistent with the metrics that Facebook allows for boosting would we recommend consideration of a paid boost. Finally, Facebook defines an impression as “the number of times an instance of an ad [or post] is on screen.”53 For our study, the activity was a study-related Facebook post. Importantly regarding the Facebook definition, there is no guarantee that an impression was actually read or not. That said, we did include a trackable URL in the Facebook post, so if an individual clicked on the link to the O2VERLAP study home page, we kept track of those clicks objectively. What we could not measure was the number of times the post was read (whether in full or in part) but not clicked.
Twitter, on the other hand, collects a fee only when a Tweet is clicked. From this perspective, paying for clicks on Twitter makes far more sense than paying for a boost that may or may not be seen on Facebook. That said, we found the Twitter audience to be significantly smaller, and it did not appear to include the demographic the study required. In further comparing Facebook and Twitter, individuals on Facebook appear to be socialized to read a wider variety of content than those who use Twitter, where most users appear to be interested in very specific kinds of content. Participation in research appears to be very rare on Twitter.
In the previous sections of this report, we provided a relatively basic background on CPAP data sharing methods and results. In this section, we provide more details on the specific strategies developed and deployed based on collaboration between the study team and the SAB that allowed the study team to move from a data sharing rate of 25% to our final sharing rate of 93%. Five specific strategies were used:
Given that the use of these strategies took several months to develop and deploy, we recommend that researchers who are trying to engage in participant medical-device data sharing take a multipronged approach to foster the sharing of such data for research:
CPAP data sharing was covered in the Methods and Results sections from the perspective of the study and in the Engagement section from the perspective of the SAB. In terms of CPAP data sharing from the perspective of a patient, patients have the right to access their medical information and data. The VHA has determined that patients own their CPAP devices and that they have a right to access their data.54 In the private sector, HME companies act as data stewards; that is, patients do not have direct and easy access to the data on their own medical devices. Instead, they must formally request permission to access their data. The study team found that this access via a third party ranges from being easy to very difficult.
The present study had limitations. The study had an ambitious time frame and depended on electronic recruitment methods that were conducted primarily through the COPD and OSA communities and PCORnet partners. The recruitment findings showed that the most success was with participants who had already been research participants and who were known to have both COPD and OSA. It may be that, in the end, these participants were more active in taking care of themselves, as demonstrated by their willingness to be part of existing patient communities and by the high CPAP adherence levels they demonstrated. Perhaps the most important finding of this study is also its most significant shortcoming. In terms of subpopulations, it appeared that in the subgroup of the most active CPAP users, greater COPD severity was associated with more use. It seems that there is a previously unidentified subgroup of patients with a high rate of CPAP use who are using CPAP during the daytime to help with their breathing. It should also be mentioned that the sample was predominately White, non-Hispanic, and well educated; therefore, research in other populations is warranted to determine generalizability of study findings or whether findings would be different in other populations.
There are several important recommendations for research based on the findings of the O2VERLAP study. From a clinical perspective, future researchers should better understand the factors associated with daytime use of CPAP in patients with overlap syndrome and whether there is a physiological benefit that supports the perceived benefit. The present study was unique in that it is 1 of the few that examined current CPAP users. Most CPAP adherence studies examine new users (ie, naive to CPAP). The interventional approaches used in this study should be performed with new users. Providers and clinics are often unsure how much support to provide to new users and such a study could help provide that information.
We have several recommendations with regard to methodological findings from this study for future research. The first concerns PCORI's contracting and negotiation process, which allows a wonderful opportunity to make significant improvements to a study. With respect to O2VERLAP, in hindsight, significantly more attention should have been paid to the changes to the study that increased the scientific rigor but also the time and effort to support that additional rigor. Closely related to this issue is the use of social media for study recruitment. Methodological studies are needed to learn how best to use social media platforms cost-effectively for research study recruitment.
The following quotes from O2VERLAP study participants were just 2 of many that showed appreciation for having a resource available to supplement the care they were receiving from the professional medical team.
I had a good chat with the overlap people today. The young lady I spoke with was very polite to me and helped me set some goals as well as some good information. (PIN 20024)
Thank you for the kindness, but it was truly my pleasure. And I learned much from the course to my benefit…. These studies and trials are the way to beat this and it would be my honor to contribute. (PIN 20152)
The O2VERLAP study did not find a difference between the intervention groups (PC and RC) in CPAP adherence or outcomes, which appears to be due in part to the shortened study time frame and use of electronic recruitment methods that resulted in very high baseline CPAP use levels in this sample of patients diagnosed with both COPD and OSA. This remarkably high level of CPAP use is a finding that has not been previously reported in the medical literature. In addition to carrying out a large, national CER study, the study team learned a great deal about (1) the advantages and disadvantages of carrying out research within PCORnet; (2) the value that an SAB brings to a large trial; (3) the pros and cons of using electronic methods (primarily direct emails and social media posts) to carry out study promotional efforts; and (4) the complex issue of the health care system providing the sharing of medical-device data with patients, specifically those who use CPAP devices.
PPRNs
CDRN
Research reported in this report was funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (PPRND-1507-31666). Further information available at: https://www.pcori.org/research-results/2016/comparing-two-methods-improve-cpap-use-among-patients-copd-and-obstructive
CPAP Adherence Interventional Studies Summary Table (PDF, 103K)
Focus Group Manuscript (PDF, 339K)
O2VERLAP Study Promotion Efforts (PDF, 225K)
Table 1. Summary of O2VERLAP Study Promotion (PDF, 166K)
Remote CPAP Adherence Monitoring Setup Process and Data Integrity Check (PDF, 186K)
Google Analytics for Trackable URLs for Each Study Promotion Campaign (PDF, 155K)
Geographic Distribution (PDF, 135K)
Table 3. O2VERLAP Study Enrollment by State (PDF, 102K)
Abbreviations/Acronyms (PDF, 68K)
Stepnowsky C, Malanga E, Martinez S, et al. (2021). Comparing Two Methods to Improve CPAP Use among Patients with COPD and Obstructive Sleep Apnea — The O2VERLAP Study. Patient-Centered Outcomes Research Institute (PCORI). http://doi.org/10.25302/09.2021.PPRND.150731666
The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.
This book is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License which permits noncommercial use and distribution provided the original author(s) and source are credited. (See https://creativecommons.org/licenses/by-nc-nd/4.0/
Your browsing activity is empty.
Activity recording is turned off.
See more...
