Peer-Driven Intervention for Care Coordination and Adherence Promotion for Obstructive Sleep Apnea: A Randomized, Parallel-Group Clinical Trial

Abstract

Rationale

Obstructive sleep apnea (OSA) is a common condition that is usually treated by continuous positive airway pressure (CPAP) therapy, but poor adherence is common and is associated with worse patient outcomes and experiences. Patient satisfaction is increasingly adopted as a quality indicator by healthcare systems.

Objectives

We tested the hypothesis that peer-driven intervention effected through an interactive voice response (IVR) system leads to better patient satisfaction (primary outcome), care coordination, and CPAP adherence than active control.

Methods

We performed a 6-month randomized, parallel-group, controlled trial with CPAP-naive patients recruited from four centers and CPAP-adherent patients who were trained to be mentors delivering support through an IVR system.

Measurements and Main Results

In 263 patients, intention-to-treat analysis of global satisfaction for sleep-specific services was better in the intervention group (4.57 ± 0.71 Likert scale score; mean ± SD) than in the active-control group (4.10 ± 1.13; P < 0.001). CPAP adherence was greater in the intervention group (4.5 ± 0.2 h/night; 62.0% ± 3.0% of nights >4 h use) versus the active-control group (3.7 ± 0.2 h/night; 51.4% ± 3.0% of nights >4 h use; P = 0.014 and P = 0.023). When compared with the active-control group, the Patient Assessment of Chronic Illness Care rating was moderately increased by an adjusted difference of 0.33 ± 0.12 (P = 0.009), Consumer Assessment of Healthcare Provider and Systems rating was not different (adjusted difference, 0.46 ± 0.26; P = 0.076), and Client Perception of Coordination Questionnaire was mildly better in the intervention group (adjusted difference, 0.15 ± 0.07; P = 0.035).

Conclusions

Patient satisfaction with care delivery, CPAP adherence, and care coordination were improved by peer-driven intervention through an IVR system. New payor policies compensating peer support may enable implementation of this approach.

Clinical trial registered with www.clinicaltrials.gov (NCT02056002).

At a Glance Commentary

Scientific Knowledge on the Subject

Obstructive sleep apnea is a common condition that is usually treated by continuous positive airway pressure therapy, but poor adherence is common, is associated with worse patient outcomes and experiences, and is a major impediment to clinical care and research in this area.

What This Study Adds to the Field

Patient satisfaction with care delivery, continuous positive airway pressure adherence, and care coordination was improved by peer-driven intervention with an interactive voice response system. New payor policies with compensation for peer support may enable implementation of this approach to patients with obstructive sleep apnea.

Obstructive sleep apnea (OSA) is a prevalent condition that is characterized by repetitive collapse of the upper airway during sleep (1). OSA is an independent risk factor for cardiovascular morbidity and mortality (2–11). OSA is commonly treated by continuous positive airway pressure (CPAP) and has been associated with improvements in health-related quality of life (HR-QOL) (12), hypertension (2), accidents (13, 14), and mortality (15–17). However, CPAP nonadherence affects a large proportion (46%) of adults with OSA (18, 19) and has been associated with increased risk for cardiovascular events and death (15, 17). Poor CPAP adherence has hampered the ability to demonstrate potential cardiovascular benefits of treating OSA by adversely influencing intention-to-treat (ITT) analyses (20–25). Our study addresses calls for reliable interventions and care-delivery models that promote CPAP adherence and reduce fragmentation of care (26–32).

Promoting adherence to therapy through peer-driven intervention (PDI) is relatively inexpensive and has met with modest success in the management of chronic medical conditions (33–37). The current study was informed by a prior pilot study that revealed that PDI may improve CPAP adherence in patients with OSA (38). Whether PDI on a larger scale can improve patient-centered outcomes in patients with OSA is currently unknown.

Demonstration of adequate CPAP adherence in the first 90 days of therapy is a requirement for continued benefits by insurance payors and, unfortunately, leads to loss of CPAP medical benefits (19, 39). The overarching aim was to study the effectiveness of a patient-centered PDI with interactive voice response (PDI-IVR) on global patient satisfaction with sleep services, patient ratings of care coordination, CPAP adherence, and health outcomes when compared with usual care. Our patient-stakeholder–driven study design identified the underpinning summative measure of all the components needed for better CPAP adherence as patient satisfaction. Patient satisfaction is deeply intertwined with better healthcare quality, improved care coordination, and better patient outcomes, including mortality, and is increasingly adopted by healthcare systems and payors as a quality indicator (40–42). We hypothesized that patients in the PDI-IVR group will experience greater patient satisfaction (43, 44) and better ratings of chronic disease management and care coordination (45, 46) than patients in the active-control group. We hypothesized that patients in the PDI-IVR group will experience greater CPAP adherence, patient activation, and self-efficacy than patients in the active-control group. We also hypothesized that patients in the PDI-IVR group will experience greater improvements in HR-QOL, vigilance, and blood pressure than patients in the active-control group.

Methods

This was a randomized, parallel-group study with an active-control comparator arm. We recruited patients with recently diagnosed OSA who had not yet initiated CPAP therapy and randomly assigned them to the PDI-IVR system to promote adherence to CPAP therapy (intervention group) or be provided with educational brochures regarding OSA and CPAP therapy (active-control group) while both groups received usual care. The raters assisting with outcome evaluations and questionnaires were blinded to the group assignment of the participant being tested. We ensured broad participation and generalizability by promoting inclusive participation with bilingual peers (peer buddies) (36). After trial commencement (April 2014), there was an institutional review board (IRB)-approved increase in maximum recruitment to account for greater-than-anticipated attrition and achieve the target sample size; the study was stopped when last participant exited the study (January 2017). IRB approval (University of Arizona IRB #1401180444) was obtained before study commencement, and all participants signed informed consent forms before study participation. Please see the data supplement for more information regarding methodology.

Selection Criteria

Patients with recently diagnosed OSA were recruited from four sleep centers. Inclusion criteria were: 1) OSA (defined by apnea–hypopnea index [AHI] ⩾ 5/h) and naive to CPAP therapy. AHI was determined by polysomnography or home sleep studies as per American Academy of Sleep Medicine (AASM) criteria (47); 2) age range 21–85 years. Exclusion criteria were: 1) central sleep apnea; 2) participation in an intervention-based research study; and 3) medical instability. Peer buddy selection criteria are provided in the data supplement.

Peer buddy training

Peer buddies were trained by research staff to share their experiences and not to provide medical advice. Sharing of coping strategies was aimed at promoting self-efficacy, outcome expectancies, risk perception, and patient activation.

Fidelity assessments, involving 20% of telephonic communications and 20% of in-person visits, were randomly recorded to verify and increase fidelity of the PDI.

Comparators

Intervention group

In the PDI-IVR intervention group, trained peer buddies with OSA were matched with the subjects regarding age, sex, race, ethnicity, socioeconomic, language (Spanish speaking), and educational backgrounds. During the 6 months, the trained peers shared experiences on coping strategies that were delineated during their training and aided by the peer buddy training manual. The interaction occurred during two supervised in-person sessions (Days 1 and 7) and telephone-based conversations through the IVR system (once a week for 1 month followed by fortnightly calls). In addition, they helped the patient navigate the healthcare delivery system and effect better care coordination through the IVR system at patient-initiated request. The 10 interactions (2 in person and 8 telephone based) in the first 3 months were mandatory and were not to exceed 30 minutes each. The interactions with the peer buddy in the subsequent 3 months were optional and on demand by the research participant (maximum eight and not more than 30 min each).

Active control with usual care description

Usual care of the patient with newly diagnosed OSA consisted of attending a CPAP initiation and education class, which is conducted either at the patient’s home or at the DME company by a respiratory therapist for both comparators. Patients were educated about the basics of the care and operation of the device, mask, and related equipment. Patients in both arms were seen in the sleep clinic at 1 and 3 months after initiation of CPAP therapy and had the option to call any of their providers. To balance the number of contacts and educational sessions between the two comparison groups, educational brochures (from AASM) and OSA and CPAP educational videos on DVD (from industry stakeholders) were mailed to participants in the active-control group at the same frequency as the scheduled contacts that occurred in the PDI-IVR arm.

Randomization and Outcomes

Random assignments to the study arms occurred after enrollment. Randomization was simple and stratified by site with participants followed for 6 months.

Outcomes were as follows: 1) Global satisfaction with the care received for CPAP support from their sleep physician and sleep center on a 5-point Likert scale (primary outcome), satisfaction with sleep physician alone and sleep center alone at 6 months (43, 48). Likert scales (5 points ranging from “strongly agree = 5” to “strongly disagree = 1”) are used to assess patient satisfaction of healthcare systems and have excellent psychometric properties (46, 49). 2) Patient Assessment of Care for Chronic Conditions measures patient-centered care within the past 6 months and consists of five subscales: patient activation, delivery system design, goal setting, problem solving, follow-up and coordination (45, 46, 50, 51). 3) Client Perception of Care Coordination measures coordination of health care as perceived by the patient (46, 52). 4) Consumer Assessment of Healthcare Providers and Systems (CAHPS v4.0) item measures the patient’s satisfaction with care on a 0–10 scale (0 = worst and 10 = best health plan possible) (44, 46, 51, 53). 5) CPAP adherence was measured from the device-internal monitoring system and analyzed over 6 months and expressed as average time used per night, the proportion of days in which CPAP was used >4 hours, and Medicare-defined CPAP adherence (whether there were 30 consecutive days with >70% of nights of >4 h use per night). This was done in three segments: 0–30 days, 31–90 days, and 91–180 days. 6) Patient Activation Measure, a validated measure that assesses patient knowledge, skill, and confidence for self-management (54). 7) Self-Efficacy Measure for Sleep Apnea has strong psychometric properties with the ability to identify patient perceptions related to treatment adherence (55). Other outcomes included sleep-specific HR-QOL (Functional Outcomes in Sleep Questionnaire [FOSQ]), Psychomotor Vigilance Task, Epworth Sleepiness Scale, blood pressure, near-misses and motor vehicle accidents related to driving, hospitalizations and healthcare utilization (56–58).

Confounders affecting CPAP adherence were also collected: age, sex, race, severity of OSA (i.e., AHI), body mass index (BMI), CPAP level, and highest level of education received (43, 59). The Charlson comorbidity index (60) and nasal congestion score were also measured as potential covariates of interest (43).

Sample Size Estimations

A difference between the lowest and highest quartile of patient satisfaction rates that was associated with mortality in hospitalized patients and better quality indicators was 6.6%, with an interquartile range of 4.9% (41, 61). Conversion of the patient satisfaction rating used in the prior study to a 5-point Likert score would yield a minimal clinically important difference of 0.20 (SD, 0.20) and require a sample size of 38 per group with 99% power to detect a difference (α = 0.05; two-sided tail). Based on pilot data for changes in FOSQ global score with two repeated measures at 3 and 6 months, we estimated a sample size of 117 participants per group (total of 234 subjects) to be adequately powered to demonstrate a moderate effect of PDI-IVR (assuming 80% power; α = 0.05; two repeated measures; correlation between baseline and follow-up measurements of 0.7). With a conservative assumption of 10% attrition, we estimated that we would need to recruit 257 patients.

Statistical Approaches

Our study involved an ITT analysis that included outcomes data from all eligible subjects after randomization. Satisfaction with care delivery was compared between intervention and active-control groups by two-sample t tests of the mean scores as published previously by our group (43, 48), as well as differences adjusted for covariates sex, BMI, and AHI using multiple linear regression. We performed a sensitivity analysis of satisfaction with care delivery with adjusted and unadjusted ordinal logistic regression for greater interpretability (62). The adjustment for factors was preidentified as the primary statistical approach. Similar analyses were performed for secondary outcome patient ratings of health services, using multiple linear regression adjusted on the 6-month measure, adjusted for baseline measure, sex, BMI, and AHI. For outcomes measured at baseline, 90 days, and 180 days, we derived group differences with Generalized Linear Mixed Model on postbaseline outcomes, adjusted for baseline outcome, sex, BMI, and AHI. Differences in CPAP adherence were adjusted for age, sex, and nasal congestion score in multivariate regression (GLMM with missingness completely at random tested by Little’s MCAR test [IBM-SPSS v24.0]). We made only one comparison comparing the primary outcome between two groups, so there was not a need to use Bonferroni correction. However, for subscales of ratings of healthcare providers and center, a Bonferroni correction was applied. We conducted preplanned heterogeneity of treatment effect (HTE) analyses based on race/ethnicity (non-Hispanic White vs. others), sex, and age (<65, ⩾65 yr) because of prior reports showing lower adherence in minorities, women, and the elderly (63, 64).

Results

A total of 263 patients were enrolled; 131 patients were randomized to the PDI-IVR system arm with usual care and 132 patients to the active-control arm with usual care. Eight participants were disqualified for reasons specified in the data supplement, and ITT analysis was performed on 255 participants (Figure 1). Attrition was similar across the two arms, with 15 (12%) and 13 (10%) in the control and intervention arms, respectively (P = 0.61; χ2 test).

Figure 1.

CONSORT diagram. A total of 263 patients were enrolled and 131 patients were randomized to the PDI-IVR system arm with usual care (UC) and 132 patients to the active-control arm with UC. Attrition was similar between the two arms, with 22 (16.7%) and 14 (10.7%) in the control and intervention arms, respectively, who did not complete the study (P = 0.16; χ² test). If the 8 disqualified participants were excluded, the attrition was 15 (12%) and 13 (10%) in the control and intervention arms, respectively (P = 0.61; χ² test). AHI = apnea-hypopnea index; CPAP = continuous positive airway pressure; PDI-IVR = peer-driven intervention with interactive voice response.

Baseline Characteristics

There were no differences in baseline characteristics between the control group and the PDI-IVR group (Table 1). Demographics of peer buddies are provided in the data supplement. A total of 130 of the 131 patients assigned to the intervention arm completed both in-person visits. One patient completed only one of two in-person visits but was still included in the ITT analysis. In all, there were 1,857 total (1,596 IVR-based and the rest in person) interactions between patients and peer buddies, for an average ± SD of 14.2 ± 2.3 interactions per subject assigned to the intervention arm (n = 131). Although in-person visits lasted for an average of 27.2 ± 4.3 minutes [SD], the average duration of phone calls was 12.2 ± 15.0 minutes, plus subsequent optional, on-demand calls. In the control arm, there were a total of 1,848 educational brochures or DVDs provided to the 132 participants, with 14.0 ± 0.0 [SD] mailings per participant, which was similar to the 14.2 ± 2.3 interactions in the intervention arm (P = 0.92). IVR system diagnostics revealed that a vast majority of interactions occurred with the peer buddy (88.2%), with a distinct minority (2.4%) of interactions with other team members and the remainder (9.4%) being invalid or dropped calls.

Table 1.

Baseline Subject Characteristics

Characteristic	Intervention (n = 130)	Control* (n = 125)
Age, yr	50.4 ± 13.1	47.9 ± 12.8
Male	66 (50.7)	70 (56.0)
Race
White	111 (85.4)	106 (84.8)
Black/AA	8 (6.2)	7 (5.6)
American Indian/Alaskan Native	2 (1.5)	3 (2.4)
Other	9 (6.9)	9 (7.2)
Hispanic	43 (33.1)	51 (40.8)
Education
Less than HS	4 (3.1)	6 (4.8)
HS diploma	44 (33.9)	40 (32.0)
Undergraduate degree	53 (40.8)	47 (37.6)
Master’s degree	14 (10.8)	14 (11.2)
Doctoral degree	3 (2.3)	5 (4.0)
Unavailable	12 (9.2)	13 (10.4)
Height, inches	67.0 ± 3.9	67.4 ± 3.9
Weight, pounds	211.6 ± 46.3	215.4 ± 44.7
BMI	33.2 ± 7.0	33.4 ± 6.8
Systolic blood pressure	129.4 ± 15.6	128.2 ± 14.4
Diastolic blood pressure	79.7 ± 10.2	79.2 ± 8.9
AHI	30.4 ± 26.3	26.3 ± 24.8
CPAP level	9.8 ± 3.1	9.7 ± 2.8
Nasal congestion score	3.5 ± 1.3	3.4 ± 1.1
Charlson comorbidity index	0.69 ± 1.3	0.95 ± 2.0
Epworth Sleepiness Scale	11.3 ± 5.6	11.7 ± 5.2

Definition of abbreviations: AA = African American; AHI = apnea–hypopnea index; BMI = body mass index; CPAP = continuous positive airway pressure therapy; HS = high school.

Data are presented as mean ± SD or n (%).

^*There was no significant difference in characteristics between study arms.

Patient Satisfaction

After adjusting for confounders, global satisfaction for support provided for CPAP treatment of OSA was better in the intervention arm than in the control arm (difference, 0.46 ± 0.13 [SD]; P < 0.001; Table 2). Patient satisfaction with the sleep center alone (difference, 0.26 ± 0.12 [SD]; P = 0.070; Table 2) and patient satisfaction with the sleep physician alone (difference, 0.21 ± 0.14 [SD]; P = 0.258; Table 2) were not significant after Bonferroni adjustment. Sensitivity analysis with ordinal logistic regression models showed that the intervention group was more likely to report a higher global satisfaction rating than the active-control group (adjusted odds ratio, 2.61; 95% confidence interval, 1.51–4.49; P = 0.001).

Table 2.

CPAP Support and Satisfaction with Sleep Care

	Completion Measures			Adjusted Difference*
Generalized Linear Mixed Model	Intervention (n = 130) (mean ± SD)	Control (n = 125) (mean ± SD)	P Value	Difference between Arms (mean ± SD)	95% CI	P Value
Global satisfaction for support provided for CPAP treatment of sleep apnea† (primary outcome)	4.57 ± 0.71	4.10 ± 1.13	<0.001	0.46 ± 0.13	0.20–0.71	<0.001
Satisfaction with care from sleep center‡	4.54 ± 0.79	4.29 ± 0.99	0.038	0.26 ± 0.12	0.02–0.50	0.070§
Satisfaction with sleep apnea care from physician‡	4.37 ± 0.96	4.16 ± 1.06	0.134	0.21 ± 0.14	−0.06–0.48	0.258§

Ordinal Logistic Regression	Unadjusted OR (SE)	95% CI	P Value	Adjusted OR (SE)	95% CI	P Value
Global satisfaction for support provided for CPAP treatment of sleep apnea† (primary outcome)	2.53 (0.69)	1.48–4.31	0.001	2.61 (0.72)	1.51–4.49	0.001
Satisfaction with sleep center alone‡	1.82 (0.50)	1.06–3.12	0.029	1.87 (0.52)	1.08–3.22	0.05§
Satisfaction with sleep physician alone‡	1.53 (0.40)	0.92–2.55	0.103	1.56 (0.41)	0.93–2.62	0.190§

Definition of abbreviations: BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; OR = odds ratio.

^*Adjusted for sex, BMI, apnea–hypopnea index; positive difference indicates intervention favorable.

^† Five-point Likert agreement that CPAP support was satisfactory: 1 = strongly disagree, 5 = strongly agree.

^‡ Five-point Likert satisfaction score: 1 = very dissatisfied, 5 = very satisfied.

^§ Bonferroni adjusted P values are not significant.

Care Coordination

Patient Assessment of Care for Chronic Conditions was better in the intervention arm than in the control arm (difference, 0.33 ± 0.12; P = 0.009; Table 3). The Rating of Healthcare Providers and Systems using the Consumer Assessment of Healthcare Providers and Systems (difference, 0.46 ± 0.26 [SE]; P = 0.076; Table 3) was not better in the intervention group. Client Perception of Care Coordination tended to be superior in the PDI-IVR intervention when compared with the control arm (difference, 0.15 ± 0.07 [SE]; P = 0.035; Table 3). The ratings for the subdomains for acceptability was significantly better in the intervention than in the control group (Table 3). The specific item of coordination of care was better in the intervention group than the control group (difference, 0.35 ± 0.13 [SE]; P = 0.006). However, there were no differences across the group for patient ratings of their satisfaction with their providers.

Table 3.

Patient Ratings of Health Services

Scale	Baseline Measures			6-mo Measures			Adjusted Change*
	Intervention (n = 130)	Control (n = 125)	P Value	Intervention (n = 130)	Control (n = 125)	P Value	Difference between Arms	P Value
Total PACIC†	3.00 ± 0.10	3.02 ± 0.09	0.788	3.35 ± 0.10	3.01 ± 0.11	0.022	0.33 ± 0.12	0.009
Patient activation	3.53 ± 0.11	3.54 ± 0.11	0.967	3.82 ± 0.11	3.48 ± 0.13	0.038	0.29 ± 0.15	0.059
Delivery system design/decision support	3.54 ± 0.10	3.59 ± 0.10	0.621	3.74 ± 0.10	3.48 ± 0.11	0.099	0.24 ± 0.14	0.087
Goal setting	2.92 ± 0.10	2.97 ± 0.11	0.629	3.35 ± 0.11	3.00 ± 0.12	0.032	0.36 ± 0.13	0.006
Problem solving/contextual counseling	3.10 ± 0.12	3.16 ± 0.11	0.522	3.53 ± 0.12	3.19 ± 0.13	0.058	0.37 ± 0.16	0.017
Follow-up/coordination	2.37 ± 0.11	2.29 ± 0.11	0.703	2.70 ± 0.12	2.31 ± 0.12	0.022	0.34 ± 0.14	0.019
Rating of healthcare providers and systems (CAHPS)†	7.47 ± 0.18	7.36 ± 0.18	0.990	8.25 ± 0.16	7.74 ± 0.21	0.056	0.46 ± 0.26	0.076
Total CPCQ†	4.07 ± 0.05	4.14 ± 0.05	0.170	4.20 ± 0.05	4.09 ± 0.06	0.180	0.15 ± 0.07	0.035
Acceptability	4.04 ± 0.07	4.18 ± 0.06	0.078	4.20 ± 0.08	4.03 ± 0.08	0.088	0.24 ± 0.10	0.014
Received care	4.08 ± 0.05	4.07 ± 0.06	0.993	4.16 ± 0.06	4.12 ± 0.06	0.655	0.06 ± 0.08	0.437
Client comprehension	4.11 ± 0.06	4.12 ± 0.07	0.704	4.28 ± 0.07	4.12 ± 0.07	0.118	0.15 ± 0.09	0.097
Client capacity	4.09 ± 0.05	4.17 ± 0.06	0.254	4.17 ± 0.07	4.07 ± 0.07	0.342	0.13 ± 0.09	0.155
General practitioner	4.06 ± 0.08	4.09 ± 0.08	0.599	4.20 ± 0.07	4.08 ± 0.08	0.279	0.16 ± 0.10	0.123
Nominated provider	4.11 ± 0.07	4.25 ± 0.07	0.102	4.20 ± 0.08	4.08 ± 0.09	0.325	0.15 ± 0.12	0.202
Coordination of care (single item)	4.11 ± 0.07	4.13 ± 0.07	0.726	4.24 ± 0.08	3.90 ± 0.11	0.010	0.35 ± 0.13	0.006

Definition of abbreviations: CAHPS = Consumer Assessment of Healthcare Providers and Systems item (0–10); CPCQ = Client Perception of Care Coordination; PACIC = Patient Assessment of Care for Chronic Conditions.

Data are given as mean ± SE.

^*Adjusted for baseline measure, sex, body mass index, apnea–hypopnea index.

^† Higher scores are better ratings.

CPAP Adherence

At 6 months, CPAP use was greater in the intervention versus control group (P < 0.0001; GLMM). After adjusting for covariates such as age, sex, and nasal congestion score, CPAP adherence remained greater in the intervention (4.5 ± 0.2 h [SD]/night) versus the control group (3.7 ± 0.2 h [SD]/night; P < 0.01; GLMM; Table 4 and Figure 2). CPAP adherence measured by the Medicare criterion was better in the intervention (62.0% ± 3.0% [SD] of nights >4 h use) than in the control arm (51.4% ± 3.0% [SD]; P = 0.023; Table 4). To save one patient from losing their CPAP benefit, the number needed to treat with the PDI-IVR system would be 8–10. Essentially, by subjecting 8–10 patients to the PDI-IVR intervention, we would make at least 1 individual to be adherent by Medicare standards to not lose possession of their CPAP device. Sensitivity analysis by coding 0 minutes adherence for dropouts as well as per-protocol analysis did not show any meaningful changes from the aforementioned findings (Table 4).

Table 4.

Adherence Data with Sensitivity Analysis

Analysis Type	CPAP Adherence, hours/night Use* (mean ± SD)		Nights >4 h of Use (mean % ± SD)		Any Consecutive 30 d with 70% Nights >4 h/night† [proportion or OR (95% CI)]		NNT
	Intervention	Control	Intervention	Control	Intervention	Control
Unadjusted
ITT	4.6 ± 0.1	3.8 ± 0.1‡	62.3 ± 1.7	52.0 ± 1.8‡	52.1%	39.9%§	8
Per-protocol	4.6 ± 0.1	3.9 ± 0.1‡	62.3 ± 1.8	52.5 ± 1.8‡	52.6%	42.9%‖	10
Adjusted (age, sex, nasal congestion score)
ITT	4.5 ± 0.2	3.7 ± 0.2‖	62.0 ± 3.0	51.4 ± 3.0¶	1.6 (1.2–2.2)‖	Ref	—
Per-protocol	4.5 ± 0.2	3.8 ± 0.2**	62.0 ± 3.0	52.4 ± 3.1¶	1.6 (1.1–2.2)‖	Ref	—
Sensitivity analysis††
Unadjusted
ITT	4.2 ± 0.1	3.4 ± 0.1‡	56.8 ± 1.8	46.7 ± 1.8‡	47.5%	35.8%§	9
Per-protocol	4.6 ± 0.1	3.9 ± 0.1‡	62.3 ± 1.8	52.5 ± 1.8‡	52.6%	42.9%‖	10
Adjusted (age, sex, nasal congestion score)
ITT	4.1 ± 0.2	3.4 ± 0.2**	56.4 ± 3.1	46.7 ± 3.2**	1.6 (1.2–2.2)‖	Ref	—
Per-protocol	4.5 ± 0.2	3.8 ± 0.2**	62.0 ± 3.0	52.4 ± 3.1‡	1.6 (1.1–2.2)‖	Ref	—

Definition of abbreviations: CI = confidence interval; ITT = intention to treat; NNT = number needed to treat; OR = odds ratio; Ref = reference.

^*Average use over 180 days.

^† Medicare criteria for adherence.

^‡ P < 0.0001.

^§ P ⩽ 0.001.

^‖ P < 0.01.

^¶ P < 0.05.

^**P < 0.10.

^†† Sensitivity analysis with dropouts and lost to follow-up as 0 minutes.

Figure 2.

In patients with obstructive sleep apnea who were prescribed continuous positive airway pressure (CPAP) therapy, device internal monitoring system downloads revealed greater 6-month CPAP adherence in the intervention group receiving peer-driven intervention through an interactive voice response system with usual care (red) compared with patients receiving active-control and usual care (green). CPAP adherence at 6 months was the average between Days 91 and 180. Data for use per night (in h/night) is shown in left panel and proportion of nights with use >4 hours is shown in the right panel.

Patient Activation and Self-Efficacy

Patient activation was not different between intervention and control arms (adjusted difference, 3.4 ± 2.0; P = 0.084 [SE]; see Table E2 in the data supplement). On the Self-Efficacy Measure for Sleep Apnea scale (range of 1–4), although risk perception and outcome expectancies were not different between the two groups, there was greater improvement in self-efficacy in the intervention arm than in the control arm (0.24 ± 0.07 [SE]; P < 0.001; Table E2).

HR-QOL

Disease-specific HR-QOL (measured by the FOSQ) was not different between the two groups (0.34 ± 0.29; P = 0.25 [SE]; Table E2). The Epworth sleepiness score was not different (−1.0 ± 0.55 [SE]; P = 0.068; Table E2). Similarly, objective measurement of momentary sleepiness performed by the psychomotor vigilance test was not different between the two groups (difference in mean reaction time, 68.9 ± 72.4 ms [SE]; P = 0.34; Table E2). BMI and blood pressure measurements were not different across the groups (Table E2).

Sensitivity Analyses

Sensitivity analysis performed by doing a per-protocol analysis gave similar results, although the magnitude of difference in the outcomes was different (Tables E5–E8). The exception was that the Epworth sleepiness score was better in the per-protocol analysis, with the intervention group manifesting a greater adjusted improvement in Epworth score (−1.23 ± 0.61 [SE]; P = 0.042), but this difference was lower than the minimal clinically important difference. HTE analyses were preplanned and are shown in Table E3 for the primary endpoint. Although there did not appear to be HTE by age, race, sex, level of sleepiness, or apnea severity, we were not powered for performing such analyses.

There was no change in a composite measure of healthcare utilization (P = 0.21), accidents or near misses (P ⩾ 0.78), or blood pressure (P > 0.65; Table E2). The intraclass correlation coefficient for peer-level data relative to total variance was well below 10% for important outcomes, suggesting that there was no evidence for clustering according to the peer buddy delivering the intervention.

Discussion

Patient Satisfaction

Global patient satisfaction with the delivery of healthcare for CPAP therapy was better in the intervention group receiving the PDI-IVR than in the control group. Importantly, the PDI-IVR led to greater CPAP adherence and likely was related to the greater self-efficacy. Patient ratings of chronic illness care and care coordination were better in the intervention than in the control group. We believe that the peer intervention provided through the PDI-IVR system was the likely mechanism for the patients to give better satisfaction ratings. Such interpretation is substantiated by the vast majority of the patient interactions occurring with the peer buddies and a distinct minority with other team members of the sleep center. Moreover, patient ratings of the sleep center alone or sleep physician alone were not different between the study arms, thereby further supporting the interpretation that the peer buddies’ efforts are the likely underlying mechanism.

Patient satisfaction measures are driving various aspects of healthcare systems, with payors adopting them as quality indicators (40–42). Patient satisfaction measures are being widely used in various areas of medicine, including sleep disorders (38, 43, 48, 65–69), and have been associated with important patient outcomes such as better survival rates in various settings (41, 61).

Care Coordination

Patients with OSA face fragmentation of care and poor experiences related to limited access to sleep physicians (26). PDI-IVR could be harnessed by primary care providers (PCPs) to promote CPAP adherence and patient experience; however, it is unclear as to how the intervention implemented in a sleep clinic setting would perform in a primary care setting. The high levels of acceptability of PDI-IVR in our study could enable efficient interaction between patients, PCPs, and specialists and is scalable and exportable to other chronic medical conditions. Essentially, the peer buddy would alleviate the burden on PCPs by task-shifting some of the educational and motivational burden from the healthcare system to the peer buddy, while providing a platform for efficient care coordination.

CPAP Adherence

Previous gap analyses have called for interventions aimed at improving CPAP adherence in patients with OSA (27, 30, 70). Prior research in this area has revealed telemedicine, text messaging, and behavioral approaches as being successful in improving CPAP adherence (71–74). However, such approaches are expensive by virtue of the need for additional providers, or there are barriers to cultural adaptations and availability of internet in underresourced clinics and disadvantaged populations. PDI-IVR, through trained peer buddies by phone or in-person visits, is more accessible and culturally appropriate and is responsive to such earlier calls. Moreover, the observed improvement in CPAP adherence in our study was greater than the minimal clinically important difference (30 min/night) identified by the AASM (75). In addition, other approaches, such as anchor-based (34.1 min/night) and dispersion methods (33.6 min/night) were also smaller than the observed improvement of 47.5 minutes/night of CPAP adherence in our study. Such improvements in CPAP adherence, when contextualized with Medicare regulations that cause patients to lose insurance coverage for CPAP therapy, reveal a modest number needed to treat of 8–10 to prevent loss of CPAP device. Although most of the interactions occurred in the first 90 days of the subject participation period, the CPAP adherence pattern over time reveals that the effects of the PDI-IVR intervention persisted at 180 days (Figure E2).

Strengths, Limitations, and Future Research

One of the strengths of our study is the potential ability to quickly disseminate and implement our findings (38, 43, 76–78). Sleep centers can begin recruitment of peer buddies from their patient population and develop the IVR in a short period, with nominal investment of time and resources. Our PDI-IVR is easily scalable within primary care practices, considering that these practices would have patients who are successfully using CPAP therapy and can be trained using our peer buddy training manual and then educate various members of their sleep apnea care team. Moreover, new Medicare policies enable compensation for patient navigators through Principal Illness Navigation services that can compensate the peer buddies (79). Considering that there is a paucity of sleep specialists (∼7,500) compared with the public health burden imposed by sleep apnea (29.4 million patients with OSA in the United States), the model of PDI-IVR as a patient-centered approach to managing OSA and providing support would be beneficial to patients (80).

Our study is generalizable, considering that it was a pragmatic study with inclusive selection criteria and bilingual peer buddies who were embedded within sleep clinics. In addition, the PDI-IVR platform can be made widely available, considering that it can be administered remotely with telehealth adaptation.

Lack of blinding poses an issue; however, the raters assisting with outcome evaluations and questionnaires were blinded to the group assignment of the participant being tested. Nevertheless, this is a limitation in many intervention-based studies for CPAP adherence (70). We did not discern any age, race, or sex-based effect, but we had not powered the study for performing such analyses. Wozniak and colleagues had identified that measures for daytime sleepiness and HR-QOL suffered from imprecision (70). These may have been partly responsible for the lack of a significant effect on FOSQ and Psychomotor Vigilance Task measures.

In conclusion, PDI-IVR improved patient satisfaction with care delivery and care coordination as well as CPAP adherence. Although this patient-centered intervention is promising, future dissemination and implementation research is needed.