Abstract
Obstructive sleep apnoea (OSA) is a highly prevalent condition but studies exploring the burden of OSA-associated comorbidities have been limited by small sample sizes with underrepresentation of women.
We queried the Truven Health MarketScan Research Databases 2003–2012, which is a collection of health insurance claims for working adults and retirees with employer-sponsored health insurance. Adults with a diagnostic code for OSA with at least 12 months of follow-up from the index date of OSA diagnosis were compared to a matched random sample. Comorbidities were assessed using International Classification of Diseases, Ninth Edition, codes. A logistic regression model was constructed to test the independent association between OSA and comorbidities.
Our cohort included 1 704 905 patients with OSA and 1 704 417 matched controls. All comorbidities were significantly more prevalent in OSA patients. Type 2 diabetes and ischaemic heart disease were more prevalent in men but hypertension and depression were more prevalent in women with OSA. In contrast, the sex differences in the prevalence of congestive heart failure, arrhythmias and stroke were less pronounced. The prevalence of comorbidities increased with age but the effect of age varied based on the specific comorbidity. The divergence between OSA and controls was more pronounced after the sixth decade of life for most cardiovascular diseases (i.e. heart failure, ischaemic heart disease, stroke and arrhythmias), while depression exhibited an opposite trend. In a fully adjusted model, the odds of all comorbidities were significantly increased in OSA patients.
In a large, nationally representative sample of working and retired people, OSA is strongly associated with significant comorbidities in both men and women with unique sex differences emerging.
Abstract
OSA is associated with comorbidities in both men and women with a unique pattern of sex and age differences http://ow.ly/V4Rg6
Introduction
The prevalence of obstructive sleep apnoea (OSA) has been increasing in parallel with the obesity epidemic. Earlier population-based studies reported a prevalence of moderate to severe OSA of 4–7% in women and 9–14% in men [1–5]. A more recent study that applied current diagnostic definitions used in clinical sleep laboratories reported a substantially higher prevalence of OSA. In this population-based study from Switzerland, 23% of women and 49% of men were found to have moderate to severe OSA [6].
OSA is a clinically relevant chronic condition characterised by recurrent episodes of complete (apnoea) or partial (hypopnea) obstruction of the upper airway causing intermittent hypoxaemia and hypercapnia, cortical microarousals, increased oxidative stress, and sleep fragmentation [1, 5]. These adverse effects of OSA are important mediators of cardiovascular, neurocognitive and metabolic risk [7–14]. OSA has been independently associated with type 2 diabetes [14], hypertension [15, 16], coronary heart disease, stroke and cardiovascular mortality [17–21]. Moreover, OSA is a well-recognized risk factor for motor vehicle and work-related accidents, and when untreated, it has been associated with overall increased health morbidity, and cognitive and mood deficits impairing work efficiency and productivity [22, 23].
Prior studies exploring healthcare expenditures and overall burden of comorbidities associated with OSA have included relatively small, single-centre clinical cohorts subject to selection and referral bias [24–28]. There is a paucity of nationally representative cohorts exploring the burden of comorbidities associated with OSA. Jennum and co-workers [29, 30] explored the Danish National Patient Registry, and reported a significantly elevated comorbidity burden and healthcare resource utilisation in patients with OSA compared to controls. However, the generalisability of the findings may be limited, given that there were only 19 438 patients with a diagnosis of OSA during an 8-year period in this administrative database. Moreover, women were underrepresented in this cohort. A more recent study from Veterans Administration hospitals and clinics in the USA reported a significant association between OSA and cardiovascular comorbidities in 21 674 patients with untreated OSA [31]. This study was limited by the fact that 96% of patients with OSA were men. Similarly, a data mining approach on >60 000 individuals with OSA uncovered a distinct occurrence peak around 55–59 years of age, and multiple comorbid conditions involving the cardiovascular, endocrine/metabolic and respiratory systems, along with increased use of healthcare services [32].
Our primary objective was to delineate more accurately the burden of comorbidities associated with a diagnosis of OSA in a very large, nationwide health insurance database for working adults and retirees with employer-sponsored health insurance coverage in the USA. We also aimed to explore sex differences as well as the potential effect of age in the burden of comorbidities associated with OSA.
Methods
Data source
The Truven Health MarketScan Research Databases are a collection of health insurance claims for working adults and retirees with employer-sponsored health insurance. For the current project, we used the MarketScan Commercial Claims and Encounters database constructed from paid claims for employee-sponsored health insurance from 2003 to 2012, representing 17 million enrolees in 2003 to 57 million enrolees in 2012. In this study, we used the inpatient service, outpatient service, inpatient admission, facility header and enrolment tables. Within these tables, we used insurance claim records from January 1, 2003, to December 31, 2012.
The study was exempt from ethical review given that the database is de-identified and publically available.
Patient selection
Between January 2003 and December 2012, there were 77.8 million individuals aged 30–94 years in the Truven Marketscan database. We chose to enrol patients starting at age 30 years based on previously published population-based studies that also enrolled individuals that were 30 years and older [1, 6]. We queried the database for patients with International Classification of Diseases, Ninth Edition – Clinical Manifestations (ICD-9-CM) diagnosis codes for OSA (table 1). A total of 5 166 379 million patients had a diagnosis of OSA. We excluded 67% of patients for not having insurance enrolment for 1 year prior to and 1 year following the date of OSA diagnosis. Therefore, a total of 1 704 905 OSA patients were included in the analysis.
Sample characteristics for patients diagnosed with obstructive sleep apnoea (OSA) and matched random samples
A matched random sample was selected from enrolees of the same sex within the same 5-year age group and state of residence (n=1 704 417). Potential matches had at least 24 months of enrolment and did not have any ICD-9-CM codes for inpatient or outpatient diagnoses of OSA, charges for polysomnograms or out-of-centre sleep testing, or charges for continuous positive airway pressure at any time during their enrolment in MarketScan. We did not exclude enrolees who did not have any contact with the healthcare system (i.e. no inpatient or outpatient claims). For example, if an individual had no claims for either OSA or any comorbidity, they could have been selected as a control.
Covariates
Age in years was ascertained on the date of index OSA diagnosis or the age 1 year into the enrolment period for the matched random sample. Age was classified into 10-year age groups from 30 to 79 years, and >79–94 years. Comorbidities were assessed using the inpatient and outpatient service, inpatient admission, and facility header tables. For OSA cases, we aimed to minimise the risk of reporting bias (i.e. new diagnosis of OSA leading the clinicians to diagnose additional comorbidities) by querying for diagnosis of comorbidities for the 6 months prior to the index diagnosis of OSA being established. In other words, all comorbidities reported in our analyses were established before the diagnosis of OSA. Comorbidity data was extracted from months 7 to 12 of enrolment for the matched random sample. Patients were considered to have comorbidity if an associated ICD-9-CM code was found at least once in the data tables. Using Elixhauser comorbidity classifications, we constructed six comorbidity categories (table 2 and supplementary table S1). In addition, we classified individuals as having morbid obesity, hypertension, type 2 diabetes, ischaemic heart disease, congestive heart failure, stroke, arrhythmias or depression using ICD-9-CM codes in table S2. To illustrate sex differences in the prevalence of comorbidities between patients with OSA and matched controls, age was classified into 5-year groups.
Groups of Elixhauser comorbidities in patients diagnosed with obstructive sleep apnoea (OSA) and matched random samples
Statistical analysis
We summarised the distribution of demographics and comorbidities for OSA cases, and the matched random sample using descriptive analyses. We performed Chi-squared tests for bivariate analyses of categorical variables. We used two-sided tests, with p-values <0.05 considered to be statistically significant. Prevalence of OSA-related comorbidities by age group, sex and group (i.e. OSA or control) were determined. Logistic regression models tested the odds ratios of having OSA-related comorbidities after adjusting for age group and sex. All statistical analyses were performed with SAS 9.3 (SAS Institute, Cary, NC, USA).
Results
Our analytic cohort included 1 704 905 patients carrying an ICD-9-CM code diagnosis for OSA. Our controls included 1 704 417 without a diagnosis of OSA matched for age, sex and state of residence (table 1). The comorbidity characteristics based on Elixhauser comorbidity categories are shown in table 2. We also used ICD-9-CM codes as described in table S2 to characterise better the prevalence of OSA-specific comorbidities (table 3). All comorbidities explored were significantly more prevalent in patients carrying a diagnosis of OSA when compared to matched controls.
Specific comorbidities in patients diagnosed with obstructive sleep apnoea (OSA) and matched random samples
Figure 1 illustrates the effect of age and sex on the prevalence of comorbidities in patients with OSA and matched controls. All comorbidities were more prevalent in patients with OSA regardless of sex when compared to matched controls, especially with increasing age. However, the prevalence of type 2 diabetes and ischaemic heart disease was higher in men with OSA compared to women with OSA. In contrast, hypertension and depression were more prevalent in women with OSA. Sex differences in the prevalence of arrhythmias and stroke in patients with OSA were much less pronounced. Type 2 diabetes, hypertension and depression were significantly more prevalent in patients with OSA compared to matched controls at all age groups.
Prevalence of a) type 2 diabetes, b) hypertension, c) arrhythmias, d) depression, e) ischaemic heart disease, f) congestive heart failure and g) stroke by age group and sex in patients diagnosed with obstructive sleep apnoea and the matched random sample.
A logistic regression model was constructed to estimate the odds of having comorbidities after adjusting for age and sex (table 4). In this model, the odds of all comorbidities were significantly increased in patients designated as having a diagnosis of OSA. Due to the nature of the administrative database, we lacked information on important confounders (i.e. body mass index, race/ethnicity, diet, physical activity and socioeconomic status) and could therefore not adjust for these covariates.
Odds ratios of having comorbidities adjusted for age and sex
Discussion
In this very large, nationwide study of health insurance claims for working adults and retirees with employer-sponsored health insurance, we found that the prevalence of important comorbidities is significantly higher in patients with a diagnosis of OSA compared to matched controls. In general, the prevalence of comorbidities increased with ageing in the matched controls, and such trends were ubiquitous among OSA-diagnosed patients as well. However, the increase in prevalence with ageing was amplified by the presence of OSA. Moreover, the joint effects of age and OSA varied based on the comorbidity: the divergence between OSA and controls was more pronounced after the sixth decade of life for most cardiovascular diseases (i.e. heart failure, ischaemic heart disease, stroke and arrhythmias), while depression exhibited an opposite trend. Sex-related differences were also apparent, as hypertension and depression were more prevalent in women with OSA. In contrast, the prevalence of type 2 diabetes and ischaemic heart disease was higher in men with OSA. Notwithstanding the sex and age influences on OSA-imposed morbidity odds, OSA was independently associated with all comorbidities examined in this very large cohort after fully adjusting for confounders. Thus, our findings not only suggest that the risk of comorbidities is increased by OSA but also that this effect also varies by sex and age.
The uniqueness of the present study resides in its inclusion of the largest sample of patients with a diagnosis of OSA reported to date, which also incorporates a robust representation of women as well as a wide age spectrum. The large sample size enabled us to explore sex differences and the effect of age in the context of OSA diagnosis and prevalent comorbidities. Our findings are concordant with prior smaller studies reporting a higher prevalence of important comorbidities in patients with OSA [24–26, 29–31]. Such an elevated comorbidity burden has far reaching implications given the increasing prevalence of OSA in the community [5, 6].
Before we discuss the implications of our findings, several methodological limitations deserve comment. Our main limitation is inherent to the analysis of a large administrative database. The diagnosis of OSA is based solely on clinicians' use of ICD-9-CM code for billing purposes, such that we cannot exclude the possibility of coding errors or reporting biases. Although we recognise that patients categorised as controls may have unrecognised or undiagnosed OSA, this would actually serve to bias towards the null, causing our reported effect size to be underestimated. However, it is also possible that the increased prevalence of comorbidities led to increased rates of screening for OSA and, therefore, overestimating the association between OSA and comorbidities. We also minimised the risk of reporting bias (i.e. new diagnosis of OSA leading the clinicians to diagnose additional comorbidities) by querying for diagnosis of comorbidities for the 6 months prior to the index diagnosis of OSA being established. To further exclude the possibility of reporting bias, we also compared the prevalence of a medical condition that is not associated with OSA and/or obesity in both groups. The prevalence of hepatitis C infection was not different between OSA and controls (0.11% versus 0.12%, respectively; p=0.24) during the 6 months of the comorbidity assessment. This finding suggests that there was no systematic bias in reporting more comorbidities in patients with OSA.
Low wage and part-time workers were underrepresented in our sample, and the database does not include unemployed individuals. Moreover, the MarketScan database does not include important demographic data that have well-established associations with the prevalence of comorbidities such as body mass index, race/ethnicity, diet, exercise habits, health literacy and socioeconomic status. This is an important limitation since obesity is a well-recognised risk factor for many of the comorbidities explored in this study as well as a risk factor for OSA. We attempted to partially overcome this limitation by matching for geographical location to the level of state of residence. Nonetheless, we acknowledge that race and obesity are important confounding variables, and their absence constitutes an important limitation, particularly when it applies to our ability to select matched controls. Additional missing information from this national database is the categorical classification of OSA severity, and estimates of adherence to a variety of treatment regimens. Lastly, although our cross-sectional analysis demonstrates a robust association between OSA and prevalent comorbidities, this finding by no means infers causality. However, there is ample rationale to support a biologically plausible link between OSA and increased prevalence of comorbidities, as assessed in our analysis.
Our limitations notwithstanding, this study has several noteworthy strengths. To the best of our knowledge, our sample size is the largest reported thus far, thereby lending more congruency to any of the findings. We purposefully selected equal proportion of men and women with OSA and matched controls to overcome the important limitation of prior studies in which women were significantly underrepresented. This strategy enabled us to perform meaningful comparisons between men and women across various age groups. Accordingly, some of the novel findings indicate that sex has differential effects on the type of comorbidity–OSA interactions, with ischaemic heart disease, for example, being more dominantly represented among male patients, while hypertension and depression show more prominent associations among female patients. The emerging awareness of the prevalence of OSA among women and its potential health-related and sex-related morbid consequences will likely lead to increasing research efforts in this area, such as to tailor more precisely sex-based interventions and outcome measures [33]. Our large sample size also enabled us to better explore the burden of comorbidities according to their distribution patterns among different age groups. When compared to matched controls, the effect of OSA on prevalent cardiovascular comorbidities became apparent around the age of 60–70 years in both sexes, and continued to increase with further ageing. Conversely, the presence of depression showed a different temporal trajectory with aging both in men and in women.
Several studies have shown that patients with OSA will present with increased risk for comorbidities starting several years before a formal diagnosis of OSA is made [25, 27, 28, 30]. This increased burden of comorbidities in patients with OSA leads to significant increase in healthcare resource utilisation, both in younger patients and the elderly [26, 34]. Women with OSA may also be more frequent users of healthcare resources than men [24]. Although the preponderance of evidence to date suggests that there is a high cardiometabolic and neurocognitive comorbidity burden associated with OSA, it remains unclear whether earlier identification and treatment of OSA leads to a reduction in the incidence or severity of these comorbidities. To shed light on this challenging and clinically relevant question would require enormous resources such as to perform adequately powered randomised controlled trials. Given the paucity of well-designed clinical trials, analysis of large administrative databases can provide useful real-life data to assist in the design and implementation of future studies [35].
In summary, using a national database that included the largest sample size of men and women with OSA and matched controls to date, the diagnosis of OSA was independently associated with significantly elevated burden of comorbidities. There are important sex differences in the comorbidities explored and the effect of age varies based on the comorbidity. Although early diagnosis and treatment of OSA appears justified due to the multifaceted end-organ morbidity burden imposed by this condition, further well-controlled clinical trials are needed to ascertain whether early identification and treatment of OSA decreases the incidence or severity of comorbidities.
Acknowledgements
We thank F. Javier Nieto from the University of Wisconsin at Madison (Madison, WI, USA) for helpful discussions.
B. Mokhlesi and D. Gozal designed the study and the analysis. B. Mokhlesi, S.A. Ham and D. Gozal performed all analyses. D. Gozal obtained study funding. B. Mokhlesi, S.A. Ham and D. Gozal participated in data interpretation and manuscript preparation. B. Mokhlesi and D. Gozal oversaw all analyses and take full responsibility for the work as a whole, including the study design, access to data, and the decision to submit and publish the manuscript. All authors approved this manuscript in its final form.
Footnotes
This article has supplementary material available from erj.ersjournals.com
Support statement: This study was supported in part by the Herbert T. Abelson Chair in Pediatrics to D. Gozal. Funding information for this article has been deposited with FundRef.
Conflict of interest: Disclosures can be found alongside the online version of this article at erj.ersjournals.com
- Received September 29, 2015.
- Accepted November 14, 2015.
- Copyright ©ERS 2016
References
