Adherence (Compliance) vs Use: A Conflict Causing Some to Lose Sleep Part I

Our paper on adherence vs use was published online last week in the journal Respiratory Care. The link to the article is on PubMed, which shows the abstract (also copied below): 

BACKGROUND: Research on subthreshold compliance with positive airway pressure (PAP) therapy in sleep apnea patients may inform clinical sleep medicine practice. We retrospectively assessed compliant and subthreshold compliant sleep apnea subjects to test whether regular but fewer hours of PAP use would demonstrate clinically meaningful improvements and correlate with outcomes.

METHODS: A chart review was conducted of 113 consecutive sleep apnea subjects, naive to treatment, who completed a titration study and filled a PAP therapy prescription. Objective data categorized subjects into 3 groups: compliant, subthreshold compliant, and minimal use. Outcome measures assessed changes in insomnia, sleepiness, and nocturia on average 7 months from PAP initiation. Correlation coefficients analyzed dose-response relationships between hours of use and changes in outcomes.

RESULTS: Among 113 PAP attempters, 104 (92%) were current users. Among 93 users with objective data, regular (consistent) PAP users included 59 compliant and 21 subthreshold compliant, and 13 subjects were minimal users. Compliant users averaging 6.6±1.3 h/night and 42.0±12.1 h/week showed the largest outcome improvements (all P < .05) with moderate to large effects for insomnia (d = 0.94), sleepiness (d = 0.58), and nocturia (d = 0.56). Subthreshold users averaging 4.1±0.7 h/night but only 18.0±5.6 h/week showed a large effect for insomnia (d = 0.76, P = .03) and nonsignificant, small effects for sleepiness (d = 0.38) and nocturia (d = 0.22). Correlation coefficients showed a trend for decrease in insomnia (P = .08; r = .20) and a significant decrease in nocturia (P = .034; r = 0.25), each in association with hours of PAP use.

CONCLUSIONS: In a clinical sample, 86% of sleep apnea subjects regularly used PAP, but adherence was 63%. Regular users showed clinical treatment effects and potential dose-response relationships, suggesting that the term use offers advantages over the term adherence. Currently, subthreshold compliance may not merit insurance coverage in many countries, an issue affecting many sleep apnea patients. 

We will spend some time dissecting this work and its clinical implications, because the concepts have great bearing on patient care for those who are nearly eligible for insurance coverage of their PAP devices, but who may actually lose their devices because of the fine-line between adherence and regular use.

As we described in the Introduction to our paper, if a diabetic were using too low a dosage of a prescribed medication yet was still receiving benefits from the drug, would an insurance company ever consider withdrawing coverage for the medication? Of course the answer is in the negative, because everyone is familiar with the adaptation process where an individual may need to start with a lower dosage to avoid side-effects. Then, gradually over time as the patient grows accustomed to the medication, the dosage can be raised to an optimal level. The identical process unfolds with many OSA/UARS patients who cannot adapt rapidly to the PAP devices, largely due to side-effects such as mask difficulties involving leak, fit, mouth breathing and skin irritation, or they cannot adjust to the pressure settings for months because PAP is an intrusive, foreign and unwieldy treatment even though it yields potent and sometimes near miraculous results. Indeed, many patients attempt PAP because they have heard numerous stories of great success from friends or family members. These friends and family members often serve as critical motivators who encourage their loved ones to give PAP a serious effort. Yet, the majority of first-time PAP users are more likely than not to report early adaptation difficulties instead of immediate success.

So, if the reality of using PAP is no different than the reality of someone starting out on antidepressant or anti-diabetic or anti-hypertensive medication, then why would insurance carriers, notably the Centers for Medicare and Medicaid Services (CMS) enact a rigid set of criteria, based on extremely limited evidence, to determine who does or who does not merit coverage for the PAP device?

Of no surprise to anyone, the answer is money. And, it certainly is a reasonable perspective as Medicare must hold itself accountable for budgeting taxpayer dollars. But as we suggest in our research paper and in the interview I conducted with Sleep Review Magazine, the Medicare approach may be particularly short-sighted, because it does not account for the more relevant factor of whether or not a patient is gaining benefit from the device regardless of the number of hours of use. In the Sleep Review piece, you will also find a comment from what appears to be a DME company attempting to further explain Medicare’s policies. I have now added a response to that comment describing more about how our research questions Medicare’s policies that do not address the issue of decreased symptoms in patients, but instead only looks at hours of use as the guiding metric.

In the research paper we first looked through the records of 113 patients in a set period of six consecutive months, and selected those patients who had never tried PAP therapy in the past and filled a prescription at their DME companies to start treatment after testing at our center. Of these 113 patients, our records confirmed only 9 patients were not using the device at the point of follow-up several months after starting treatment. Of the remaining 104 patients, we obtained objective data downloads (ODD) on 93 patients, after which we first divided them into two main groups of patients, those who minimally used their PAP devices (13 patients; 14% of the sample) and those who regularly used their PAP devices (80 patients; 86% of the sample). We then divided the latter group into two new groups, those who used regularly and were Medicare compliant (59 patients; 63% of the sample) and those who used regularly but were not Medicare compliant (21 patients; 23% of the sample). Note how combining both sets of regular users equals the 80 patients and 86% of those deemed to be using regularly.  

By hours of use, minimal users averaged fewer than 4 hours of PAP per week; the non-compliant regular users average more than four times the amount of the minimal users, at 18 hours/week; and the compliant users averaged more than ten times the amount of minimal users at 42 hours/week.  Yet, both the non-compliant and compliant users achieved important improvements in symptoms of insomnia, sleepiness, and nocturia. As expected, the compliant users achieved medium to large gains with their biggest changes in the symptom of insomnia. Sleepiness and nocturia (trips to the bathroom at night to urinate) showed moderate improvements. All these changes achieved statistical significance. In the non-compliant regular users, there was also a large and statistically significant improvement in the insomnia symptoms. For the other two symptoms of sleepiness and nocturia, the changes were in the small to medium range, which did not achieve statistical significance, yet which did demonstrate effect sizes likely to be clinically relevant.

As an aside, an effect sizes is typically a more reliable indicator of a change due to a therapeutic intervention.   It measures the improvement in a way that strongly suggests whether clinical change is real or not. Thus, it is quite common to find clinically meaningful effect sizes in small samples of patients under study, but due to the small numbers in the sample statistical significance is not achieved. In theory, when a future study is conducted with more patients in the sample, the effect sizes usually persists and statistical significance is also achieved. Thus, small sample studies sometimes serve as pilot or precursor research to guide the development of larger studies.

Although the average patient may not fully grasp this side note on statistics, it is still important to know that when you are discussing different therapies with your doctors, you might hear a physician or other provider commenting on whether a treatment showed statistical significance. You would do well in that instance to inquire about the effect size found with the treatment. Believe it or not something could be statistically significant and yet the effect size could be negligible suggesting minimal clinical effect. On the other hand as we are describing above, without statistical significance, it is valuable information to learn whether or not effect sizes demonstrate real clinical changes, and small to medium effects sizes sometimes mean just that. Unfortunately, many physicians untrained in statistics look only for the famous “p-value,” and if it is not significant (P < .05), then they are prone to disregard the research without looking deeper at the effect sizes.

Now, back to our study. Again, as expected the patients in the minimal user group had negligible hours of use and showed relatively small to minimal changes in symptoms. Still, there information was valuable in the last part of the data analysis where we attempted to look for a dose-response relationship between the hours of use and the change in symptoms. In this analysis, we looked at all patients combined and used all their data to plot a graph that would show whether the change in one variable (hours of use) correlated with a change in another variable (e.g. insomnia or nocturia).   The graphs appear in the research paper and demonstrated that when the number of hours increases for PAP, the symptoms of insomnia and nocturia decrease proportionally. For both symptoms, the correlation coefficient was consistent with a small effect and thus supported the idea of a dose-response relationship, which generally makes sense along the lines that if you take the full dose of a recommended medication, then you are likely to see the most benefits.

For these reasons, we are clear that the sub-threshold patients are gaining benefits, but from an even more practical viewpoint we say with confidence someone who is receiving benefit is more likely to continue using a therapy. In fact, they are more likely to continuously strive to increase their use of the therapy with the belief they would experience further gains. This, then, was the main point of the paper, because we show that sub-threshold patients are perhaps more accurately understood as patients in transition. Averaging only 18 hours per week yet receiving clear-cut benefits, they are using less than half the hours of the Medicare compliant group at 42 hours per week. However, because they currently reside in a “transitional zone,” we have high hopes they might prove more motivated in moving their number of hours into the 20 or 30 per week range. Moving into this range would bring them much closer to compliance, but even then it is worth noting you can total 27 hours of PAP in one week by using 3.9 hours per night for 7 nights and yet not qualify for Medicare coverage of the device.

This last point goes to the crux of the matter, because undoubtedly someone using PAP 27 hours per week is likely to gain benefits for insomnia, sleepiness, and nocturia, and yet eventually they could lose their device if they lost insurance coverage and did not have the funds to pay for it out of pocket. In the next post, I want to review some of the idiosyncrasies of insurance coverage and how it might affect patient care.

Barry Krakow MD


Dr Krakow’s 27 years of sleep research have focused on the complex relationship between physiological and psychological sleep disorders. Dr Krakow currently operates private sleep medical center, Maimonides Sleep Arts & Sciences, Ltd., and serves as Classic SleepCare’s paid Medical Director.

Leave a comment

Comments have to be approved before showing up.