Introduction

Obstructive sleep apnea (OSA) is a common, chronic sleep-related breathing disorder characterized by repeated narrowing and obstruction of the upper airway during sleep, resulting in repeated episodes of decreased or completely halted airflow, intermittent oxygen desaturation, and sleep disruption [1, 2]. In the United States (US), an estimated 163.2 million adults aged 30–69 years have mild to severe OSA and almost 24 million have moderate to severe OSA [3]. Obesity, a major modifiable risk factor for OSA [2, 4], is projected to affect half of US adults by 2030 [5]. Approximately 40% of individuals with obesity also have sleep apnea [6], and results from the Wisconsin Sleep Cohort Study indicate that 41% of OSA in adults, including 58% of moderate-to-severe OSA cases, can be attributed to overweight or obesity [7].

Both OSA and obesity are associated with substantial direct and indirect economic burden, driven by healthcare resource utilization (HCRU). OSA remains highly underdiagnosed and undertreated, with an estimated ~80% of affected individuals lacking a formal diagnosis [8]. Undiagnosed OSA alone cost the US ~$149.6 billion in 2015, driven by workplace and motor vehicle accidents, comorbid disease management, and lost productivity [8]. Similarly, the economic and social cost of obesity in the US rose from $0.976 trillion in 2014 to an estimated $1.389 trillion in 2018 [9]. Adults with obesity in the US compared to those without obesity experience significantly higher annual medical care costs, which significantly increase with the severity of obesity [10]. Obesity-related complications, such as type 2 diabetes (T2D), cardiovascular diseases, and other comorbidities, further elevate healthcare resource utilization and associated costs [11].

Positive airway pressure (PAP) therapy is the first-line treatment for OSA, effectively reducing disease severity, daytime sleepiness, systemic blood pressure, and motor vehicle accidents, while significantly improving health- and sleep-related quality of life [12]. In addition, evidence from a large observational study in France demonstrated that long-term adherence to continuous PAP (CPAP) therapy is associated with lower all-cause mortality. Moreover, a recent meta-analysis confirmed the potential beneficial effects of CPAP on both all-cause and cardiovascular mortality in patients with OSA, highlighting its critical role in OSA management [13, 14].

For individuals with both OSA and overweight or obesity, clinical guidelines recommend PAP therapy in conjunction with weight management interventions [4, 15]. Although comprehensive lifestyle modifications, pharmacological interventions, and bariatric surgery can substantially improve OSA severity and related comorbidities, weight loss alone rarely resolves OSA completely [16, 17] Therefore, many patients will require continued PAP therapy even after weight reduction. Regular clinical reassessment, including repeat sleep studies, is essential to guide ongoing treatment decisions objectively [18].

Previous studies have found an association between adherence to PAP therapy and lower HCRU and costs [19]. In a general OSA population, Sterling et al. found during the first year after the start of PAP therapy, PAP adherence was significantly associated with fewer emergency room visits, all-cause hospitalizations and lower total healthcare costs, with similar results in the second year of PAP use [20]. This beneficial relationship has been observed as well, specifically among patients with OSA and comorbid conditions such as chronic obstructive pulmonary disease (COPD) [21], T2D [22], heart failure [23, 24], atrial fibrillation [25], and depression [26].

Given the substantial clinical and economic burden associated with obesity in OSA patients [27, 28], it is important to understand how PAP adherence affects healthcare outcomes in patients with OSA and comorbid obesity. This study was designed to test the hypothesis that adherence to PAP therapy would predict HCRU and costs among patients with obesity who were newly diagnosed with OSA and initiating PAP therapy in a real-world setting.

Methods

Data source

This was a retrospective cohort study conducted among patients with obesity and newly diagnosed OSA who initiated PAP therapy between January 2015 and May 2021. De-identified payer-sourced medical and pharmacy claims data from over 160 Medicare Advantage, Medicaid, and commercial health plans (Inovalon Insights LLC, Bowie, MD) were linked with objective individual patient PAP usage data collected from cloud-connected devices (AirView™; Resmed, San Diego, CA). The claims data included information about healthcare encounters, prescription fills, diagnosis and procedure codes, and the device data included detailed PAP therapy usage metrics. The databases were linked through a tokenization process and the resulting linked database underwent a third-party expert determination to ensure compliance with the Health Insurance Portability and Accountability Act. The study design was reviewed by an institutional review board (IRB) (Advarra, Ref number Pro0004005) and deemed exempt from further IRB oversight.

Study cohort

The study cohort included adults (aged ≥18 years) with obesity who completed a sleep test and subsequently received an OSA diagnosis within 60 days (International Classification of Diseases, Tenth Revision-Clinical Modification [ICD-10-CM]: G47.33), had PAP device setup (index date) within 1 year following the sleep test, used an AirSense™ 10 or 11 PAP device (Resmed, San Diego, CA), and had medical and pharmacy claims data available for ≥1 year prior to and the full 2 years after the index date. As such, all included subjects had data for the full study period, with no censoring. Obesity was identified by the presence of at least 2 healthcare encounter claims with an ICD-10 diagnosis code of obesity or morbid obesity or an ICD-10 code indicating a BMI value ≥ 30 kg/m2 in the year prior to device setup (Supplementary Table 1).

Patients were excluded if they had evidence of central sleep apnea, nocturnal hypoventilation, pregnancy, or weight reduction surgery/procedures at any time during the study period, or evidence of PAP re-supply in the year prior to the sleep test date. Patients were also excluded if they had end-stage renal disease or dialysis, which are associated with high healthcare resource use and costs [29].

PAP adherence

Adherence to PAP therapy over 2 years was the primary predictor of interest. PAP usage was objectively measured by patients’ PAP device for each night of use over the first 2 years. For reimbursement purposes, the US Centers for Medicare and Medicaid Services (CMS) considers a patient adherent to PAP therapy if the device is used ≥4 h per night on ≥70% of nights during a consecutive 30-day period during the first 90 days of initial use [30]. Three levels of adherence were evaluated in this analysis: (1) adherent patients who met CMS criteria for all 8 consecutive 90-day time frames (quarters) within the first 2 years; (2) intermediately adherent patients who met CMS criteria in at least 1 but no more than 7 quarters; and (3) nonadherent patients who did not meet CMS criteria in any of the 8 quarters.

Outcomes

The HCRU primary outcomes of interest were all-cause hospitalizations, and all-cause emergency room (ER) visits within the first and second years after PAP initiation. Healthcare costs were secondary outcomes and included inpatient hospitalizations, ER visits, and total costs (with and without apnea equipment). Proxy costs for all resource use were provided by Inovalon Insights LLC based on their proprietary Proxy Financials algorithm, which is based on CMS Medicare prospective payment system fee schedules.

Covariates

The following covariates were included to control for baseline differences among adherence groups: (1) demographic characteristics (age, region, sex, payer); (2) comorbid conditions based on ICD-10-CM diagnosis codes observed in the year prior to index (Supplementary Table 1; coronary artery disease, atrial fibrillation, pulmonary hypertension, hypertension, cerebrovascular disease, other arrhythmia, atrial flutter, heart failure, stroke, hyperlipidemia, asthma, chronic obstructive pulmonary disease, pneumonia, psychotic disorders, other mood disorders, depression, anxiety, type 1 and type 2 diabetes, hyperlipidemia, gastroesophageal reflux disease (GERD), cancer, musculoskeletal pain, psoriasis, and osteoarthritis); (3) glucagon-like peptide 1 receptor agonist (GLP-1) medication fills; (4) claim for weight management lifestyle intervention (Supplementary Table 2); and (5) pre-index all-cause hospitalizations and ER visits.

Additionally, wellness visits, flu vaccines, and adherence to beta blocker, statin, and antihypertensive medications were included as proxies to control for a possible healthy user effect [31, 32] For each medication category, among patients who filled ≥1 prescription 181–360 days before starting PAP therapy, the proportion of days covered (PDC) was measured from the medication fill date to the end of the measurement year (PDC score range: 0.0–1.00).

Statistical analyses

Baseline demographic and clinical characteristics and covariates were assessed overall, and across adherence groups using descriptive statistics. A multinomial logistic regression model was run to calculate propensity scores estimating the likelihood of being in each adherence group, based on baseline covariates. Using these scores, inverse probability of treatment weighting (IPTW) was applied to balance adherence groups on baseline covariates to control for measured confounding [33, 34]. The IPTW approach weights the sample so that the distribution of covariates is similar across adherence groups and mirrors the distribution of the overall cohort. Balance between groups was assessed using standardized mean differences (SMD), with a value < 0.1 indicating good balance [34]. To assess the association between adherence level and outcomes, pairwise differences (adherent vs intermediately adherent, adherent vs nonadherent, intermediately adherent vs nonadherent) in the mean number of healthcare visits and mean healthcare costs were assessed using weighted Wilcoxon rank-sum tests. In order to assess robustness of results to potential unmeasured confounding, e-values were reported for the adherent to nonadherent comparison. E-values denote the minimum strength of association an unmeasured confounder would need to have with both the predictor and the outcome to fully explain observed differences, beyond any measured covariates [35]. All analyses were performed using R statistical software version 4.2.2, PSWeight version 1.1.8 package [33, 36].

Results

Baseline characteristics

The final cohort included 173,691 patients with obesity and newly diagnosed OSA who initiated PAP therapy (Fig. 1). Of these, 46.6% were female, the mean age was 51.6 ± 12.0 years, and over two-thirds of patients (68.0%) had commercial health insurance (Table 1). Comorbid conditions were common, with 95.3% of patients having at least 1 comorbid condition. The mean number of comorbid conditions was 4.3 (2.7) in the overall cohort, with the most common being hypertension (experienced by 67.0% of patients), hyperlipidemia (56.3%), musculoskeletal pain (55.5%), GERD (31.0%), type 2 diabetes (30.1%), depression (26.8%), and anxiety (26.4%). Almost one-quarter of patients had a claim for a lifestyle intervention for weight management; baseline fills for GLP-1 receptor agonists for the treatment of obesity and/or type 2 diabetes was low (3.8%). During baseline, 34.7% of patients had an ER visit and 12.3% had an inpatient hospitalization for any cause.

Fig. 1: Study cohort selection flowchart.
Fig. 1: Study cohort selection flowchart.The alternative text for this image may have been generated using AI.
Full size image

BMI body mass index, CSA central sleep apnea, ESRD end stage renal disease, ICD-10 International Classification of Diseases, Tenth Revision, kg/m2 kilograms per meters squared, NH nocturnal hypoventilation, OSA obstructive sleep apnea, PAP positive airway pressure.

Table 1 Unadjusted baseline characteristics, healthcare resource use, and costs, overall and by adherence group.
Full size table

In the unadjusted cohort prior to IPTW, patients who were adherent to PAP therapy were significantly more likely than patients who were intermediately adherent or nonadherent to be male, older, and be commercially insured. Baseline healthcare resource use and costs were lower among adherent patients compared to those of intermediate and nonadherent patients. All healthy user proxy variables assessed during baseline (statin medication PDC, antihypertensive medication PDC, beta blocker medication PDC, wellness visit, flu vaccination) differed significantly by adherence group (Table 1). After applying IPTW, the adherence groups were well balanced on all baseline characteristics, including prior year HCRU, and healthy user proxies (absolute value of all SMDs <0.1; Supplementary Table 3).

PAP adherence

In the first 2 years following PAP initiation, 35.2% of patients (n = 61,171) were considered adherent to PAP therapy, 39.1% (n = 67,938) were intermediately adherent, and 25.7% (n = 44,582) were nonadherent. PAP usage overall and for each adherence group is shown in Table 2. In the overall cohort, 67.9% met the CMS criteria for PAP adherence in the first 90 days. In adherent patients, PAP usage was consistent over 2 years, averaging almost daily, full night use. Average PAP use for intermediately adherent and nonadherent patients was lower than that for adherent patients and decreased over time.

Table 2 Summary of positive airway pressure (PAP) usage in years 1 and 2, overall and by adherence group.
Full size table

PAP adherence and healthcare resource use and costs

After applying IPTW, over the first year of PAP use, patients who were adherent to PAP therapy had significantly fewer (mean ± SD) all-cause ER visits (adherent: 0.55 ± 1.77 vs. intermediate: 0.62 ± 1.68; vs. nonadherent: 0.69 ± 1.78) and hospitalizations (adherent: 0.10 ± 0.44 vs intermediate: 0.13 ± 0.53; vs. nonadherent: 0.14 ± 0.56) than intermediate or nonadherent patients (all p < 0.001). Results were similar in the second year with adherent patients continuing to have significantly fewer ER visits (adherent: 0.50 ± 1.51 vs. intermediate: 0.59 ± 1.60; vs. nonadherent: 0.65 ± 1.69) and hospitalizations (adherent: 0.09 ± 0.40 vs. intermediate: 0.13 ± 0.53; vs. nonadherent: 0.13 ± 0.56) (all p < 0.001). (Table 3, Fig. 2). Healthcare resource use for intermediately adherent patients fell between that of adherent and nonadherent patients, with significantly lower mean all-cause ER visits and hospitalizations compared to patients who were nonadherent to PAP therapy over 2 years. E-values for ER visits were 1.85 and 1.95 and for hospitalizations were 2.22 and 2.52 in the first and second years, respectively.

Fig. 2: Emergency room visits and hospitalizations before and after PAP therapy initiation.
Fig. 2: Emergency room visits and hospitalizations before and after PAP therapy initiation.The alternative text for this image may have been generated using AI.
Full size image

Annual per-person emergency room visits (A) and all-cause hospitalizations (B) in year prior to, 1st, and 2nd years of positive airway pressure (PAP) therapy after applying inverse probability treatment weighting. All comparisons between adherence groups for ER visits and all-cause hospitalizations were significant at P < 0.001 except for the comparison of intermediate adherence vs non-adherence groups on all-cause hospitalizations, which was significant at P < 0.05.

Table 3 Healthcare resource use and costs in years 1 and 2 in inverse probability of treatment weighted cohort, by adherence group.
Full size table

Total costs in year 1, were significantly lower for adherent patients than intermediately adherent or nonadherent patients (adherent: $11,332 ± $21,160 vs. intermediate: $11,947 ± $19,325 [p = 0.005]; vs. nonadherent: $11,927 ± $20,212 [p < 0.001]), with an absolute difference in mean costs of $595 for adherent vs. nonadherent patients. Results were similar when excluding sleep apnea equipment, with a difference in mean costs of $991 for adherent vs. nonadherent patients (p < 0.001). Inpatient hospitalization costs (adherent: 1384 ± 7022 vs. intermediate: 1854 ± 8902; vs. nonadherent: 2116 ± 9723) and ER visit costs (adherent: 536 ± 2236 vs. intermediate: 608 ± 1794; vs. nonadherent: 677 ± 1852) were also significantly lower (all p < 0.001).

Results in the second year were consistent. Mean total costs for adherent patients were $695 less than for nonadherent patients (10,531 ± 19,106 vs. 11,226 ± 21,706 [p < 0.001]), and $1023 less when excluding sleep apnea equipment (10,092 ± 19,056 vs. 11,115 ± 21,665 [p < 0.001]). Adherent patients also had significantly lower costs compared to intermediately adherent patients when sleep apnea equipment was excluded (10,092 ± 19,056 vs. 11,159 ± 21,319 [p < 0.001]). Inpatient hospitalization (adherent: 1239 ± 6658 vs. intermediate: 1886 ± 9118; vs. nonadherent: 1984 ± 9436) and ER visit (adherent: 503 ± 1935 vs. intermediate: 594 ± 1779; vs. nonadherent: 649 ± 1802) costs remained significantly lower for adherent compared to intermediately and nonadherent patients (all p < 0.001) (Table 3).

Discussion

In this real-world study among patients with obesity and newly diagnosed with OSA, we found that sustained adherence to PAP therapy over 2 years was associated with significantly lower HCRU, including fewer emergency room (ER) visits and hospitalizations, and substantially reduced healthcare costs, compared with patients who did not adhere to PAP therapy. While the per-patient annual savings may be modest at the individual level, ~$600–$700 with equipment in years 1 and 2, respectively, they could translate to substantial cumulative savings at the population level and may represent cost-savings opportunities if the ~80% of OSA patients who remain undiagnosed were treated. For most outcomes, patients with intermediate PAP adherence also demonstrated reductions in HCRU and healthcare costs relative to nonadherent patients, suggesting a dose-dependent relationship. This finding is consistent with previous research indicating measurable clinical and economic benefits even from modest nightly PAP usage (1–3 h of PAP use per night) [37]. The observed decline in PAP adherence over time in the intermediate adherence group underscores the need for targeted interventions, particularly within the first year of treatment, to sustain initial adherence and maximize long-term benefits [38].

Until recently, the impact of untreated or inadequately treated OSA on economic outcomes has been understudied [39, 40]. A limited number of US real-world studies, employing other datasets and methodologies, have found an association between PAP use and/or PAP adherence and lower HCRU and/or healthcare costs [39,40,41]. However these studies have several limitations including small sample sizes, lack of objective PAP adherence data, and inadequate control for confounding factors such as the healthy user bias and PAP adherence [39]. Our study addresses these limitations by leveraging objective PAP adherence data from cloud-connected devices linked to administrative claims and a bigger sample size strengthens the evidence that better adherence is associated with lower HCRU and costs.

Our findings reinforce and expand on previous analyses linking administrative claims data and objective PAP usage, which have demonstrated lower HCRU with better adherence to PAP therapy in a general population newly diagnosed with OSA [20] and in OSA populations with major pre-existing comorbid conditions, including COPD [21], T2D [22], heart failure [23, 24], atrial fibrillation [25], and depression [26]. Across these populations, increased adherence to PAP therapy was consistently and significantly associated with fewer all-cause ER visits and hospitalizations and lower healthcare costs in the first and second years of PAP use. In analyses that included patients with intermediate adherence [23,24,25,26], there was evidence of a dose-dependent relationship, with intermediate patients generally having worse outcomes than adherent patients but better outcomes than nonadherent patients, a pattern consistent with the results of the present analysis which further emphasizes the clinical importance of promoting even modest improvements in adherence.

Lower HCRU and costs among patients who are adherent to PAP therapy may be explained at least in part by improvements in the comorbid conditions associated with OSA and obesity. The combination of untreated OSA and obesity increases the risk of certain conditions (e.g., hypertension, cardiovascular and cerebrovascular disease, atrial fibrillation, T2D, and dyslipidemia) and complications from those conditions. Improvement in symptoms or the progression of comorbid conditions due to treatment for OSA may be reflected in lower HCRU and costs. For example, data from recent meta-analyses have shown that OSA is an independent risk factor for prediabetes and T2D [42], and that treatment with CPAP improves hemoglobin A1c and long-term glycemic control in patients with T2D [43]. Since poor glycemic control has been associated with high direct medical costs of treating T2D [44], the improvements in glycemic control with PAP adherence [42] may partly explain the lower costs observed in our study.

Our findings should be interpreted considering several limitations inherent to observational studies using administrative claims data [45]. First, claims data are primarily generated for billing and reimbursement of purposes and lack detailed clinical information such as patient-reported outcomes, laboratory results, and comprehensive data on social determinants of health. Future studies integrating electronic medical records or patient-reported outcomes data could address these gaps and provide deeper insights. Second, although obesity identification through ICD codes is highly specific, obesity is known to be underreported in claims databases [46]. As a result, our study may not have captured all individuals with obesity, potentially influencing generalizability. However, the high specificity (>97%) of obesity ICD codes provides confidence in the accurate characterization of our sample [46]. Third, the findings represent insured individuals within the US healthcare system, which limits the generalizability to uninsured populations or international contexts. Additionally, healthcare utilization not captured by claims (e.g., out-of-pocket or non-reimbursable services) was not assessed, potentially underestimating true healthcare resource use. Fourth, healthcare cost estimates were derived from Inovalon’s proprietary Proxy Financials algorithm. While this method provides standardized estimates, it may not reflect actual paid amounts and could limit the precision of cost-related results. Observational research inherently carries a risk of healthy user bias—patients adherent to PAP therapy may also engage in other health-promoting behaviors [47, 48]. To control for any potential healthy user bias, we included several proxy variables for healthy user behavior as covariates in our analysis, all of which were well-balanced after IPTW adjustment. Furthermore, based on the calculated E-values, unmeasured confounders, including a healthy user effect, would have to be associated with both PAP adherence and outcomes by at least a factor of ~2 (E-values of 1.85 and 1.95 for ER visits and 2.22 and 2.52 for hospitalizations in the first and second year, respectively) to explain fully the observed effect. This suggests our findings are relatively robust to unmeasured confounding. As this association would have to be beyond any associations already captured, this degree of residual confounding is unlikely given the extensive set of covariates included in this analysis. Finally, patients with Medicare fee-for-service coverage and patients using PAP devices manufactured by other companies were not included in our data set, thereby possibly further limiting the generalizability of these findings.

Despite these limitations, our study has several strengths. Importantly, the use of administrative claims data linked to objective PAP data from cloud-connected devices. Objective PAP use data to classify adherence groups has been shown to perform better in terms of accuracy than claims-based definitions of PAP adherence [49]. The linkage of objective adherence data with comprehensive administrative claims significantly enhances the validity and clinical relevance of our findings. Our large and diverse study cohort in terms of geographic region, payer type, and patient characteristics, enhances the generalizability of findings across the US healthcare landscape. Additionally, our robust analytical approach using IPTW created well-balanced groups and minimized the effects of measured confounding without loss of sample size or statistical power. Use of IPTW also allowed for the analysis of more than 2 groups; including the intermediate adherence group increased the generalizability of the results and better characterized real-world adherence patterns [20].

Conclusion

In this large real-world cohort of patients with obesity and newly diagnosed OSA, sustained adherence to PAP therapy over two years was associated with significantly lower healthcare utilization and costs. These results highlight that supporting long-term PAP adherence, could lead to significant reduction healthcare utilization and costs, and ultimately improve overall patient quality of life for patients living with obesity and OSA.