Positive airway pressure (PAP) adherence data are a routine aspect of clinical care for obstructive sleep apnea (OSA), but not uniformly available. We hypothesized that mask refills are a measure of PAP adherence.
We measured PAP use over the first 90 days of treatment in 123 patients with OSA. The number and timing of mask refills was assessed over 18 months. Demographic and medical information was obtained from the electronic medical record.
Average PAP use in the first week of more than 4 h/d was a predictor of adherence over the first 90 days (P < .0001). PAP use over 90 days was greater in datacard-derived data dependent on patients presenting a datacard to the clinic compared to modem-derived data (P = .0006). A mask refill within the first 30 days of treatment was associated with a 1.3 h/d lower PAP usage in the first 90 days (P = .0044). Conversely, the number of mask refills between 30 days and 18 months was associated with higher PAP adherence during the first 90 days, with an additional 0.61 h/d of use for each additional refill (P = .0015).
In a retrospective cohort of veterans with OSA, first week PAP use was a strong predictor of 90-day use. Use of autonomously transmitted modem data avoided potential selection bias in adherence estimates. Mask refills in the first month were associated with less 90-day PAP use, whereas more mask refills after 30 days were associated with greater PAP use.
Scharf MT, Keenan BT, Pack AI, Kuna ST. Mask refills as a measure of PAP adherence. J Clin Sleep Med. 2017;13(11):1337–1344.
Obstructive sleep apnea (OSA) is a prevalent disease in the United States and worldwide.1 Positive airway pressure (PAP), the mainstay of therapy for OSA, is effective, safe, and widely available. Objective PAP adherence monitoring is currently a part of routine clinical practice. The use of objective data is important because subjective reports may not accurately reflect PAP use.2 With modern PAP units, autonomously transmitting modems are generally provided for at least the first few months of use, and sometimes for longer periods of time. Beyond this time, adherence information is generally only available when the patient brings in a datacard provided with the PAP unit into the clinic.
Comprehensive adherence monitoring is further constrained by the limited accessibility of adherence information. Each PAP unit manufacturer has its own proprietary software. In addition, even with online interfaces, each patient is associated with a particular durable medical equipment company and the referring clinic. This adherence information is not available to other providers or researchers. Furthermore, because autonomously transmitting modems are not often operable for the life of the PAP unit, acquisition of data may be dependent on patients returning to clinic. These constraints make large-scale acquisition of adherence data impossible. For investigators who study populations in databases such as the Medicare database, measurement of PAP adherence requires a surrogate measure.
Current Knowledge/Study Rationale: PAP adherence monitoring is a routine aspect of clinical care in patients with OSA, but direct PAP adherence data are not uniformly available. Use of an indirect measure of PAP adherence such as mask refills could allow for PAP adherence monitoring in situations where data would be otherwise unavailable.
Study Impact: Our study demonstrates that the number of mask refills correlates with directly measured PAP adherence. This suggests that the number of mask refills could be used as a measure of PAP adherence where adherence data are otherwise unavailable.
One potential surrogate measure for PAP use is the frequency of PAP mask refills. Masks deteriorate over time and patients are generally eligible for a new mask at least once every 3 to 6 months. In some systems, patients are required to initiate the order for a new mask, whereas in others a new mask is routinely provided as long as patients are continuing to use PAP. It seems likely that patients who are more adherent to PAP therapy would obtain mask refills at a greater frequency compared to patients who are less adherent; this makes the frequency of mask refills a potential surrogate for PAP adherence.
The primary purpose of this study was to determine whether mask refills were associated with PAP adherence and, thus, could potentially be used as a surrogate measure of adherence. We hypothesized that an increased number of mask refills would be associated with higher PAP usage, based on 90-day, primarily modem-derived PAP usage—the most comprehensive data we had available. We first characterized PAP adherence in our Veterans Administration (VA) population and validated the tools used to measure PAP adherence. Specifically, we addressed the issues of whether use of data-card-derived data dependent on patients bringing a datacard to clinic affected the measured adherence compared to patients who did not bring a datacard to clinic (ie, for whom we only had modem-derived data). Finally, we assessed the correlation between the number of mask refills over 18 months and the directly measured 90-day PAP adherence.
Design and Study Population
We performed a retrospective cohort study on the population served by the Crescenz VA Medical Center sleep center and affiliated clinical-based outpatient centers. Eligible veterans received a diagnosis of OSA, defined as an apnea-hypopnea index (AHI) ≥ 5 events/h, and had a PAP device ordered between June and December 2013. In order to be included, patients had to have received a PAP device and have lived ≥ 18 months since receiving the PAP unit. Of the 145 patients for whom an initial PAP order was placed, 4 patients died within the 18-month study period, 7 patients had the PAP order canceled, and 5 patients never received a PAP unit. Therefore, 129 patients met the inclusion criteria. The institutional review board of the Crescenz VA Medical Center approved this study and ethical standards were observed during the study.
A sleep study was ordered for patients presenting to the sleep center for whom OSA was suspected. For 8 patients with OSA, the diagnostic sleep study was done at an outside institution and a repeat sleep study was not obtained. Of the 115 sleep studies conducted, sleep studies were predominantly done with a home sleep apnea test (n = 103), but some patients underwent in-laboratory polysomnography (n = 12). The methods for these studies and their scoring have been published previously.3,4
PAP Initiation and Mask Orders
PAP therapy was ordered for patients in whom OSA was diagnosed. More than 98% of the PAP units were auto-CPAP (continuous positive airway pressure) units. On rare occasions, an in-laboratory PAP titration was conducted and a bilevel PAP (2 patients) was ordered. A durable medical equipment company made arrangements with the patients to deliver the PAP unit and recorded the day of PAP initiation. The patient was fitted for a mask by the durable medical equipment therapist and the mask was provided to the patient at the initial setup. The patients were instructed how to call the Prosthetics Department at the VA medical center for mask refills. A patient was eligible for a mask refill at least every 6 months and sooner if problems were encountered (eg, mask breakage or switching to another model or size).
PAP adherence data were obtained from autonomously transmitting modems and/or downloadable datacards. Every new PAP machine issued during this time period included a modem that transmitted data for 90 days, in addition to storing the data on a datacard. The autonomously transmitted modem data were available online via the Philips Respironics EncoreAny-where website. Patients were instructed to bring their datacard to all follow-up clinic appointments.
We retrieved individual data from the electronic medical record. The demographic data were taken from when the patient was started on PAP therapy. Variables included: age, sex, race, body mass index (BMI), presence of diabetes or hyper-tension, Epworth Sleepiness Scale (ESS) score, AHI on diagnostic sleep study, type of diagnostic sleep apnea testing, and type of PAP therapy prescribed. We also assessed whether the patients had any follow-up sleep center visits. To determine whether patients were still seeking medical care within the system, we assessed for the presence of any visits and/or medication refills ≥ 18 months after PAP initiation. This final determination was done at least 6 months after the termination of the study period.
All PAP units and mask refills were ordered by the Prosthetics Department at the VA medical center. The Prosthetics Department provided a list of all the PAP mask orders for the period of June 2013 to June 2015. Therefore, all masks that were ordered within 18 months of PAP initiation were assessed.
The measure of PAP adherence was average daily hours of usage, including days used and days not used. For 120 of the patients (98%), the PAP adherence data were obtained by modem transmission. For the remaining 3 patients (2%), modem data were not available, but 90-day adherence data were available from a datacard. We used 90-day adherence data because it allowed us to include more than 95% of eligible patients and, therefore, avoid a large selection bias (discussed in the following section; see Results).
Continuous variables are summarized using means and standard deviations and compared across groups using parametric t tests or analyses of variance (ANOVA), based on the number of groups. Categorical variables were described using frequencies and percentages and compared across groups using chi-square or Fisher exact tests.
Our primary outcome measure was PAP adherence (h/d) over the initial 90 days of use (ie, “90-day PAP adherence”), as described previously. Within this sample, we also summarized PAP usage over the first 7 and 30 days; longer term data (18 months) were available in a small sample of patients (n = 19). Primary analyses examined associations between 90-day PAP adherence and the number of mask refills over 18 months. Initial analyses examined the number of mask refills over the entire initial 18 months (categorized as 0, 1, 2, or ≥ 3 refills). However, given the hypothesized relationship between early mask refills and poor adjustment to PAP therapy, we also analyzed the number of mask refills separately during the first 30 days (any versus none) and the remaining 17 months (0, 1, 2, or ≥ 3 refills). Associations between mask refills and 90-day PAP adherence were assessed using unadjusted t tests (2 groups) or analysis of variance (ANOVA; > 2 groups) and linear regression models controlling for age, sex, race, BMI, diabetes, hypertension, ESS, AHI, and subsequent sleep visit. When looking at number of refills as a categorical measure (0, 1, 2, or ≥ 3 refills), we first tested the global null hypothesis of no differences among refill groups; if this null hypothesis was rejected then we performed pairwise comparisons assessing between-group differences. In addition, we examined whether there was evidence for a linear trend (or dose response) across groups by analyzing the groupings as an ordinal variable. In addition to these primary analyses examining the relationship between mask refills and 90-day PAP adherence, we examined the relationship between PAP adherence during the first week (< 4 h/d versus ≥ 4 h/d) and 90-day adherence using similar methods.
Next, to assess the potential effect of selection bias in studies relying on datacards or modems only, we examined the difference in 90-day PAP adherence in patients who brought datacards to clinic versus those for whom only modem data were available. Moreover, we used Bland-Altman analyses to examine the agreement between datacard and modem data among individuals with data available from both sources. Briefly, Bland-Altman analyses involve a series of analyses to robustly examine agreement, including: (1) testing whether there is a significant difference in 90-day adherence between the 2 methods (calculated as modem minus datacard values); (2) calculating a “limits of agreement,” equal to ± 2 standard deviations around the mean difference, to measure the expected range of differences in the 2 methods; and (3) examining the correlation between the difference and the average usage on the 2 methods to determine whether the amount of bias is related to the estimated values (eg, whether differences between modem and datacard are larger/smaller for higher estimates of adherence).
Finally, to examine how well 90-day PAP adherence associates with longer-term adherence, we explored associations between 90-day adherence group (< 4 h/d versus ≥ 4 h/d) and hours of PAP use in 90-day intervals from days 91 to 540. Analyses were performed using a linear mixed regression model to calculate both the overall difference in longitudinal 90-day adherence measures between the groups defined over the first 90 days and the difference between initial 90-day adherence groups within each 90-day interval.
Given the single primary outcome of interest (90-day adherence), statistical significance was based on an α = 0.05.
Analyses were performed using SAS, Version 9.4 (SAS Institute, Cary, North Carolina, United States) or Stata, Version 14 (StataCorp, College Station, Texas, United States).
One hundred twenty-nine patients met study criteria (see Methods). Of these patients, 123 (95%) had complete data sufficient for inclusion in the 90-day adherence analysis. The study cohort had a mean (SD) age of 55.6 (11.4) years and BMI of 34.0 (6.1) kg/m2; 114 patients (92.7%) were male, 60 (48.8%) were Caucasian, and 55 (44.7%) were African American. The level of OSA severity was moderate, with a mean (SD) AHI of 23.6 (17.6) events/h (see Table 1). Overall, patients had moderate levels of PAP adherence, which declined on average over the monitoring period. Specifically, participants used PAP for an average of 3.2 ± 2.7 h/d over the first 7 days, 2.8 ± 2.5 h/d over the first 30 days, and 2.5 ± 2.4 h/d over the first 90 days (see Table 1).
Summary of demographic, follow-up care, and PAP usage characteristics.
Summary of demographic, follow-up care, and PAP usage characteristics.
To gauge whether the patients were continuing to receive medical care at the VA through the 18-month study period, we examined whether patients had any visits and/or medication refills ≥ 18 months after PAP initiation. Ninety-nine patients (78.9%) presented for at least 1 visit, 101 (80.5%) had at least 1 medication refill, and 108 (86.2%) met at least 1 of these follow-up criteria. Similarly, 96 patients (75.6%) presented for a subsequent sleep center follow-up visit after obtaining a PAP unit (Table 1). Thus, most of the patients were continuing to receive at least some medical care at the VA after the study period and most had at least 1 follow-up visit at the sleep center.
Association Between First 7 Days of PAP Usage and 90-Day Adherence
Average PAP use of more than 4 h/d during the first week was associated with significantly greater 90-day adherence when compared to patients using PAP < 4 h/d in the first week. Specifically, compared to patients with less than 4 h/d of use in the first week, those who had more than 4 h/d of PAP use had an almost 3 h/d greater 90-day usage (4.3 ± 2.0 versus 1.4 ± 1.8 h/d; P < .0001). This association remained significant (P < .0001) after controlling for a number of covariates, including age, sex, race, BMI, diabetes, hypertension, ESS, AHI, and a subsequent sleep visit. Because the first week of data are included in the 90-day data, we carried out a subsequent analysis looking at more or less than 4 h/d of use in the first week and analyzed PAP usage over the subsequent 83 days. The results were similar to the 90-day usage data. Of the baseline covariates assessed, only higher baseline AHI (P = .0003) and subsequent sleep visit (P = .0040) were significantly associated with greater 90-day PAP adherence.
Effect of PAP Measurement Technique: Datacard Versus Modem
One potential problem in the interpretation of PAP adherence data is selection bias. For example, if patients lost to follow-up are excluded from analysis, the remaining study group may not be representative of the target population. To determine whether assessing only patients who presented for a follow-up with a datacard would affect the adherence assessment, we compared datacard data from patients who brought in a datacard to modem data from patients who did not bring in a datacard. Patients for whom datacard data were available were 5.4 years older on average (P = .009); other baseline characteristics were similar. As expected, 100% of the patients for whom datacard data were available had presented for a follow-up visit, compared to 51.6% of the patients for whom only modem data were available. Datacard-derived 90-day adherence was 3.6 ± 2.3 h/d, compared to 1.5 ± 2.0 h/d in modem-derived data (P < .0001; Figure 1); results remained significant after controlling for covariates (covariate adjusted P = .0006). These data demonstrate the potential bias toward increased adherence if measured PAP adherence only includes patients who returned to clinic with a datacard. In other words, not returning for care is an indication of poor adherence to PAP.
Comparison of 90-day PAP use based on datacard or modem device.
Patients who did not return for a follow-up sleep clinic visit and have modem only data have significantly worse 90-day PAP adherence compared to patients who brought a datacard to the clinic. * = adjusted for age, sex, race, BMI, diabetes, hypertension, Epworth Sleepiness Scale, apnea-hypopnea index, and subsequent sleep visit. PAP = positive airway pressure.
Comparison of 90-day PAP use based on datacard or modem device.
Modem use for PAP adherence is increasingly ubiquitous, but data are sparse validating modem accuracy in real-world settings. For the 90-day adherence data, a subset of 58 patients had both modem and datacard data. Modem data were very similar to datacard data (3.5 ± 2.3 and 3.7 ± 2.3 hours of use/d, respectively) and the average difference between the 2 groups was −0.18 h/d (P = .195, Figure S1 in the supplemental material). Importantly, we note that in 2 instances there were large differences between the modem and datacard measures, which are likely due to modem failures. When excluding these technical failures, there was almost no difference between the 2 measures (difference = 0.01 hours, P = .638, Figure S2 in the supplemental material) and a Bland-Altman limits of agreement ranging from −0.17 to 0.18 hours, suggesting that approximately 95% of the differences are expected to be within 10 minutes. Altogether, these data verify the accuracy of modem-transmitted data.
Association Between the Initial 90-Day PAP Adherence and Longer Term Adherence
In this study, we opted to use 90-day PAP usage data, as these were available on more than 95% of patients. Numerous studies have shown a correlation between short-term PAP use and long-term PAP use (see Discussion). To examine whether the initial 90-day PAP adherence was associated with longer-term adherence, we assessed whether more or less than 4 h/d of use in the initial 90 days was associated with PAP usage over the 5 subsequent 90-day periods ending at 18 months (see Figure S3 in the supplemental material). Overall, we found that patients who used PAP for ≥ 4 h/d in the initial 90 days had on average 2.1 hours (95% confidence interval [CI]: 0.62, 3.50) more PAP usage over the 5 subsequent 90-day periods (P = .0077). Results were generally similar when examining each 90-day period separately (see Figure S3). These data are consistent with results reported in the literature and further support the use of PAP usage data in the first 90 days of treatment as a measure of 18-month PAP usage.
Relationship Between Mask Refills and 90-Day PAP Adherence
We next assessed the relationship between mask refills over an 18-month period and PAP adherence over the initial 90 days. In general, we would expect excellent users to have at least 3 mask refills during the 18-month time period, as patients were instructed to get a mask refill at least once every 6 months. The baseline characteristics assessed were not significantly different in patients with 0, 1, 2, or ≥ 3 mask refills. Over the entire follow-up period, there was generally an association between increased number of mask refills and more PAP usage in both unadjusted (PANOVA < .0001, Ptrend = .0001) and covariate adjusted (PANOVA = .0011, Ptrend = .0364) analyses. Patients with no mask refills had 2.4 ± 2.7 hours of use, whereas patients with 1, 2, and at least 3 refills had 1.4 ± 1.8, 2.9 ± 2.3, and 4.1 ± 2.2 hours of use, respectively (see Figure 2). In adjusted pairwise comparisons, 1 mask refill was associated with less PAP use than 2 (P = .0260) or 3 (P = .0001) mask refills, and there was a nonsignificant trend toward less PAP use with 2 versus 3 mask refills (P = .145). Although these associations suggested a positive association between number of mask refills and adherence, interestingly, the group without refills had greater adherence compared to the group with only 1 refill in adjusted analyses (P = .0139). As a result of this observation, combined with clinical experience, we questioned whether the timing of mask refills could affect the relationship between the number of refills and adherence.
Comparison of 90-day PAP use by number of mask refills.
The number of mask refills over 18 months is associated with 90-day PAP adherence. * = adjusted for age, sex, race, body mass index, diabetes, hypertension, Epworth Sleepiness Scale, apnea-hypopnea index, and subsequent sleep visit. PAP = positive airway pressure.
Comparison of 90-day PAP use by number of mask refills.
Specifically, we reasoned that early mask refills may be the result of difficulty with PAP use, whereas mask refills later would be associated with better use. Indeed, we found that the presence of 1 or more mask refills in the first 30 days was associated with less PAP use when compared to no refills (2.8 ± 2.4 versus 1.9 ± 2.2 hours of use; Figure 3) in both unadjusted (P = .0317) and covariate adjusted (P = .0044) analyses. Excluding the first 30 days, we found that over the remaining 17-month period, a higher number of mask refills was linearly associated with greater PAP use in both unadjusted (PANOVA < .0001, Ptrend < .0001) and covariate adjusted (PANOVA = .0131, Ptrend = .0015) analyses (see Figure 4). Over this 17-month period, patients with no mask refills had 1.6 ± 2.2 h/d of use, and patients with 1, 2, or ≥ 3 mask refills had 2.2 ± 2.2, 3.1 ± 2.4, and 4.3 ± 2.0 h/d of use, respectively. In adjusted pairwise comparisons, patients with at least 3 refills had significantly greater usage compared to those with no refills (P = .0022) or one refill (P = .0105), and a nonsignificantly greater average usage than those with 2 refills (P = .240). As illustrated in Figure 4, there was a clear linear relationship between number of refills and PAP usage, with an estimated 0.86-hour increase in PAP usage for each additional mask refill in unadjusted analyses (β [95% CI] = 0.86 [0.52, 1.20]; P < .0001) and a 0.61-hour increase per refill in covariate adjusted models (β [95% CI] = 0.61 [0.24, 0.99]; P < .0015).
Comparison of 90-day PAP use by mask refills in 30 days.
The presence of one or more mask refills within the first 30 days is associated with worse 90-day PAP use. * = adjusted for age, sex, race, body mass index, diabetes, hypertension, Epworth Sleepiness Scale, apnea-hypopnea index, and subsequent sleep visit. PAP = positive airway pressure.
Comparison of 90-day PAP use by mask refills in 30 days.
Comparison of 90-day PAP use by mask refills from 1–18 months.
With exclusion of mask refills within the first 30 days, the number of mask refills between 1–18 months is linearly associated with increased 90-day PAP use. * = adjusted for age, sex, race, body mass index, diabetes, hypertension, Epworth Sleepiness Scale, apnea-hypopnea index, and subsequent sleep visit. PAP = positive airway pressure.
Comparison of 90-day PAP use by mask refills from 1–18 months.
Therefore, the number of mask refills over an 18-month period was associated with 90-day PAP adherence. Moreover, differentiating between early (first 30 days) and later (1–18 months) mask refills helped to clarify the observed association—mask refills over the initial 30 days were associated with worse PAP adherence and refills over the subsequent 17 months were associated with increased PAP use.
The purpose of this study was to determine whether mask refills could be used as a surrogate of PAP adherence. We demonstrated that the number of mask refills was associated with PAP use. More refills within the first 30 days of PAP use was associated with decreased PAP use, likely as a result of difficulty with PAP use. But over the subsequent 17 months there was a linear relationship between the number of mask refills and increased PAP use.
In addition to these primary results, we further examined predictors of and associations with PAP adherence over the initial 90 days in our population. We demonstrated the reliability of autonomously transmitted modem data and, further, that use of a data-card-based adherence measure dependent on patients returning to clinic caused a marked selection bias. Moreover, we showed temporal relationships between PAP adherence in the first week and over the initial 90 days, as well as between the adherence in the initial 90 days and longer-term usage in a smaller sample.
CPAP use in the short-term period after CPAP initiation has been shown to reliably predict long-term CPAP use 1 or more years later. In a multicenter study with sites in China, Australia, and New Zealand, CPAP use during the initial month of therapy was a predictor of CPAP use at 12 months.5 Specifically, every 1 hour of PAP use during the initial month of therapy was associated with 38 minutes of PAP use 12 months later.5 Similarly, numerous studies have demonstrated that using a threshold of CPAP use during the short-term period of initial use (between 12 days and 6 months) is a powerful predictor of long-term use between 1 and 5 years later.6–9 Therefore, patients with good adherence during the short-term period following CPAP initiation are far more likely to have good adherence in the long term.
In our study, we used 90-day PAP adherence as a primary measure of PAP adherence. In addition to the aforementioned studies that have shown the reliability of early CPAP use as a predictor of long-term CPAP use, we assessed this directly in the small subset of patients for whom we had 18-month PAP compliance data. We demonstrated that patients who used PAP for more than 4 h/d during the initial 90 days had greater PAP usage throughout the entire 18-month period. Our finding is consistent with a prior study which showed that CPAP use during the initial 3 months of therapy for more than 2 h/night was a powerful predictor of long-term use (median = 22 months, hazard ratio = 13.8).7 These results suggest that 90-day adherence data are likely to be representative of 18-month PAP use.
One of the strengths of our study was the large proportion of eligible patients that we were able to include. By utilizing primarily modem-derived data, we were able to include 95% of eligible patients (123/129) and therefore reduce the possibility of selection bias. In fact, we demonstrated that patients who brought a datacard to clinic had more than double the 90-day PAP use compared to those for whom only modem-derived data were available. Because one of the potential applications of assessing mask refills as a measure of PAP adherence is to identify nonadherent patients, exclusion of a large percentage of nonadherent patients would markedly limit the applicability of this study. We think that this is a significant limitation of a similar study looking at the association between CPAP supply refills and CPAP usage.10 In that study, the authors showed that 0.7 mask refills per year could be used as a threshold to distinguish adherent from nonadherent patients. However, only patients with datacard data for more than 1 year were included. This may have introduced a significant selection bias to their sample, as it could have excluded patients with poorer adherence who did not return for clinic follow-up.
As our study was heavily reliant on autonomously transmitted modem data, we sought to validate modem-derived data by comparing modem-derived and datacard-derived data directly. In the subset of patients for whom we had both sets of data, we demonstrated a high fidelity of modem-transmitted data, such that there was only an average difference of ∼10 min/d of PAP use between the 2 methods of data acquisition. Therefore, modem transmission of data was a reliable method of data acquisition.
Our study has shown that the number of mask refills is associated with 90-day PAP use. We hypothesized that it may not be simply the number of mask refills ordered that correlates with PAP use, but that the timing of those refills may be important. We found that the presence of 1 or more mask refills in the initial 30 days was associated with worse PAP adherence. After excluding mask refills ordered within the first 30 days from the analysis, there was a linear relationship between number of mask refills and PAP usage.
Consideration of the purpose of mask refills suggests that mask refills at these 2 different times are likely measuring 2 different things. Mask refills within the first 30 days likely indicate problems with adaptation to PAP therapy. Although some of these problems may be alleviated with a replacement mask, some of the issues precipitating a mask refill may not reflect a problem with the mask per se, but rather a general problem with tolerating PAP. As time progresses, mask refills likely reflect issues related to continued PAP use, such as wearing out of the mask, rather than the initial adaptation to PAP. Therefore, with exclusion of mask refills within the initial 30 days of therapy, a robust linear relationship between more mask refills and increased PAP adherence becomes apparent.
One limitation of our study was that it was done within the VA system, thereby limiting its generalizability. The VA system represents a demographic that is predominantly male and, by definition, has military service. Furthermore, the VA is a single-payer system with its own billing requirements and characteristics. For example, a new PAP order often generates a copayment for a patient with Medicare and private carrier health insurance. However, at the VA, ordering a PAP for a patient does not generally generate any charges directly to the patient. In other systems, the copay may lead to patients buying replacement PAP masks from a web-based supplier. Additionally, many health insurance plans require a certain degree of PAP adherence in order to pay for the unit and PAP supplies. At the VA, once a unit is ordered there is no adherence requirement. Therefore, whether a similar correlation between PAP usage and mask refills exists in other populations utilizing other health care systems remains to be determined.
Another limitation of our study was the relatively low PAP use. Patients averaged only 2.5 h/d over the first 90 days of use. Some of the poor adherence may be due to the inclusion of a relatively high percentage of patients (see above). Further, many factors have been implicated in modulating PAP adherence, including degree of sleepiness,2 severity of OSA,11 age,12 race,12,13 socioeconomic factors,3,13 and personality characteristics.14 There may also be other modulators in the Veteran population compared to the population at large. For example, combat-related posttraumatic stress disorder has been shown to worsen CPAP adherence.15 It is possible that posttraumatic stress disorder and other sequelae of military service contribute to the relatively poor PAP adherence observed in our Veteran population. Therefore, whether a similar relationship between mask refills and PAP compliance is observed in a more adherent population will need to be determined.
Most of the patients in our study received a diagnosis of OSA based on a home sleep apnea test and treated with auto-CPAP. The efficacy of this approach compared to in-laboratory polysomnography and PAP titration has been established.16,17 The average severity of OSA in our patients was moderate, with an average AHI of 23.6 events/h. At this severity, it is unlikely that the use of home sleep apnea tests, which may underestimate OSA severity, would have resulted in a clinically meaningful difference. However, whether a similar relationship is observed between mask refills and PAP compliance in patients with mild OSA and tested and treated via a different algorithm will need to be determined.
In summary, mask refills over an 18-month period correlated with PAP adherence over 90 days, which is associated with longer term adherence. Mask refills within the first 30 days were associated with worse PAP use. However, once mask refills within the first 30 days were excluded, more mask refills were associated with more PAP use in a linear fashion. Modems accurately transmitted PAP usage data and use of modem-derived data eliminated a large selection bias compared to using data derived from datacards only, which is available only in the subset of patients that present their datacards to the clinic.
Our data support the notion that mask refills may be a useful surrogate of PAP adherence when assessing outcomes of therapy, particularly in studies using data from large databases such as the Medicare database. Although there is extensive literature on the effect of PAP on outcomes such as hypertension,18 diabetes,19 atrial fibrillation,20 and cardiovascular mortality,21 the ability to effectively assess these outcomes in a real-world setting is important. The availability of large databases containing data from millions of patients is an important resource to utilize. Use of surrogates of PAP adherence can allow for not only the correlation between PAP and various outcomes, but also for determination of how actual use of PAP may modulate these outcomes in routine clinical care. With greatly increasing amounts of “big-data” available, we propose that use of a PAP adherence surrogate will be a valuable tool for studying the link between OSA and other diseases.
This work was supported by the National Institute of Health (HL094307). Work for this study was performed at the Corporal Michael J. Crescenz VA Medical Center, Philadelphia, PA. All authors have seen and approved the manuscript. The authors report no conflicts of interest.
analysis of variance
continuous positive airway pressure
durable medical equipment
Epworth Sleepiness Scale
obstructive sleep apnea
positive airway pressure
posttraumatic stress disorder
The authors thank the staff of the sleep center at the Corporal Michael J. Crescenz VA Medical Center for technical assistance.
Punjabi NM. The epidemiology of adult obstructive sleep apnea. Proc Am Thorac Soc. 2008;5(2):136–143. [PubMed Central][PubMed]
Boyaci H, Gacar K, Baris SA, Basyigit I, Yildiz F. Positive airway pressure device compliance of the patients with obstructive sleep apnea syndrome. Adv Clin Exp Med. 2013;22(6):809–815. [PubMed]
Platt AB, Field SH, Asch DA, et al. Neighborhood of residence is associated with daily adherence to CPAP therapy. Sleep. 2009;32(6):799–806. [PubMed Central][PubMed]
Fields BG, Behari PP, McCloskey S, et al. Remote ambulatory management of veterans with obstructive sleep apnea. Sleep. 2016;39(3):501–509. [PubMed Central][PubMed]
Chai-Coetzer CL, Luo YM, Antic NA, et al. Predictors of long-term adherence to continuous positive airway pressure therapy in patients with obstructive sleep apnea and cardiovascular disease in the SAVE study. Sleep. 2013;36(12):1929–1937. [PubMed Central][PubMed]
Popescu G, Latham M, Allgar V, Elliott MW. Continuous positive airway pressure for sleep apnoea/hypopnoea syndrome: usefulness of a 2 week trial to identify factors associated with long term use. Thorax. 2001;56(9):727–733. [PubMed Central][PubMed]
McArdle N, Devereux G, Heidarnejad H, Engleman HM, Mackay TW, Douglas NJ. Long-term use of CPAP therapy for sleep apnea/hypopnea syndrome. Am J Respir Crit Care Med. 1999;159(4 Pt 1):1108–1114. [PubMed]
Furukawa T, Suzuki M, Ochiai M, Kawashima H, Yokoyama N, Isshiki T. Long-term adherence to nasal continuous positive airway pressure therapy by hypertensive patients with preexisting sleep apnea. J Cardiol. 2014;63(4):281–285. [PubMed]
van Zeller M, Severo M, Santos AC, Drummond M. 5-years APAP adherence in OSA patients--do first impressions matter? Respir Med. 2013;107(12):2046–2052. [PubMed]
Patel N, Sam A, Valentin A, Quan SF, Parthasarathy S. Refill rates of accessories for positive airway pressure therapy as a surrogate measure of long-term adherence. J Clin Sleep Med. 2012;8(2):169–175. [PubMed Central][PubMed]
Krieger J, Kurtz D, Petiau C, Sforza E, Trautmann D. Long-term compliance with CPAP therapy in obstructive sleep apnea patients and in snorers. Sleep. 1996;19 9 Suppl:S136–S143. [PubMed]
Budhiraja R, Parthasarathy S, Drake CL, et al. Early CPAP use identifies subsequent adherence to CPAP therapy. Sleep. 2007;30(3):320–324. [PubMed]
Billings ME, Auckley D, Benca R, et al. Race and residential socioeconomics as predictors of CPAP adherence. Sleep. 2011;34(12):1653–1658. [PubMed Central][PubMed]
Brostrom A, Stromberg A, Martensson J, Ulander M, Harder L, Svanborg E. Association of Type D personality to perceived side effects and adherence in CPAP-treated patients with OSAS. J Sleep Res. 2007;16(4):439–447. [PubMed]
Collen JF, Lettieri CJ, Hoffman M. The impact of posttraumatic stress disorder on CPAP adherence in patients with obstructive sleep apnea. J Clin Sleep Med. 2012;8(6):667–672. [PubMed Central][PubMed]
Rosen CL, Auckley D, Benca R, et al. A multisite randomized trial of portable sleep studies and positive airway pressure autotitration versus laboratory-based polysomnography for the diagnosis and treatment of obstructive sleep apnea: the HomePAP study. Sleep. 2012;35(6):757–767. [PubMed Central][PubMed]
Kuna ST, Gurubhagavatula I, Maislin G, et al. Noninferiority of functional outcome in ambulatory management of obstructive sleep apnea. Am J Respir Crit Care Med. 2011;183(9):1238–1244. [PubMed]
Pepperell JC, Ramdassingh-Dow S, Crosthwaite N, et al. Ambulatory blood pressure after therapeutic and subtherapeutic nasal continuous positive airway pressure for obstructive sleep apnoea: a randomised parallel trial. Lancet. 2002;359(9302):204–210. [PubMed]
Botros N, Concato J, Mohsenin V, Selim B, Doctor K, Yaggi HK. Obstructive sleep apnea as a risk factor for type 2 diabetes. Am J Med. 2009;122(12):1122–1127. [PubMed Central][PubMed]
Kanagala R, Murali NS, Friedman PA, et al. Obstructive sleep apnea and the recurrence of atrial fibrillation. Circulation. 2003;107(20):2589–2594. [PubMed]
Marin JM, Carrizo SJ, Vicente E, Agusti AG. Long-term cardiovascular outcomes in men with obstructive sleep apnoea-hypopnoea with or without treatment with continuous positive airway pressure: an observational study. Lancet. 2005;365(9464):1046–1053. [PubMed]