An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Jerry Krishnan, MD, PhD, Richard Casaburi, MD, PhD, Scott Cerreta, RRT, Kristen Holm, PhD, Janos Porszasz, MD, PhD, Valentin Prieto-Centurion, MD, Robert Sandhaus, MD, PhD, and Jamie Sullivan, MPH.
Author Information and AffiliationsUnderuse and overuse of long-term oxygen therapy (LTOT) at home is common in patients with chronic obstructive pulmonary disease (COPD). The need for evidence-based interventions to promote appropriate use of LTOT in this population was identified as a critical knowledge gap in multi-stakeholder COPD workshops.
(1) Engage patients with COPD prescribed home oxygen and their caregivers to identify gaps in knowledge, self-management skills, and outcomes of importance to patients. (2) Pilot test study procedures for the PELICAN trial. (3) Evaluate the comparative effectiveness of proactive vs reactive PELICAN interventions vs usual care on adherence to oxygen (primary outcome) and multiple patient-centered outcomes in a multicenter pragmatic clinical trial of patients with COPD prescribed LTOT 24 hours per day, 7 days per week. (4) Examine heterogeneity of treatment effects of proactive and reactive PELICAN interventions in patient subgroups. (5) Understand barriers and facilitators of successfully implementing the PELICAN proactive intervention.
Patients with COPD prescribed LTOT were randomly allocated 1:1:1 to proactive coaching (patient-directed educational materials and phone-based peer coaching delivered during 5 sessions over 60 days); reactive coaching (same educational materials, with support from peer coaches offered only in response to inbound calls by participants); or usual care (more limited set of patient-directed educational materials). Research coordinators masked to intervention assignment conducted follow-up visits by phone over the course of 90 days. The primary outcome was adherence to LTOT, defined as mean use of ≥17.7 hours per day over the 0- to 60-day interval (yes vs no), after accounting for potential confounders in multivariable logistic regression models. We calculated LTOT use using oxygen concentrator meter readings and number of compressed oxygen tanks used, as recorded by participants on worksheets developed for the study. The mean level of LTOT use in participants who had improved survival in the previous clinical trial of LTOT was 17.7 hours per day; we therefore used this cut point to define adherence to LTOT in the current study. Adherence to LTOT in the reactive vs usual care and in proactive vs usual care groups over 60 days (0-60 days after randomization) were prespecified as coprimary comparisons (using a 2-sided P < .025 and a 97.5% CI to identify a significant difference for each comparison). Secondary analyses employed a 2-sided P < .05 and a 95% CI to identify significant differences and included comparisons of adherence to LTOT at other time intervals (0-30, 30-60, and 60-90 days) and LTOT use as a continuous outcome. Secondary outcomes included Patient-Reported Outcomes Measurement Information System measures of physical, emotional, and social health, and patient-reported acute care utilization.
Of 444 participants (142 usual care, 148 reactive, 154 proactive), the proportion adherent to LTOT (mean use ≥17.7 hours/day) over the 0- to 60-day interval was 74% in the usual care, 84% in the reactive coaching, and 70% in the proactive coaching groups. Reactive coaching (adjusted odds ratio [OR] for adherence to LTOT vs usual care: 1.77; 97.5% CI, 0.80-3.90; P = .10) and proactive coaching (adjusted OR for adherence to LTOT vs usual care = 0.70; 97.5% CI, 0.34-1.46; P = .28) groups did not significantly differ in the odds of adherence to LTOT in the 0- to 60-day interval compared with the usual care group. We observed similar findings when we compared LTOT use as a continuous outcome (secondary analyses). However, proactive coaching significantly reduced the odds of adherence to LTOT compared with the reactive coaching (adjusted OR for adherence to LTOT = 0.40; 95% CI, 0.20-0.79; P = .008; secondary analyses). In addition, proactive coaching significantly reduced depressive symptoms compared with both usual care (adjusted difference in T score: −3.2; 95% CI, −5.1 to −1.3; P < .01; secondary outcome) and with reactive coaching (adjusted difference in T score: −2.5; 95% CI, −4.4 to −0.6; P < .01; secondary outcome), and reduced sleep disturbance compared with usual care (adjusted difference in T score: −2.0; 95% CI, −3.9 to −0.1; P = .04; secondary outcome) in the 0- to 60-day interval. We did not find a significant difference in acute care utilization among the study groups.
Proactive or reactive peer coaching did not significantly improve adherence to LTOT compared with usual care. The potential for proactive peer coaching as a strategy to promote improvements in depressive symptoms and sleep disturbance requires further study.
Missing oxygen use data (27% of participants) and the risk of nonsystematic and systematic reporting errors for oxygen use are potential limitations. We did not observe significant evidence of heterogeneity of treatment effects on oxygen use.
Chronic obstructive pulmonary disease (COPD) is a chronic respiratory disorder that affects about 24 million adults in the United States and is responsible for 750 000 hospitalizations per year.1,2 In 2010, COPD surpassed stroke to become the third leading cause of death in the United States.3 The annual total cost of COPD is approximately $50 billion in the United States alone.1,4 In recognition of the substantial and increasing public health impact of COPD, the US Centers for Disease Control and Prevention; the National Heart, Lung, and Blood Institute; and others are collaborating to increase awareness, understanding, and use of evidence-based preventative and treatment approaches for COPD.2,3
Patients with sufficiently severe COPD to require long-term oxygen therapy (LTOT) are at higher risk of poor outcomes, including hospitalizations and death.5 In patients with COPD and an oxygen saturation of 88% or lower at rest (severe resting room air hypoxemia), clinical trials indicate that LTOT improves survival.6 Use of LTOT with a mean of 17.7 hours per day (ie, 73.8% of the day) is associated with a survival benefit among patients with COPD.7 Inadequate understanding about the risks and benefits of home oxygen therapy, confidence about the appropriate use of various types of home oxygen equipment, and a perceived stigma of using supplemental oxygen in public contribute to an underuse of prescribed oxygen, with various studies suggesting adherence rates of 45% to 70%.8
The need for evidence-based interventions to promote appropriate use of LTOT has been identified as a critical knowledge gap in multi-stakeholder workshops.9,10 Peer coaches, individuals who provide education and coaching based on their experiences to those who are affected by the same condition, offer a potential solution. Emerging evidence supports the use of peer coaches to provide self-management education for people with diabetes.11 Additionally, a pre-post intervention study of phone-based coaching by trained peers in patients with α1-antitrypsin deficiency-associated COPD demonstrated improvements in self-management (including LTOT adherence) and health outcomes (COPD exacerbations, acute care health encounters).12 However, it is not clear whether such results can be generalized to a broader population, including those with COPD. Moreover, results of studies examining telephone-based coaching have not been uniformly effective. For example, telephone-based coaching did not improve the likelihood of smoking abstinence, adherence to medications, functional status, or quality of life post–hospital discharge in patients with acute coronary syndrome.13
The goal of the PELICAN (Peer-Led O2 InfoLine for Patients and Caregivers) study was therefore to design, implement, and evaluate the effectiveness of a peer coaching intervention on oxygen use in patients with COPD prescribed oxygen 24 hours per day. The PELICAN study addressed the patient-centered question “How can clinicians and the care delivery systems they work in help me make the best decisions about my health and health care?” by directly comparing 2 forms of peer-led phone-based coaching services against usual care to support appropriate use of LTOT. Based on the input of patients, caregivers, clinicians, and other stakeholders, we then conducted a pragmatic clinical trial to evaluate the effectiveness of a telephone-based proactive peer coaching intervention (a series of scheduled coaching calls to deliver an oxygen curriculum over a 60-day period) vs reactive peer coaching (assistance by peer coaches in response to inbound calls by study participants) vs usual care. The primary outcome was LTOT adherence, and secondary endpoints included patient-reported measures of physical, mental, and social health, and acute care utilization.
The PELICAN study included the following specific aims:
We intended the results of the PELICAN study to inform the design and implementation of care models by patient advocacy groups, health systems, and payers (including the Centers for Medicare & Medicaid Services) seeking to improve the care and outcomes of a high-risk population with COPD.
The multidisciplinary PELICAN stakeholders included 16 patient advocacy groups and patients with COPD and their caregivers; 7 professional societies and clinicians who care for patients with COPD; 21 researchers; 15 representatives from the industry, including durable medical equipment (DME) and other medical supply companies; 32 hospitals and health systems; and 4 universities that train clinicians who manage COPD.
Throughout the study, we emphasized the importance of input from all stakeholders. We gave the information obtained from the patient and caregiver focus groups and interviews priority in defining the most important curriculum topics to be included in the intervention and outcomes to be measured in the study. We relied on the input from a multidisciplinary external advisory committee (EAC) to ensure that the study design addressed their expressed needs regarding the target population, importance of outcomes, and practicality of the intervention.
We used “snowball” recruitment to identify additional stakeholders and build new partnerships. Study investigators had a history of collaborating with COPD stakeholders (eg, COPD Foundation). Early in the study planning stage, we elicited recommendations from existing stakeholder partners to identify additional partners.
We employed 3 formal methods of stakeholder engagement during the study: (1) in-person focus groups for patients and caregivers in English and Spanish, (2) conference with the multidisciplinary EAC on a yearly basis, (3) partnering with stakeholders to deliver the intervention (COPD Foundation), and (4) including 2 patient research advocates (patients with COPD who were not study participants) as members of our staff to inform study operations form a patient perspective.
We developed the research question in collaboration with our stakeholders, especially patients and their caregivers.14
Patient and caregiver input was critical in selecting the topics for the intervention curriculum. For example, patients underscored the importance of discussing the known benefits of oxygen in the first session, as they often focused more on negative feelings and beliefs about LTOT. Additional feedback suggested the need to encourage patients to regularly follow up with their providers to reassess their need for LTOT, and instructions about how to use pulse oximeters that are sometimes prescribed for patients. Patient and caregiver input was critical in decisions about the length and frequency of sessions, and the practicality of implementation by phone. We used patient, caregiver, clinician, and other stakeholder preferences in defining the primary and secondary outcomes. Patients and caregivers also served as peer coaches, who agreed to deliver the intervention, building on the existing COPD Foundation InfoLine. Stakeholders in the EAC also helped shape the study eligibility criteria and recruitment/retention plan.
Participation of the COPD Foundation enhanced the rigor and quality of the study, as the foundation had already developed a peer-to-peer coaching program that could support clinical trials (eg, structured training program, recorded calls to permit training and quality control processes). The stakeholders supported the need for a clinical trial design that included 2 active comparators (reactive coaching and proactive coaching).
The independent EAC and data and safety monitoring board (DSMB) regularly reviewed study procedures and updates.
Our stakeholders, including patient advocacy groups and professional organizations, have expressed interest in disseminating study findings to various stakeholder channels (including payers). Stakeholders including the COPD Foundation, the American Association for Respiratory Care, and the Respiratory Health Association have already posted on their websites information indicating forthcoming results. The American Association for Respiratory Care has also conducted webinars with respiratory therapists, a key target end-user for the PELICAN project. Depending on study findings, stakeholder participation could be broadened to promote adoption into practice.
We conducted patient and caregiver interviews as part of aim 1 to inform the design of the clinical trial proposed in aim 3. The methods and findings of this study have previously been described in a peer-reviewed manuscript.14 In brief, we conducted in-person focus groups and individual discussions via telephone and email with 25 patients with COPD and 5 caregivers to elicit feedback to refine the intervention and clarify outcomes of highest importance to patients. The study was a 3-arm pragmatic clinical trial: Proactive PELICAN, Reactive PELICAN, and Usual care (Figure 1). To generate evidence about the real-world effectiveness of peer coaching, the study design features (1) inclusive eligibility criteria and recruitment from a national sample of patients, (2) minimal interactions between research staff and participants to mimic the real world, and (3) interventions that could easily be implemented in clinical practice. In the PELICAN study, the interventions are delivered by patients and caregivers who together constitute the COPD InfoLine staff, rather than researchers. The COPD InfoLine, which existed before the PELICAN study, is a free service available to any patients with COPD or their caregivers who call the toll-free number.15 They are peers in the sense that they also have COPD or serve as caregivers of patients with COPD. All peer coaches undergo a rigorous 45-hour training process on customer service and call etiquette, service to sales, Health Insurance Portability and Accountability Act compliance, COPD disease management information, and COPD Foundation program information.16
PELICAN Study.
We employed block-stratified randomization to promote balance in the number of participants in each of the 3 PELICAN groups within the following strata: months since last hospitalization, duration of LTOT, and DME provider. We selected DME provider as a randomization stratum because the support and education patients receive while on LTOT may differ across providers. Staff who enrolled participants and collected outcome data were masked to the allocation sequence.
Patients were enrolled into the main trial if they met all of the inclusion criteria: (1) age ≥18 years; (2) English as primary language; (3) patient-reported physician diagnosis of COPD; and (4) reported a physician prescription for LTOT 24 hours per day, 7 days per week. We employed the following exclusion criteria: (1) patient declines to provide consent, (2) patient declines to receive at least 1 study phone call after enrollment, and (3) patient reports life expectancy <6 months or reports enrollment in a palliative care/hospice program. We did not require confirmation of the diagnosis of COPD by spirometry as this would be unlikely to be required by the program if implemented in routine practice. Following informed consent, participants were asked to identify their primary caregiver, who was also offered enrollment into the study.
With assistance from our network of stakeholders who were part of the research team or members of the multidisciplinary EAC (eg, COPD Foundation, Pulmonary Hypertension Association, Coalition for Pulmonary Fibrosis, American Association for Respiratory Care, Drive Medical-DeVilbiss Healthcare, patients with COPD and their caregivers), we identified multiple IRB-approved sources of patients to support recruitment goals (IRB #2014-0385). Engagement with stakeholders early in the study design phase allowed us to benefit from their networks to raise awareness of the PELICAN study and to refer potentially eligible patients to contact the study call center and learn more about the study via a toll-free phone number: (1) Clinicians (eg, physicians, respiratory therapists) throughout the United States referred patients to contact the study call center; (2) DME providers throughout the United States referred patients to contact the study call center; (3) the study team used research registries at the University of Illinois at Chicago developed by the study PI to identify, recruit, and enroll eligible patients; and (4) information about the PELICAN study was posted on various online venues (eg, Facebook, COPD Foundation website, DME provider websites) to invite interested patients to contact the study call center. Our use of a phone-based recruitment strategy (combination of inbound calls to our study call center from patients interested in learning more about the PELICAN study and outbound calls to individual patients who were part of our existing research registries) constituted a pragmatic approach to identifying interested patients.
On the initial phone call, the call center staff interviewed patients using a standardized, IRB-approved phone script to identify eligible patients, obtain informed consent via telephone, and collect contact and baseline information (baseline visit A). Study participants were then mailed a copy of the consent document and other PELICAN study materials (welcome/education packet #1; see Table 1). The process and documentation of informed consent was reviewed and approved by the University of Illinois IRB. See the flow diagram of patients through eligibility assessment, enrollment, intervention allocation, follow-up, and data analyses in Figure 2.
Self-management Educational Materials.
CONSORT Diagram for the PELICAN Trial, n = 444 Randomized.
The peer coaches performed the calls to deliver the intervention from their homes using the COPD InfoLine telephone system. The peer coaches lived in homes located in 10 different states across the country; thus, the PELICAN study was a multicenter trial. Our stakeholder-supported approach to recruitment allowed the PELICAN call center, located in Chicago, Illinois, to recruit and engage patients in all 50 states.
Patients and their caregivers were randomly allocated to 1 of 3 groups during the randomization/baseline visit B, as described in the “Aim 1: Engage Patients With COPD Prescribed LTOT and Their Caregivers to Identify Gaps in Knowledge, Self-Management Skills, and Outcomes of Importance to Patients” section.
Following randomization, the PELICAN study call center transferred patients in the proactive PELICAN group to a COPD Foundation InfoLine peer coach (“warm hand-off”) located in 1 of 10 states to receive their first of 5 specially designed telephone-based peer coaching sessions. These sessions were delivered over a 60-day period and built on written education packets that we adapted from self-management materials developed by the COPD Foundation's medical and scientific experts (Table 1).16 All educational materials were written at the sixth- to eighth-grade level. The PELICAN welcome/education packet #1, mailed after the baseline A visit, included information about the PELICAN study, the toll-free COPD InfoLine phone number, hospital-to-home care transitions, home oxygen equipment, and how to recognize COPD exacerbations. Two additional education packets provided information about how to personalize LTOT use with a pulse oximeter, recognizing and treating exacerbations, pulmonary rehabilitation, and the role of caregivers (education packet #2, mailed after the randomization/baseline B visit), and other information (education packet #3, mailed at end of the 90-day follow-up period).
Peer coaching was delivered via the COPD InfoLine by individuals who lived in 10 different states. The peer coaches who delivered the proactive PELICAN curriculum (to 7 patients with COPD prescribed LTOT and 3 caregivers of patients with COPD prescribed LTOT) underwent study-specific training review the materials in education packets #1 and #2 with study participants during 5 phone call sessions scheduled over 60 days to promote self-management skills regarding the use of LTOT, as well to address patient questions and concerns (Table 2). We designed peer-coaching sessions to be one-on-one, unless the participant's caregiver was available to jointly participate. Proactive participants were also encouraged to contact the COPD InfoLine peer coaches as needed for additional assistance and information.
Proactive PELICAN Peer-Coaching Curriculum, by Phone Session.
Following randomization, the PELICAN study call center reviewed the contact information for the COPD InfoLine peer coaches included in education packet #1 and encouraged participants to contact peer coaches for assistance as needed. However, the call center did not provide a warm hand-off to peer coaches. As in the proactive group, education packet #2 was mailed to the reactive group after the randomization/baseline B visit. Packet #3 was mailed to participants at the end of the study.
Following randomization, the PELICAN call center confirmed the contact information for the study participants. Although participants in the usual care group were mailed the welcome/education packet #1 at the same time as the participants in the proactive and reactive PELICAN groups (after baseline A visit), they were mailed both education packet #2 and education packet #3 at the end of the study.
PELICAN call center research staff collected baseline and follow-up assessments via phone at the randomization/baseline visit B, and then at 1 month (window: 37-43 days after baseline visit), 2 months (67-73 days) and 3 months (90-97 days; Table 3). The call center was masked to intervention group. In a sample of 44 participants who lived in Chicago, Illinois, we also conducted home visits and attached devices that could identify pressure fluctuations in the nasal cannula consistent with respiration (Breath Trackers; Reference LLC, Elkader, IA) using stationary oxygen concentrators (Table 4) to evaluate the accuracy of the measured adherence to LTOT using patient-reported information.17
Outcomes Assessment Schedule.
Oxygen Delivery Equipment Reported by Patients in the PELICAN Study.
PELICAN is intended to test strategies to promote adherence to patients' LTOT prescription (primary outcome), since LTOT underuse is common and mean LTOT use of at least 17.7 hours per day (73.8% of 24-hour day) increases survival in patients.7 We did not rely on participant-reported LTOT use for this study in order to minimize the risk of overestimating adherence; instead, we asked participants for information about their oxygen equipment and calculated LTOT use based on this information. During the randomization/baseline visit B, participants (and, if available, caregivers) were extensively trained to record information about their oxygen equipment using a worksheet (written at sixth- to ninth-grade reading levels) developed by study staff, and to report this information to the PELICAN call center during the outcome assessment calls that occurred at 30, 60, and 90 days after randomization.
We instructed participants to report the stationary and portable oxygen concentrator meter readings during outcome assessment calls. We used these meter readings to calculate the total number of hours of use of the equipment over the time interval (eg, 60 days for the 0- to 60-day interval), including the day the reading was recorded. We calculated the mean number of hours of use for each oxygen concentrator by dividing the total hours of use by the total number of days in the interval between the outcomes assessment calls. If oxygen concentrators were replaced between outcome assessment calls, participants were instructed to record the final meter reading and date of reading for the old equipment as well as the initial meter reading and date of reading for the replacement oxygen concentrator.
Participants with portable compressed oxygen tanks (which lack usage meters) were trained to record in worksheets the type and number of tanks emptied and their initial pressure, and to report this number during outcome assessment calls, along with any changes to their prescribed oxygen flow rates. We calculated the hours per day of oxygen delivery based on the initial pressure, the tank type, and the prescribed flow rate. For those participants who used transfill systems to fill portable tanks, the meter readings indicating the number of hours in which tanks were being filled was recorded. We ascertained the rate of tank filling for the transfill device (liters per minute) for the transfill device model. These data, as well as the patient's prescription for ambulatory oxygen flow rate, allowed us to calculate the total number of hours of transfill oxygen use in a given recording interval. For participants who used conserver devices with their tank oxygen, we calculated a device-specific conserver ratio and multiplied the total tank oxygen use in a given interval by this ratio. We obtained the conserver ratio from literature as well as data from manufacturers on their conserver characteristics.18 As in the case of oxygen concentrators, we calculated the mean portable compressed oxygen tank use per day by dividing the total hours of use by the number of days in the interval between the calls.
We combined the mean use (hours per day) for each piece of oxygen delivery equipment to calculate the overall mean LTOT use in the interval from the previous outcome assessment call. Data for calculating adherence to LTOT was considered as able to be evaluated for a given interval (eg, 0-60 days) only if meter readings for all pieces of oxygen equipment were available and the data corresponded to at least 80% of the days of an interval (eg, the data provided by the participant included at least 48 days of the 0- to 60-day interval). Otherwise, we recorded oxygen adherence as missing for that interval. The analyses prespecified measuring the primary outcome over the 0- to 60-day interval, as it coincides with the end of the coaching intervention in the proactive group. We intended to supplement this information with the data from DME providers that supplied the oxygen delivery equipment for each patient (eg, concentrator readings, number and types of tanks delivered), thereby linking patient-reported data with that recorded by the DME. However, as described in the Methods, Conduct of Study section, DME providers were unable to provide these data to the study team. We used the Patient-Reported Outcomes Measurement Information System (PROMIS) to assess patient-reported physical function (v1.0, SFA4a), fatigue (v1.0 SF4a), sleep disturbance (v1.0 SF4a), depression (v1.0 SF4a), anxiety (v1.0 SF4a), ability to participate in social roles and activities (v2.0 SF4a), and satisfaction with social roles and activities (v2.0 SF4a) as secondary outcomes.14,19 PROMIS scores are standardized such that the national population has a mean T score of 50 with an SD of 10. Studies suggest that a 2- to 5-unit or greater difference in T score may be clinically important, though minimum important differences specific to patients with COPD have not been established.20-23 Caregiver outcomes consisted of the PROMIS ability to participate in social roles and activities (v2.0 SF4a), and satisfaction with social roles and activities (v2.0 SF4a) instruments. We also collected patient-reported acute health care utilization (ie, all-cause emergency department visits, hospitalizations) during follow-up calls. The reliance on patient-reported measures of health status and health care utilization mimics what often occurs in clinical practice; even when clinician offices include an electronic health record, these records are not usually integrated with the electronic health records of health systems in the same or other regions.
We designed data collection processes to minimize participant burden to facilitate recruitment and retention. We accommodated individual preferences as much as possible when scheduling calls; used a relatively short (90-day) follow-up period for the study; minimized data collection to information necessary to answer the study questions; completed follow-up activities by phone, rather than requiring patients to travel for in-person study visits; permitted participants to complete follow-up visits, even if they discontinued the interventions; and provided a small reimbursement to participants for their time ($25 per participant per completed assessment call). We attempted to contact participants by phone a maximum of 3 times to complete data collection, and then recorded reasons for the inability to complete data collection (eg, lost to follow-up, withdrew consent). We also convened staff meetings every 1 to 2 weeks to review study progress and troubleshoot difficulty with study activities, including data collection or management.
We performed analyses in 3 phases. Phase 1 consisted of exploratory analyses to identify, correct, and confirm values or missing data, and provide descriptive statistics (ie, frequency [proportions], mean [SDs]). We also employed Bland-Altman plots to evaluate the agreement (mean difference; 95% limits of agreement) between patient-reported LTOT use data collected by the PELICAN call center vs the Breath Tracker data.24
Phase 2 focused on bivariate analyses to compare primary and secondary outcomes by treatment group. We used t tests, Wilcoxon rank sum tests, chi-square tests, and Fisher exact tests, as appropriate, for pairwise comparisons of reactive coaching vs usual care, proactive vs usual care, and proactive vs reactive coaching.
Phase 3 included multivariable logistic or linear regression models, as appropriate, to account for potential confounders for the primary and secondary outcomes. We prespecified the primary analysis as the results of the multivariable logistic regression models comparing adherence in days 0 to 60 in the reactive vs usual care, and proactive vs usual care treatment groups, after adjusting for potential confounders (ie, adjusted odds ratios [ORs]). We predefined potential confounders as baseline characteristics that were associated with home oxygen use in patients with COPD in previous studies or differed between the 3 study groups at the baseline visit: age, gender, marital status, use of a portable oxygen concentrator, patient-reported physician diagnoses of anxiety or depression, all-cause hospitalizations in the previous 12 months, and all-cause ED visits in the previous 12 months.8,25 To minimize the risk of bias in specifying potential confounders, we prespecified potential confounders for the multivariable analyses while being masked to treatment group when examining baseline characteristics.
We also conducted exploratory analyses to assess the potential for heterogeneity of treatment effects for the primary outcome by examining the consistency of the adjusted ORs in prespecified subgroups (across levels of each baseline characteristic) in a series of logistic regression models with covariates that included the treatment group indicator, the treatment by subgroup interaction terms, and all other covariates in the multivariable models. We determined the P values for consistency of adjusted ORs across subgroups by Wald chi-square tests, as described in a previous study.26 The PELICAN study was not specifically powered to assess heterogeneity of treatment effects, so all such analyses should be considered exploratory and provided so that they could be used in meta-analyses of multiple studies.
We conducted all analyses using a modified intention-to-treat principle in which we ignored missing data and data from participants with liquid oxygen (as we were unable to collect their LTOT use data; see Methods, Conduct of Study section). We employed a 2-sided α of .025 (and 97.5% CIs) for the 2 pair-wise analyses of the primary outcome in the multivariable analyses (LTOT adherence in reactive vs usual care; proactive vs usual care at 0-60 days), and a 2-sided α of .05 for all other hypothesis tests. We selected the first 60 days (0- to 60-day interval) for conducting the primary assessments, since this period coincided with the end of the peer coaching intervention. Secondary intervals for assessments included 0 to 30 days, 30 to 60 days, and 60 to 90 days (end of the follow-up period). In a sensitivity analysis for the 0- to 60-day interval, we compared adjusted differences in LTOT use as a continuous outcome.
We also considered various approaches for handling missing data, including multiple imputation strategies.27 However, results of imputation strategies may present spurious results when missing data are not missing at random. In studies targeting adherence, analyses are more challenging since missing data may be related to patient characteristics or the underlying and unobserved degree of adherence.28 We elected, therefore, to follow the approach for missing data in a recently published study of LTOT by reporting the frequency of missing data for all analyses26 and comparing baseline characteristics of those with vs without evaluable data. For the primary outcome, we supplemented these analyses with a “best-case” and “worst-case” sensitivity analyses in which we replaced all missing adherence data as adherent and as nonadherent, respectively.
Power: Assuming 450 study participants (150 per group), 10% loss to evaluable data, and adherence of 45% to 70% in the usual care group,8 we estimated at least 90% power (with 2-sided α = .025) for a minimum detectable difference of 18% to 24% in the proportion adherent to LTOT in each of 2 coprimary comparisons (reactive coaching vs usual care; proactive coaching vs usual care).
We originally proposed enrolling patients newly initiated on LTOT immediately following a hospitalization for a COPD exacerbation. These participants were to be identified by a national DME provider that had agreed to collaborate in this study; the DME provider also had initially agreed to provide the study investigators data regarding oxygen equipment they obtained during home visits (which would be used to calculate oxygen adherence). However, the pilot study indicated that such an approach would be infeasible because the DME provider who was part of the research team indicated they would no longer be able to assist with recruitment activities or data about oxygen utilization. We had viewed a commitment from a leading DME company as a strength of the study, but our experience suggests stakeholder engagement in research studies can be complicated by financial decisions that contribute to how industry partners make decisions about continued study participation.29-31 Following discussions with our EAC, DSMB, and the Patient-Centered Outcomes Research Institute (PCORI) program officer, we elected to modify the study protocol to broaden the study population and include any patient with COPD prescribed LTOT 24 hours per day, 7 days per week. Therefore, with IRB approval, we modified the study eligibility criteria to recruit patients with COPD in the ambulatory setting, whether or not they had recently started on LTOT. Our stakeholders helped identify multiple sources of potential study participants and supported the revised study design that targeted a broader COPD population. To minimize the opportunity for confounding, our stakeholders recommended stratifying the randomization according to the duration of LTOT, time since last hospitalization, and the DME provider that delivered the supplemental oxygen delivery equipment.
Due to the withdrawal of our principal DME partner, we were also unable to obtain information about the home oxygen delivery equipment, including a copy of physician prescriptions for LTOT. As a result, we relied exclusively on patient-reported data about oxygen delivery equipment to calculate LTOT adherence and were unable to calculate adherence to liquid oxygen equipment. Calculating adherence in patients using liquid oxygen requires the weight of the liquid oxygen equipment at the time of delivery as well as subsequent time points (eg, 30, 60, and 90 days after randomization). Obtaining the weight of the liquid oxygen equipment from the participants was deemed to be impractical and to potentially pose a safety concern. We therefore were unable to include 63 (12% of 507) participants who reported using liquid oxygen equipment. Compared with participants who had not used liquid oxygen, liquid oxygen users were older, were more likely to have education beyond high school, used oxygen at a higher flow rate, had been on LTOT for a longer period of time, were less likely to use a stationary concentrator or compressed oxygen tanks, had less anxiety and depressive symptoms, and had greater ability to participate in social roles and activities (Table 5). Patients on liquid oxygen appeared to represent a different COPD population and were therefore excluded in the analyses examining the primary and secondary outcomes in this report. The analyses we present therefore represent a modified intention to treat.
Baseline Characteristics of Patients, Stratified by Use of Liquid Oxygen.
With IRB approval, we amended the study protocol to conduct home visits among participants recruited at 1 site (Chicago, Illinois) to assess the validity of patient-reported data regarding meter readings on stationary concentrators and portable concentrators, as compared with the Breath Tracker data. We plan to develop secondary manuscripts for peer-reviewed journals that describe (1) differences between patients using liquid oxygen vs other home oxygen equipment; (2) effects of proactive, reactive, and usual care on outcomes other than adherence to oxygen (secondary outcomes in this report) among the 63 patients using liquid oxygen; and (3) unanticipated challenges when engaging for-profit stakeholders in patient-centered research.
Eleven individuals participated in focus groups (Table 5), 8 individuals participated in interviews, and 11 COPD InfoLine staff provided feedback via email.14 Feedback from patients and caregivers helped the research team refine multiple aspects of the study (Table 6). With the exception of sexual function, patients and caregivers endorsed the importance of including multiple patient-reported outcomes in the study (see the Methods section).
Focus Group Participants.
Stakeholder Feedback and Revisions to the Study Design (Specific Aim 1).
Eight hospitalized patients were screened, of which 1 was ineligible. Seven patients (and 1 caregiver) were enrolled in the study, far fewer than the 20 patients we had originally proposed to enroll. Six participants completed the study and 1 participant died before the 90-day outcome assessment. The principal challenge was the slow rate of enrollment during hospital-to-home transitions. Our primary DME provider was unable to support the recruitment role for the PELICAN study, which led to a redesign of the study eligibility criteria, approach to recruitment, and other study procedures (see Methods, Conduct of Study section). Inclusion of a pilot study following stakeholder engagement activities proved to be critical in finalizing the study design.
From April 2015 to May 2016, 732 patients from all 50 states were assessed for eligibility. Of these patients, 221 were ineligible and 27 declined to provide informed consent. The primary reason for ineligibility was not being prescribed LTOT 24 hours per day (89% of those ineligible). Of the 484 individuals who provided informed consent, 444 completed the baseline/randomization visit (26 were lost to follow-up, 10 withdrew consent, and 4 were terminated during the run-in period for other reasons). Of the 444 who were randomized, 142 were allocated to usual care, 148 to reactive coaching, and 154 to proactive coaching (Figure 2). Retention through the 3-month follow-up visit was high in all 3 participant groups (95%, 97%, and 94%, respectively) and in 73% of the enrolled caregivers.
Overall, the study cohort was predominantly 65 years or older, female, and non-Hispanic White; had been on LTOT for more than 12 months; used a stationary concentrator and compressed oxygen tanks; and reported a 1 SD lower physical function and ability to participate in social roles and activities compared with the US national average (Table 8;). A little more than half of the participants reported being hospitalized in the previous 12 months, and nearly all participants reported having had a visit with a health care provider in the past 12 months. The treatment groups were similar at baseline except that individuals in the proactive group were more likely to be at least 65 years old (vs usual care: 65% vs 57%) and female (vs usual care: 72% vs 62%), and were less likely to be married or cohabiting (vs reactive: 32% vs 52%) and using a portable concentrator (vs reactive: 27% vs 36%).
Baseline Characteristics by Treatment Group (n = 444).
A total of 272 (61.2% of 444) participants reported having 2 different types of home oxygen equipment, 158 (35.6%) reported having 3 different types, 12 (2.7%) reported having 4 types, and 2 (0.5%) reported having a single type. Surprisingly, 9 different combinations of oxygen equipment existed among study participants (Table 9); the combination of oxygen equipment was similar across the study groups. The most common equipment combination was a stationary concentrator and compressed oxygen tanks (210 participants, or 47%). These results have been submitted for presentation at a research conference.32
Supplemental Oxygen Equipment Combinations for the 0- to 60-Day Interval.
Combining LTOT use across the various types of equipment, we had evaluable data for calculating adherence in n = 325 (73% of 444; Table 9) participants in the 0- to 60-day interval. A higher proportion of individuals in the reactive group had evaluable adherence data for the 0- to 60-day interval, followed by the usual care and proactive groups (78%, 73%, and 68%, respectively). We had evaluable data in n = 314 participants (71% of 444 participants) in the 0- to 30-day interval, n = 319 (72% of 444 participants) in the 30- to 60-day interval, and n = 335 (76% of 444 participants) in the 60- to 90-day interval. Incomplete oxygen equipment worksheets for 1 or more pieces of equipment, errors in recording data on oxygen equipment worksheets, and not having the oxygen equipment worksheet at the time of the PELICAN follow-up call (eg, because the call was completed from a location other than home) contributed to missing LTOT adherence data. Missed PELICAN follow-up calls (approximately 5% of visits) also contributed, to a lesser extent, to missing data on LTOT adherence.
Compared with the 325 participants with evaluable LTOT adherence data for the 0- to 60-day interval, the 119 participants without evaluable data were more likely to be non-Hispanic Black or other minority race/ethnicity (P = .05), were less likely to have education beyond high school (64% vs 53%; P = .04), had more depressive symptoms at baseline (mean T score 53.7 vs 55.9; P = .02), and were twice as likely to have been hospitalized in the previous 30 days (12% vs 24%; P < .01; Table 10). Among the participants with both evaluable patient-reported data for LTOT use (reported to the PELICAN call center) and Breath Tracker data, the agreement between the 2 was very high (0- to 30-day interval κ = 0.97; 30- to 60-day interval κ = 0.90; 60- to 90-day interval κ = 0.86).
Characteristics of Patients With and Without Evaluable Data for LTOT Adherence 0 to 60 Days (Primary Outcome).
Among all 325 study participants with evaluable data for the 0- to 60-day interval, 247 (76.0%) were LTOT adherent (mean LTOT use ≥17.7 hours per day; Figure 3). Participants used stationary oxygen delivery equipment for 92% of the time they used LTOT (mean use of stationary oxygen delivery equipment of 18.3 hours per day of a total of 19.9 hours per day of mean daily LTOT use). Unexpectedly, the proactive group had significantly lower adherence compared with the reactive group (69.5% vs 83.6%; P = .01) in the 0- to 60-day interval, but the difference in adherence between either the peer coaching group or the usual care group (74.0% adherent) was not significant. The proportion of participants who were adherent to LTOT in the usual care, reactive, and proactive groups for the 0- to 30-day interval was 73.8%, 82.6%, and 72.5%, respectively. The corresponding values for LTOT adherence in the usual care, reactive, and proactive groups for the 30 to -60-day interval were 72.9%, 87.5%, and 75.2%, respectively, and for the 60- to 90-day interval were 75.9%, 82.9%, and 76.7%, respectively. In the 30- to 60-day interval, both the usual care group and the proactive group had a significantly lower adherence compared with the reactive group (P = .005 and P = .016, respectively). Although other pair-wise comparisons for the 0- to 30-, 30- to 60-, and 60- to 90-day intervals were not significant, the pattern of differences across the 3 study arms was consistent.
Adherence to Long-Term Oxygen Therapy by Study Group, 0- to 60-Day Interval.
Results of multivariable logistic regression analyses for LTOT adherence were consistent with the bivariate analyses (Table 11). After accounting for age, gender, marital status, use of a portable concentrator, patient-reported physician diagnoses of anxiety or depression, all-cause hospitalizations in the past 12 months, and all-cause ED visits in the past 12 months, we did not observe a significant difference in LTOT adherence during in the 0- to 60-day interval between usual care and either the reactive (P = .10) or proactive (P = .28) coaching groups (coprimary outcomes). However, we did observe significantly lower adherence in the proactive vs reactive group in the 0- to 60-day interval (adjusted OR = 0.40; 95% CI, 0.20-0.79; P = .008), as well as in the 30- to 60-day interval. The reactive group also had a significantly greater adherence compared with the usual care group in multivariable analyses for the 30- to 60-day interval (adjusted OR = 2.69; 95% CI, 1.32-5.47; P = .007), but not in other intervals. Results of sensitivity analyses that compared adherence to LTOT as a continuous outcome produced similar results (Table 11). We also examined the sensitivity of results to missing data by replacing all missing adherence values in the multivariable models as adherent (best case) or all missing adherence values as nonadherent (worst case); these additional sensitivity analyses produced similar findings (Table 11).
Multivariable Logistic Regression Models for LTOT Adherence (Primary Outcome).
We had evaluable data in a high proportion of participants needed to calculate a 0- to 30-day change in patient-reported measures of physical function, n = 400 (90.1% of 444 participants); fatigue, n = 416 (93.7%); anxiety, n = 418 (94.1%); sleep disturbance, n = 419 (94.4%); depression, n = 416 (93.7%); satisfaction with participation in social roles, n = 386 (86.9%); and ability to participate in social roles and activities, n = 378 (85.1%). There were similarly high proportions of participants with evaluable data for these measures in the 0- to 60-day (91.0%, 93.7%, 93.9%, 93.9%, 93.9%, 85.8%, and 86.0%, respectively) and 0- to 90-day (92.3%,92.8%, 94.1%, 94.4%, 93.9%, 86.9%, and 86.0%, respectively) intervals.
For the 0- to 30-day interval, no other significant differences in patient-reported measures of physical, emotional, or social health existed (data not shown). For the 0- to 60-day interval, bivariate analyses and multivariable linear regression models indicate a greater reduction in depressive symptoms with proactive coaching compared with usual care (adjusted difference in T scores in multivariable linear regression model vs usual care: −3.2; 95% CI, −5.1 to −1.3; P < .01), and reactive coaching (adjusted difference −2.5; 95% CI, −4.4 to −0.6; P = .01; Figure 4). The proactive group also had a significantly greater reduction in sleep disturbance compared with usual care group (adjusted difference −2.0; 95% CI, −3.9 to −0.1; P = .04; Figure 5). No other significant differences in patient-reported measures of physical, emotional, or social health existed (Figures 6-10).
Patient-Reported Depression (PROMIS) T Score Differences for the 0- to 60-Day Interval.
Patient-Reported Sleep Disturbance (PROMIS) T Score Differences for the 0- to 60-Day Interval.
Patient-Reported Physical Function (PROMIS) T Score Differences for the 0- to 60-Day Interval.
Patient-Reported Fatigue (PROMIS) T Score Differences for the 0- to 60-Day Interval.
Patient-Reported Anxiety (PROMIS) T Score Differences for the 0- to 60-Day Interval.
Patient-Reported Satisfaction With Participation in Social Roles and Activities (PROMIS) T Score Differences for the 0- to 60-Day Interval.
Patient-Reported Ability to Participate in Social Roles and Activities (PROMIS) T Score Differences for the 0- to 60-Day Interval.
For the 0- to 90-day interval, proactive coaching significantly reduced depressive symptoms compared with usual care (adjusted difference −2.4; 95% CI, −4.2 to −0.6; P = .01) and with reactive coaching (adjusted difference −2.1; 95% CI, −3.9 to −0.3; P = .02). In the 0- to 90-day interval, the proactive group also significantly reduced anxiety compared with usual care (adjusted difference −2.6; 95% CI, −4.6 to −0.7; P < .01) and with reactive coaching (adjusted difference −2.2; 95% CI, −4.1 to −0.2; P = .03). No other significant difference existed in patient-reported physical, emotional, or social health in the 0- to 90-day interval.
Missing data occurred in <5% of participants for patient-reported all-cause acute care utilization at days 30, 60, and 90. The cumulative proportion hospitalized within 30, 60, and 90 days of randomization was 9%, 16%, and 19%, respectively. We noted similar trends for emergency department visits (12%, 21%, and 27%, respectively) and all-cause hospitalization or emergency department visits (14%, 23%, and 30%, respectively). No significant differences existed between study groups in all-cause emergency department visits or hospitalizations at 60 days in bivariate analyses or in multivariable models that accounted for possible confounders (reactive vs usual care: adjusted OR = 1.09, 95% CI, 0.61-1.95, P = .76; proactive vs usual care: adjusted OR = 1.23, 95% CI, 0.69-2.20, P = .48; proactive vs reactive: adjusted OR = 1.13, 95% CI, 0.64-1.99, P = .68). Results were similar at 30 and 90 days (data not shown).
Among the 444 participants randomized, 94 participants were accompanied by their caregivers at 1 or more time points in the study. We did not find significant differences among the different study groups in the caregivers' satisfaction with participation in social roles scores (n = 71) or ability to participate in social roles and activities scores (n = 70) over the 0- to 60-day interval.
None of the prespecified baseline characteristics significantly modified the observed effects of the treatment groups on LTOT adherence in multivariable logistic regression models. The P values for Wald chi-square tests for consistency of adjusted ORs for different levels of each of the baseline characteristics in separate models were age (P = .77); gender (P = .19); race (P = .21); marital status (P = .55); highest level of education (P = .23); length of LTOT prescription (P = .46); prescribed oxygen flow rate at rest (P = .25), with activity (P = .36), and during sleep (P = .09); compressed oxygen tanks (P = .54); portable concentrator (P = .63); home transfill (P = .61); number of LTOT delivery devices (P = .90); continuous positive airway pressure or bilevel positive airway pressure (P = .12); patient-reported physician diagnosis of anxiety (P = .28), depression (P = .38), or sleep apnea (P = .07); participation of a caregiver in the study (P = .91); PROMIS depression score (P = .16); PROMIS anxiety score (P = .10); and all-cause hospitalizations or emergency department visits in the past 12 months (P = .59).
The COPD Foundation InfoLine peer coaches completed a high proportion of the 5 coaching calls: 1 (98%), 2 (95%), 3 (94%), 4(91%), and 5 (86%). At least 3 of 5 coaching calls (most sessions) were completed in 94% of participants in the proactive group. We audited recordings from a random sample of 16 peer coaching sessions, and we debriefed with 8 COPD InfoLine peer coaches and 5 PELICAN staff to identify barriers and facilitators of delivering a phone-based peer coaching intervention for home oxygen equipment. Three themes emerged as potential barriers or facilitators of phone-based peer coaching (Table 12): (1) personalized curriculum: extent to which the coaching sessions met patient needs and ability to participate; (2) understanding the study procedures: ensuring that patients understand the goals of the program, and what they should expect from participation; and (3) relationship of peer coach with: how patient participation was affected by relationship with the peer coach, personality differences, and expertise of peer coach compared with patient needs. These findings presented opportunities to improve the implementation of a peer coaching program and enhancing its effectiveness.
Barriers and Facilitators of Peer Coaching Sessions.
In this patient-centered pragmatic clinical trial in 444 patients with COPD prescribed LTOT 24 hours per day, proactive and reactive peer coaching by phone did not significantly change adherence compared with usual care in the 0- to 60-day interval (coprimary outcomes). However, proactive coaching significantly and unexpectedly reduced adherence to LTOT compared with reactive coaching (adjusted OR for use ≥17.7 hours per day = 0.40, 95% CI, 0.20-0.79, P < .01; secondary comparison). We screened patients for eligibility in all 50 US states, but our approach to study recruitment favored enrolling patients who were interested and able to participate in phone-based coaching interventions delivered in English, which may limit the applicability to other study populations. As there was no statistically significant difference between each active comparator and usual care in the primary outcome of the study, our study does not justify the use of peer coaching to improve adherence to LTOT. The lower levels of adherence in the proactive coaching group compared with reactive coaching was an unexpected result of a secondary analyses and therefore needs further study; this finding could represent more appropriate use of LTOT since studies suggest overuse of oxygen is common in this population, but this possible explanation requires confirmation.33,34
Proactive coaching significantly improved some secondary outcomes. The proactive coaching intervention significantly improved symptoms of depression and sleep disturbance compared with usual care (and symptoms of depression compared with reactive coaching) in the 0- to 60-day interval. The magnitude of improvements in these secondary outcomes in the proactive group (mean difference in T scores of 2-3.2) suggests that the observed effects are likely to be meaningful.
Patients reported several factors that could serve as barriers or facilitators to implementing phone-based peer coaching in this population, including the extent to which the educational curriculum was personalized, the extent to which study participants understand study procedures, and the strength of the relationship with the peer coach.
Some randomized clinical trials have reported health benefits with peer coaching (eg, glycemic control, improvements in blood pressure),11,35,36 and a pre-post intervention study in patients with α1-antitrypsin deficiency–associated COPD demonstrated improvements in self-management (including adherence to LTOT) and health outcomes (COPD exacerbations, acute care health encounters) with phone-based coaching by specialized staff.12 To our knowledge, the PELICAN study is the first randomized clinical trial to evaluate phone-based peer coaching in patients with COPD to support appropriate use of LTOT. The lack of a significant difference in the primary outcomes (LTOT adherence over the 0- to 60-day interval) with peer coaching (proactive or reactive) compared with usual care provides justification for the design and conduct of rigorous, controlled studies to evaluate care models in the context in which they are to be used, rather than simply repurposing interventions that have been shown to benefit other patient populations.
We also observed a significant improvement in 2 secondary outcomes (depression symptoms and sleep disturbance) with proactive peer coaching compared with usual care. To our knowledge, ours is the first study to suggest improvements in various aspects of patient-reported health in patients with COPD on LTOT using a proactive phone-based peer coaching model. It is possible that the social support provided through peer coaching had an overall positive effect on mental health that was captured across multiple domains of the PROMIS questionnaires. The size of the effect on depressive symptoms is likely to be clinically meaningful, though the minimum important difference for these measures has not been defined in patients with COPD using LTOT.20-23
Given the lack of effect on the primary outcomes, we do not advocate implementing our phone-based peer coaching intervention to promote appropriate use of LTOT in COPD. Our ability to recruit and implement a nationwide telephone-based peer coaching program and to collect patient-reported outcomes suggests this type of infrastructure is promising for delivering and scaling some interventions. This infrastructure, for example, could serve as a template for future studies, including those to confirm our findings regarding the effect of proactive coaching on depression symptoms and sleep disturbance. We would recommend that future studies ensure that peer coaching models are coordinated with the patient's prescribing clinician, which could have assisted us in securing physician prescriptions for LTOT. We also experienced significant challenges to collaborating with a national DME provider, even though we had engaged the provider's leadership early on (in the design of the application to PCORI) and continuously during the study period (see Methods, Conduct of Study section). Additional work is needed to understand and successfully overcome barriers to collaborating with the DME industry.
The US Centers for Disease Control and Prevention (CDC) used data from the 2011 Behavioral Risk Factor Surveillance System (BRFSS) to generate its first comprehensive report on the epidemiology of COPD in the United States.37 Compared with individuals in this CDC BRFSS report, PELICAN study participants had a similar proportion of women (about two-thirds) and non-White participants (about a fifth). However, the PELICAN participants were older (65 years or older: approximately 60% vs 50%) and more likely to have education beyond high school (approximately 61% vs 45%). Similar reference statistics for patients prescribed LTOT 24 hours per day in the United States are not available, so it is difficult to know if our study population differs in important ways from patients with COPD prescribed LTOT 24 hours per day in the United States. The following are factors that favor increased generalizability: (1) We used multiple sources of recruitment that covered all 50 US states and (2) we used inclusive study eligibility criteria and had limited losses to follow-up (see the “Aim 2: Pilot-Test Study Procedures for the PELICAN Trial” section). However, our approach to study recruitment favored enrolling patients who were interested and able to participate in phone-based coaching interventions delivered in English, which may limit the applicability to other study populations. This recruitment strategy may have also contributed to enrolling an “engaged” population that was generally more adherent (74% were adherent in the usual care group) than that of other studies (range of adherence 45% to 70%).
We conducted exploratory analyses to evaluate the potential for heterogeneity of treatment effects in various subpopulations defined by age; gender; race; marital status; highest level of education; duration of LTOT; oxygen flow rates; types of home oxygen equipment; patient-reported diagnoses of depression, anxiety, and sleep apnea; and acute care utilization in the past 12 months. These exploratory analyses do not suggest that the observed effects on LTOT adherence (primary outcome) vary significantly in any of these subpopulations.
This study has some limitations. The principal limitation is the risk of bias due to missing data for the primary outcome in just more than a quarter (27%) of study participants. Participants with missing data were more likely to be non-Hispanic Black or other minority race/ethnicity, to have less formal education, to have depressive symptoms, and to have been recently hospitalized. All of these factors have been associated with differences in adherence to medical therapies, though their association with adherence to LTOT is less well understood.38,39 Findings were unchanged, however, if we replaced missing adherence values as adherent (best-case) or nonadherent (worst case; see Table 11). Nevertheless, we acknowledge that we cannot exclude the potential for bias to limit inferences, particularly given that baseline differences existed between participants with evaluable primary outcome data and those without. Additionally, we did not collect LTOT use data while participants were hospitalized, which may have led to underestimating LTOT adherence; or the specific brand/model of O2 delivery device participants had, which may have affected its use. Given the pragmatic design of the study, we also did not collect information on how often participants randomized to the reactive and usual care groups-initiated calls to the COPD InfoLine. However, this is unlikely to have had a significant effect on LTOT adherence, as these participants would have received education on general COPD self-management and, unless requested by the participant, not focused on LTOT. Also, we were unable to determine if the lower LTOT adherence in the proactive coaching group compared with the reactive coaching group was due to worsening underuse (a potential harm of proactive coaching) or successfully limiting overuse (a benefit of proactive coaching). In our original study design, we had planned to collect data from home visits conducted by DME providers to supplement the information we gained from study participants, but we were unable to collect such data.
Additional research is needed to confirm the potential benefits of phone-based proactive peer coaching on depressive symptoms and sleep disturbance in patients with COPD prescribed LTOT. Also, the role of proactive peer coaching to facilitate LTOT discontinuation (deimplementation) when clinically indicated should be evaluated. We also recommend studies to examine the role of proactive peer coaching in patients recently initiated on LTOT (>80% of study participants in the current study had been on LTOT for at least 12 months). Moreover, future studies should include baseline measures of patient activation, caregiver engagement, and perceived level of self-management skills, all of which may serve as mediators or moderators of the peer-coaching intervention.
Given the lack of effect on the primary outcomes, we do not advocate implementing our phone-based peer coaching intervention to promote adherence to LTOT in patients with COPD. We experienced significant challenges to retaining DME providers as study partners, a significant barrier to research to improve the care and outcomes of patients on LTOT. Our results do provide favorable preliminary data about the feasibility of implementing peer coaching by phone in this high-risk COPD population, and about some potential benefits of proactive peer coaching, which require confirmation in future research.
Research reported in this report was [partially] funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (CE-1304-6490). Further information available at: https://www.pcori.org/research-results/2013/using-phone-based-peer-health-coaching-improve-home-oxygen-use-and-health
Krishnan J, Casaburi R, Cerreta S, et al. (2018). Using Phone-Based Peer Health Coaching to Improve Home Oxygen Use and Health in Patients with Chronic Obstructive Pulmonary Disease – the PELICAN Study. Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/11.2018.CE.13046490
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...
