Issue Navigator

Volume 15 No. 09
Earn CME
Accepted Papers

Scientific Investigations

Association of Positive Airway Pressure Use With Acute Care Utilization and Costs

Douglas B. Kirsch, MD1; Hongmei Yang, PhD2; Andréa L. Maslow, PhD2; Michael Stolzenbach, MHA3; Andrea McCall, MHA, RN, CPHQ4
1Sleep Medicine, Atrium Health, Charlotte, North Carolina; 2Information and Analytics Services, Atrium Health, Charlotte, North Carolina; 3Continuing Care, Atrium Health, Charlotte, North Carolina; 4Quality Division, Atrium Health, Charlotte, North Carolina


Study Objectives:

The current value-based medical climate has led to scrutiny of all medical costs. Given the relationship between obstructive sleep apnea (OSA) and many significant comorbid medical conditions, treating patients with OSA via positive airway pressure (PAP) therapy might reduce health care expenditures. Our goal was to determine the relationship between PAP use and acute care utilization and costs.


This was a retrospective cohort study of adult patients who initiated PAP therapy after a diagnosis of moderate-severe OSA at a large integrated health system during 2014–2016.


The study consisted of 1,098 patients, of which 60% were on PAP > 4 h/night for ≥ 70% of the nights. The average h/night were 5.3 (standard deviation 2.4). Increasing PAP usage was associated with reduced inpatient (rate ratio [RR] 0.92, 95% confidence interval [CI] 0.86–0.98) and overall acute care visits (RR 0.96, 95% CI 0.92–0.99). The linear relationships were supported by the threshold effects identified in the categorical adherence measures. No linear association was identified with emergency department visits and inpatient stays. However, lower number of emergency department visits (RR 0.78, 95% CI 0.62–0.98) and inpatient stays (RR 0.56, 95% CI 0.35–0.91) were observed among adherent (> 4 h/night for ≥ 70% of the nights) patients.


PAP usage was linearly associated with reduced number of inpatient and overall acute care visits, and lower likelihood of having positive costs from these visits. PAP usage was also associated with less emergency department visits and inpatient stays when comparing adherent patients to less adherent patients.


Kirsch DB, Yang H, Maslow AL, Stolzenbach M, McCall A. Association of positive airway pressure use with acute care utilization and costs. J Clin Sleep Med. 2019;15(9):1243–1250.


Current Knowledge/Study Rationale: Many, though not all, research studies have demonstrated improvement in several comorbid medical conditions when treating obstructive sleep apnea (OSA) with positive airway pressure (PAP) therapy. This study was performed to assess whether the hours of use of PAP therapy for patients with moderate-severe OSA impact acute care utilization and costs

Study Impact: Improved treatment of patients with moderate-severe OSA with PAP therapy appeared to lead to reduction in acute care costs. Cost reduction has become increasingly important in the current medical economic climate, thus focus on diagnosis and effective treatment of OSA may prove valuable for both patients and health systems.


Obstructive sleep apnea (OSA) is a common medical condition in which the upper airway repetitively collapses during sleep, causing an arousal from sleep and/or an oxygen desaturation.1 OSA has been associated with increased risk for several health conditions, including hypertension, heart attacks, and stroke.2 A common treatment for OSA is positive airway pressure (PAP) therapy, which eliminates the collapse of the airway via nasal or nasal-oral air pressure. PAP therapy has been demonstrated to reduce the medical risks associated with OSA. Although PAP is the most effective treatment for stabilization of the airway during sleep, population data suggests that patients are not always tolerant to or willing to continue PAP therapy in the long term.3 Adherence tracking for PAP therapy has become more commonplace in the last several years, with modem-based tracking commonplace in most PAP devices from major manufacturers. The continuous data flow from PAP devices eases evaluation of large populations of PAP users and is fairly unique among medical treatments.

In the current medical economic climate, patients are paying increasing attention to medical costs, particularly when they are involved in plans with significant deductibles. Inpatient medical care, when it occurs, accounts for significant individual medical cost. Similarly, health systems with value-based medical models or who have risk-based contracts with insurance companies are also strongly focused on control of inpatient costs.

This study was performed to assess the impact of use of PAP therapy for moderate or severe OSA on hospital-based care and costs in a large health system. Moderate or severe OSA was selected as a criterion for the study based on trials which have suggested these patients are more prone to comorbidities.4,5 Atrium Health, a large integrated health care system in the southeastern United States, has the advantage of a comprehensive sleep center with a large patient volume, a home medical equipment company with resulting centralization of patient PAP data, and is a large health system with multiple hospitals that share a centralized charge service.


Study Population and Data Sources

A retrospective cohort study was conducted among patients who had a diagnostic or split-night in-laboratory sleep study with Atrium Health between January 1, 2014 and December 31, 2016. Patients were eligible if they: (1) were at least 18 years old, (2) were diagnosed with OSA (ICD-9-CM code 327.23 or ICD-10-CM code G47.33, inclusively) during the diagnostic study, (3) initiated PAP therapy, (4) had an apnea-hypopnea index (AHI) ≥ 15 events/h (defining moderate OSA) and (5) had a central apnea index (CAI) ≤ 15 events/h (which defines moderate central sleep apnea) at the time of the diagnostic study. Eligible patients were also required to survive at least 18 months since the onset of PAP therapy (ie, the index date) to November 10, 2017 (ie, the end of the study) to allow for time for potential changes in health care resource use and costs. Expiration status was identified using the Social Security Death Index and/or records in the electronic medical records.

Patients’ daily PAP usage data was obtained from Somnoware, a proprietary cloud-based respiratory care management platform (Somnoware Healthcare Systems Inc., Charlotte, North Carolina). Other data were obtained from the Atrium Health’s electronic data warehouse which contains patients’ clinical, billing, and scheduling data.


Primary outcomes were health care utilization. Visit counts to a health care setting has been employed to measure health care utilization in the literature.6,7,8 We measured health care utilization through the number of inpatient (IP), emergency department (ED) and observation (OBS) visits, and length of inpatient stays. Overall number of visits from acute care settings, which was the sum of IP, ED, and OBS visits, was also examined. The values of these variables were determined from hospital billing data for a period of 18-months starting on the index date. Elective visits were excluded, including transplant surgeries, maintenance chemotherapy, rehabilitation, and nonacute admissions for a scheduled procedure.9

Secondary outcomes were health care costs from IP, ED and OBS visits, and overall acute care costs over the 18-month period. Costs in the study included fixed and variable costs incurred to Atrium Health while delivering care to patients, such as salaries, supplies and implants, laboratory tests, pharmaceuticals, room and board, etc. Costs from elective visits were excluded. Cost data from different years were converted to 2017 US dollars based on the Consumer Price Index data from US Bureau of Labor Statistics for Medical care.10

PAP Adherence

PAP usage was measured as the average hours used per night (h/night) over all nights in the 18 months. The denominator was the number of nights over the 18-month (ie, 547 nights). The numerator was the total hours of PAP use during the same 18-month period. As there has been no agreed-upon minimum adherence level, we examined three measures of adherence based on usage: (1) numeric PAP h/night, (2) categorical PAP h/night, and (3) adherent status based on Medicare PAP coverage criteria. The numeric measure was used as a continuous variable to estimate the linear dose-response relationship between PAP usage and acute care utilization and costs. The categorical measure, grouped into four categories according to its distribution: 0–2, > 2–4, > 4–6 and > 6 h/night, was used to explore potential threshold effects. The Medicare reimbursement measure in which patients were considered being adherent if they used PAP > 4 h/night on ≥ 70% of the nights over the 18 months, was examined to assess if the usage duration was sufficient. This final measure was based on a metric utilized by the Centers for Medicare and Medicaid Services (CMS) to determine if a patient is deemed adherent to PAP therapy.11


To account for potential differences in baseline risk, the following covariates were considered in the study: (1) sociodemographic characteristics (age, sex, marital status, race and primary health insurance plan at the onset of PAP therapy); (2) Charlson comorbidity score12 and individual comorbidities (cardiovascular diseases, chronic respiratory conditions, chronic kidney diseases, endocrine conditions, depression, and insomnia); (3) concomitant medication use (antihypertensive agents, lipid-lowering agents, antiplatelet agents or coagulants, insulin and other antidiabetic medicines, sedatives and psychoactive agents); (4) history of health care resource usage (number of acute care visits, overall and by visit type, 12-month prior to the index date); (5) weight, body mass index, and AHI at the time of sleep diagnostic study; and (6) annual wellness check-up (yes/no) as a proxy for the healthy adherer effect.13 Comorbidities were identified based on the ICD-9-CM and ICD-10-CM diagnosis codes appearing on inpatient and outpatient billing records 12-month prior to the index date. Concomitant medications were defined based on patient’s prescriptions recorded by medical providers at Atrium Health during the 12-month period prior to the index date.

Statistical Analysis

Associations between PAP usage and acute care utilization and costs were examined for each of the three measures of adherence—numeric and categorical PAP h/night, and Medicare PAP coverage criteria. For each measure, the relationship between PAP usage and acute care utilization was analyzed using multivariate negative binomial regression, adjusting for baseline risk. The final model included only covariates that remained significant at a .05 significance level. Rate ratio (RR) and 95% confidence interval (CI) values were reported.

For each measure, the relationship between PAP usage and acute care costs was analyzed using multivariate hurdle regression.14 The two-stage model first employed multivariate logistic regression to estimate the probability of having a nonzero cost with adjusting for baseline risk that remained significant at a .05 significance level in the final model. Odds ratio (OR) and 95% CI values were reported. Then, a truncated-at-zero model controlling for relevant covariates was used for patients with positive costs. Inverse gaussian distribution was specified as the variances were proportional to the cube of the means. RR and 95% CI values were reported.

No offset variable was indicated in the above multivariate analyses as all patients had same study period. Model fit was assessed by Akaike information criterion (AIC) and P based on Pearson chi-square. Statistical analyses were performed with SAS version 9.4 (SAS Institute, Inc., Cary, North Carolina). The Institutional Review Board at Atrium Health reviewed and approved the study.


Baseline Characteristics

The study consisted of 1,098 patients, of which 60% (n = 665) were on PAP > 4 h/night for ≥ 70% of the studied nights (ie, adherent using CMS criteria). Table 1 presents the baseline characteristics of the study population. Compared to their counterparts, adherent patients were more likely to be Caucasian white, married and have commercial health insurance. They had a lower Charlson comorbidity index (CCI) score and a lower prevalence of ischemic heart disease, heart failure, hypertension, diabetes, and chronic kidney diseases. Adherent patients also had lower number of IP, ED, and overall acute care visits.

Baseline characteristics of the study population.


table icon
Table 1

Baseline characteristics of the study population.

(more ...)

PAP Adherence Data

PAP usage summaries are presented in Table 2. The average percentage of nights with PAP usage > 4 h/night was nearly 70% (standard deviation [SD] 30%). The average h/night being on PAP over all nights was 5.3 (SD 2.4), including 141 (13%) used PAP for an average 0–2 h/night, 166 (15%) > 2–4, 294 (27%) > 4–6, and 497 (42%) > 6 h/night.

PAP usage summaries.


table icon
Table 2

PAP usage summaries.

(more ...)

Association of PAP Usage With Acute Care Utilization

Descriptive statistics of acute care utilization and costs over the 18-month study period are presented across PAP adherence levels in Table 3.

Acute care utilization and costs over the 18-month study period.


table icon
Table 3

Acute care utilization and costs over the 18-month study period.

(more ...)

After adjusting for significant covariates, association between PAP usage and acute care utilization was still evident in all three measures of adherence (Figure 1). Increasing PAP usage was negatively associated with IP and overall acute care visits. For every 1 h/night increase in PAP usage, there was 8% decrease in IP visits (RR 0.92, 95% CI 0.86–0.98) and 4% decrease in overall visits (RR 0.96, 95% CI 0.92–0.99). Such linear relationships were supported by the threshold effects identified in the two categorical adherence measures. Adherent patients (> 4 h/night for ≥ 70% of the nights) had significantly lower IP (RR 0.62, 95% CI 0.44–0.87) and overall visits (RR 0.74, 95% CI 0.61–0.90). Compared to those on PAP 0–2 h/night, patients on PAP > 6 h/night had lower IP and a trend toward lower overall visits as well.

Association of PAP usage with acute care utilization, adjusting for significant confounding factors.

Acute care visits adjusting for baseline acute care visits, annual wellness check-up, depression, BMI, race, and health insurance. IP visits adjusting for baseline acute care visits, CCI, heart failure, diabetes, sedatives. ED visits adjusting for baseline acute care visits, annual wellness check-up, CCI, depression, race, and health insurance. OBS visits adjusting for baseline acute care visits, CCI, hypertension and insomnia. Inpatient stays adjusting for baseline acute care visits, CCI, sedatives, weight, and hypertension. BMI = body mass index, CCI = Charlson Comorbidity Index, ED = emergency department, IP = inpatient hospitalization, OBS = inpatient observation, PAP = positive airway pressure.


Figure 1

Association of PAP usage with acute care utilization, adjusting for significant confounding factors.

(more ...)

Although PAP usage was not linearly associated with inpatient stays and ED visits, significantly lower number of inpatient stays (RR 0.56, 95% CI 0.35–0.91) and ED visits (RR 0.78, 95% CI 0.62–0.98) were observed among adherent patients.

Association of PAP Usage With Acute Care Costs

Figure 2 presents the relationship between PAP usage and acute care costs, with top figures indicating the odds of having positive costs from each visit type, and bottom figures indicating the cost ratio among those with positive costs. Increasing PAP usage was associated with less likelihood of having positive cost from IP (OR 0.93, 95% CI 0.86–1.00) and overall acute care visits (OR 0.94, 95% CI 0.89–1.00), but it was not associated with the level of cost once such cost occurred. Such relationships were also observed when comparing adherent patients with nonadherent patients but were not apparent when comparing patients on PAP > 6 h/night with those on PAP 0–2 h/night.

Association of PAP usage with acute care costs, adjusting for significant confounding factors.

Acute care cost: logistic regression (stage 1) controlling for baseline acute care visits, annual wellness check-up, depression and race; truncated-at-zero (stage 2) model controlling for CCI, dementia, weight and marital status. IP cost: logistic regression (stage 1) controlling for baseline acute care visits, CCI, hypertension, and sedatives; truncated-at-zero (stage 2) model controlling for annual wellness check-up, dementia, and race. ED cost: logistic regression (stage 1) controlling for baseline acute care visits, annual wellness check-up, race, and health insurance; truncated-at-zero (stage 2) model controlling for baseline acute care visits, dementia, psychoactive agents, antihypertensive agents, and marital status. OBS cost: logistic regression (stage 1) controlling for baseline acute care visits, CCI, insomnia, lipid-lowering agents, oral antidiabetic medicines; truncated-at-zero (stage 2) model controlling for hyperlipidemia, marital status, and race. CCI = Charlson Comorbidity Index, ED = emergency department, IP = inpatient hospitalization, OBS = inpatient observation, PAP = positive airway pressure.


Figure 2

Association of PAP usage with acute care costs, adjusting for significant confounding factors.

(more ...)

There was no linear relationship between PAP usage and the occurrence and level of ED cost. However, among patients with ED cost, using PAP > 4 h/night on ≥ 70% of the nights was associated with 27% lower ED cost (RR 0.73, 95% CI 0.55–0.96).


The primary goal of this study was to assess the impact of PAP adherence on health care utilization and cost, to assess whether patients with more effectively controlled moderate or severe OSA would benefit from that treatment by reducing inpatient care. Previous studies have suggested that patients with untreated OSA have increased health care utilization compared to control populations.15,16 As well, a recent study from a military academic medical center suggested that outpatient health care utilization was reduced with use of PAP therapy.17 The current medical climate has led to scrutiny of all medical costs and the concept of value-based medicine emphasizes the importance of maintaining or increasing quality of care while reducing overall health care expenditures. While there is a long history of research studies that have correlated OSA with hypertension, cardiac disease, and mortality,18 recent studies, such as the SAVE trial, have raised the question of whether treating OSA in certain populations changes patient outcomes, such as mortality.19

This study evaluated nearly 1,100 patients with moderate to severe OSA on PAP therapy for at least 18 months to assess whether PAP adherence was correlated with a change in inpatient use and cost. There are clear criteria suggested by insurance companies and governmental payers for minimum short-term PAP adherence (such as 70% of days > 4 hours) with limited data to suggest these criteria are appropriate.20,21 However, there is no agreement about what a minimum PAP adherence level might be in the long term for patients. Some of the discordance likely stems from the differing outcomes under evaluation (eg, sleepiness, quality of life, mortality, etc.)

PAP adherence is commonly cited as being less than 50% in research trials, depending on the trial and the metrics applied.22 In our study, adherence was assessed in two primary different methods. One method involved averaging hourly use over the 18-month period; the other was applying the 70% of nights > 4 hours over the 18-month period. The study patients used PAP > 4 hours nearly 70% of all nights, with 69% of patients using PAP on average > 4 h/night and 60.5% of the total study population adherent based on the > 4 hours on 70% of nights metric. Research studies report that 50% of patients do not take their cholesterol medication as directed,23 which is similar to data on patient adherence to PAP therapy in the early 1990s.24 However, this study demonstrates that the current work done by the clinical staff to promote PAP adherence in this cohort appears to be more effective than in observed in those earlier studies.

Broadly, use of PAP above 4 h/night over the 18-month period was associated with a reduction in overall inpatient visits and costs. There was a linear response to hours of PAP usage to reduction in acute care visits. Using CMS criteria to PAP changed some of the subcategory associations, but generally led to similar conclusions of reduced visits and costs with patients using PAP > 4 hours on 70% of nights (a slightly higher bar than use averaging more than 4 h/night).

There are several potential reasons that increased use of PAP therapy may be associated with decreased health care utilization. Treatment of OSA improves daytime sleepiness, which in turn, reduces risk of a sleepiness-related injury such as a motor vehicle crash25 (a cause of inpatient or ED visits). Without treatment of their nocturnal breathing disorder, patients with OSA have an estimated crash RR of 1.2–4.89 when compared to controls.26 PAP therapy of OSA also likely has downstream effects on reducing severity of some medical comorbidities, such as hypertension or arrhythmia.27 Reducing the impact of these comorbidities may lower the risk for or reduce the acuity of a resultant inpatient visit. Patients who use PAP at a high rate may also be more adherent to treatment of other health conditions, thus PAP adherence in this case would be an association rather than a causative effect. The finding of our study can be contrasted with a recently published article suggesting that PAP dispensation does not impact utilization; it would appear that adherence to PAP treatment (rather than just dispensation of device), may have relevance to the reduction in utilization observed in our study.28

There are several potential limitations in the current study. First, although objective device-based daily usage data was used to assess PAP adherence, misclassification of PAP adherence level may exist due to technical problems and other man-made reasons. Problems such as failure of data downloads from the PAP device, interrupted wireless transmission, and failure to insert or replace the memory cards will result in underestimate of PAP usage. Second, acute care utilization and costs reported in the study may be underestimated because we were not able to account for health care use outside of Atrium Health, whether using a different health care system locally or more remotely. However, to our best knowledge, there was no evidence showing that patients with different adherence level might seek health care outside Atrium Health differently. Third, measures of PAP usage and the outcomes were aggregated summaries over the 18-month study period. Patient visits to acute care settings over the 18 months might result in changes in PAP use during the study period. Therefore, the relationship identified in the study between PAP use and acute care utilization and costs may not represent a solely causal relationship. Fourth, patients with different adherence levels differed significantly in some baseline risk factors, including several medical comorbidities. Some medical comorbidities may be associated with worse PAP adherence and increased medical visits. Although we specifically adjusted for significant confounding factors, it is debatable whether the confounding effects have been completely accounted for by the statistical model. More rigorous study design, such as difference-in-differences or pre-post with a parallel control group, may help address this challenge and enhance studies in the field.29,30 Fifth, we were not able to adjust for confounders that were not measurable or not available in our data, such as lower income which may contribute to both suboptimal adherence and increased acute care visits. Nevertheless, we used Medicaid as a surrogate for low socioeconomic status and adjusted for it during the analyses. Sixth, there was a self-selection bias, in that all patients agreed to initiate PAP therapy. Finally, our data comprised patients from an urban county in the southeastern US and may not be generalizable to other populations.

In conclusion, this retrospective study demonstrated the association of higher levels of PAP adherence with decreasing inpatient visits and lower inpatient costs in patients from a single health care system. The study results were notable given the fairly large number of study patients meeting criteria and the relatively short time of the study (18 months) in which a statistically significant improvement was seen. Given the increasing amount of data available from modem-based PAP units, studies of larger populations of patients or a longer window of PAP treatment should be considered to assess the stability and longevity of the relationship. As well, this study only included some health care costs, but did not include postacute or ambulatory costs; further research would ideally benefit from costs across the continuum of acute and ambulatory care and include all potential health systems used by an individual patient. As patients and health care systems evaluate methods to reduce medical costs, treating OSA effectively should be considered part of the solution. Researchers and clinicians in the field of sleep medicine should continue to focus on outcomes-based research to ensure that patients, payers, and hospital systems all understand the value that treatment of sleep disorders provides.


All authors have reviewed the manuscript and approved of the contents. Work for this study was performed at Atrium Health, Charlotte NC. The authors report no conflicts of interest.



apnea-hypopnea index


Akaike information criterion


central apnea index


Charlson Comorbidity Index


confidence interval


emergency department






obstructive sleep apnea


positive airway pressure


rate ratio


standard deviation



Young T, Peppard PE, Gottlibe DJ. Epidemiology of obstructive sleep apnea. Am J Respir Crit Care Med. 2002;165(9):1217–1239. [PubMed]


Jordan AS, McSharry DG, Malhotra A. Adult obstructive sleep apnoea. Lancet. 2014;383(9918):736–747. [PubMed]


Weiss P, Kryger M. Positive airway pressure therapy for obstructive sleep apnea. Otolaryngol Clin North Am. 2016;49(6):1331–1341. [PubMed]


Marshall NS, Wong KK, Cullen SR, Knuiman MW, Grunstein RR. Sleep apnea and 20-year follow-up for all-cause mortality, stroke, and cancer incidence and mortality in the Busselton health study cohort. J Clin Sleep Med. 2014;10(4):355–362. [PubMed Central][PubMed]


Young T, Finn L, Peppard PE, et al. Sleep disordered breathing and mortality: eighteen-year follow-up of the wisconsin sleep cohort. Sleep. 2008;31(8):1071–1078. [PubMed Central][PubMed]


Dick A, Liu H, Zwanziger J, et al. Long-term survival and healthcare utilization outcomes attributable to sepsis and pneumonia. BMC Health Serv Res. 2012;12(1):432


Diehr P, Yanez D, Ash A, Hornbrook M, Lin DY. Methods for analyzing health care utilization and costs. Annu Rev Public Health. 1999;20(1):125–144


Wang SX, Ganguli AX, Bodhani A, Medema JK, Reichmann WM, Macaulay D. Healthcare resource utilization and costs by age and joint location among osteoarthritis patients in a privately insured population. J Med Econ. 2017;20(12):1299–1306. [PubMed]


Centers for Medicare and Medicaid Services. 2015 Measure Information About the 30-Day All-Cause Hospital Readmission Measure, Calculated for the Value-Based Payment Modifier Program. Accessed August 27, 2018.


Bureau of Labor Statistics. Consumer Price Index. Accessed June 28, 2018.

11 Obstructive+Sleep+Apnea+LCD+and+PA. Accessed August 1, 2019.


Charlson ME, Charlson RE, Peterson JC, Marinopoulos SS, Briggs WM, Hollenberg JP. The Charlson comorbidity index is adapted to predict costs of chronic disease in primary care patients. J Clin Epidemiol. 2008;61(12):1234–1240. [PubMed]


Bansilal S, Castellano JM, Garrido E, et al. Assessing the impact of medication adherence on long-term cardiovascular outcomes. J Am Coll Cardiol. 2016;68(8):789–801. [PubMed]


Mullahy J. Specification and testing of some modified count data models. J Econom. 1986;33(3):341–365


Kapur V, Blough DK, Sandblom RE, et al. The medical cost of undiagnosed sleep apnea. Sleep. 1999;22(6):749–755. [PubMed]


Ronald J, Delaive K, Roos L, Manfreda J, Bahammam A, Kryger MH. Health care utilization in the 10 years prior to diagnosis in obstructive sleep apnea syndrome patients. Sleep. 1999;22(2):225–229. [PubMed]


Walter RJ, Hagedorn SI, Lettieri CJ. Impact of diagnosing and treating obstructive sleep apnea on healthcare utilization. Sleep Med. 2017;38:73–77. [PubMed]


Ayas NT, Taylor CM, Laher I. Cardiovascular consequences of obstructive sleep apnea. Curr Opin Cardiol. 2016;31(6):599–605. [PubMed]


McEvoy RD, Antic NA, Heeley E, et al. CPAP for prevention of cardiovascular events in obstructive sleep apnea. N Engl J Med. 2016;375(10):919–931. [PubMed]


Billings ME, Kapur VK. Medicare long-term CPAP coverage policy: a cost-utility analysis. J Clin Sleep Med. 2013;9(10):1023–1029. [PubMed Central][PubMed]


Naik S, Al-Halawani M, Kreinin I, Kryger M. Centers for Medicare and Medicaid Services positive airway pressure adherence criteria may limit treatment to many Medicare beneficiaries. J Clin Sleep Med. 2019;15(2):245–251. [PubMed Central][PubMed]


Weaver TE, Grunstein RR. Adherence to continuous positive airway pressure therapy: the challenge to effective treatment. Proc Am Thorac Soc. 2008;5(2):173–178. [PubMed Central][PubMed]


Osterberg L, Blaschke T. Adherence to medication. N Engl J Med. 2005;353(5):487–497. [PubMed]


Kribbs NB, Pack AI, Kline LR, et al. Objective measurement of patterns of nasal CPAP use by patients with obstructive sleep apnea. Am Rev Respir Dis. 1993;147(4):887–895. [PubMed]


Karimi M, Hedner J, Häbel H, Nerman O, Grote L. Sleep apnea related risk of motor vehicle accidents is reduced by continuous positive airway pressure: Swedish traffic accident registry data. Sleep. 2015;38(3):341–349. [PubMed Central][PubMed]


Tregear S, Reston J, Schoelles K, Phillips B. Obstructive sleep apnea and risk of motor vehicle crash: systematic review and meta-analysis. J Clin Sleep Med. 2009;5(6):573–581. [PubMed Central][PubMed]


Farrell PC, Richards G. Recognition and treatment of sleep-disordered breathing: an important component of chronic disease management. J Transl Med. 2017;15(1):114[PubMed]


Derose SF, Zhou H, Huang BZ, Manthena P, Hwang D, Shi JM. Does providing positive airway pressure for sleep apnea change health care utilization? Med Care. 2018;56(11):901–907. [PubMed]


Dimick JB, Ryan AM. Methods for evaluating changes in health care policy: the difference-in-differences approach. JAMA. 2014;312(22):2401–2402. [PubMed]


Wing C, Simon K, Bello-Gomez RA. Designing difference in difference studies: best practices for public health policy research. Annu Rev Public Health. 2018;39(1):453–469