Undiagnosed obstructive sleep apnea increases risk of hospitalization among a racially diverse group of older adults with comorbid cardiovascular disease

Abstract

Study Objectives:

Undiagnosed obstructive sleep apnea (OSA) is associated with increased risk for subsequent cardiovascular events, hospitalizations, and mortality. The primary objective of this study was to determine the association between undiagnosed OSA and subsequent hospitalizations among older adults with preexisting cardiovascular disease (CVD). A secondary objective was to determine the risk of 30-day hospital readmission associated with undiagnosed OSA among older adults with CVD.

Methods:

This was a retrospective cohort study of a 5% sample of Medicare administrative claims data for years 2006–2013. Beneficiaries aged 65 years and older diagnosed with CVD were included. Undiagnosed OSA was defined as the 12-month period prior to OSA diagnosis. A similar 12-month period among beneficiaries not diagnosed with OSA was used for the comparison group (no OSA). Our primary outcome was the first all-cause hospital admission. Among beneficiaries with a hospital admission, 30-day readmission was assessed for the first hospital admission only.

Results:

Among 142,893 beneficiaries diagnosed with CVD, 19,390 had undiagnosed OSA. Among beneficiaries with undiagnosed OSA, 9,047 (46.7%) experienced at least 1 hospitalization whereas 27,027 (21.9%) of those without OSA experienced at least 1 hospitalization. Following adjustment, undiagnosed OSA was associated with increased risk of hospitalization (odds ratio 1.82; 95% confidence interval 1.77, 1.87) relative to no OSA. Among beneficiaries with ≥ 1 hospitalization, undiagnosed OSA was associated with a smaller but significant effect in weighted models (odds ratio 1.18; 95% confidence interval 1.09, 1.27).

Conclusions:

Undiagnosed OSA was associated with significantly increased risk of hospitalization and 30-day readmissions among older adults with preexisting CVD.

Citation:

Kirk J, Wickwire EM, Somers VK, Johnson DA, Albrecht JS. Undiagnosed obstructive sleep apnea increases risk of hospitalization among a racially diverse group of older adults with comorbid cardiovascular disease. J Clin Sleep Med. 2023;19(7):1175–1181.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Obstructive sleep apnea is highly comorbid with cardiovascular disease and increases risk for numerous adverse health consequences, yet it is underdiagnosed and consequently undertreated. We evaluated the effect of undiagnosed obstructive sleep apnea on hospitalization and 30-day hospital readmission among Medicare beneficiaries hospitalized with comorbid obstructive sleep apnea and cardiovascular disease.

Study Impact: In this nationally representative sample of older adults with obstructive sleep apnea and comorbid cardiovascular disease, undiagnosed obstructive sleep apnea was associated with higher odds of hospitalization and 30-day readmission. These results highlight the importance of screening for obstructive sleep apnea among individuals hospitalized with cardiovascular disease.

INTRODUCTION

Cardiovascular disease (CVD) is the leading cause of mortality in the United States, causing 25% of all deaths annually.^1,2 Among older adults in particular CVD is also the single most common chronic condition, affecting more than 75% of adults 60 years of age and older.² CVD is associated with a significant social and economic burden, including elevated health care use with a total cost exceeding $360 billion yearly.^1–3 Obstructive sleep apnea (OSA) is a sleep-related breathing disorder that is highly comorbid with CVD, with prevalence estimates ranging from 40–60% among individuals with CVD.^4,5

OSA involves a decrease or complete cessation of airflow, resulting in fragmented and poor sleep quality, oxyhemoglobin desaturations, sympathetic activation, and intrathoracic pressure swings.⁶ OSA is associated with numerous adverse health consequences, including increased risk for CVD, exacerbation of medical and psychiatric comorbidities such as diabetes and depression, increased risk for motor vehicle collisions, and diminished quality of life.^7–15 Undiagnosed OSA is associated with a substantial economic burden annually, including direct treatment costs as well as indirect costs such as all-cause health care use, diminished workplace productivity, and increased workplace accidents.^5,16–24 Treatment of OSA with continuous positive airway pressure might reduce these risks and result in positive economic benefit.^22,25–28 Unfortunately, OSA is highly underdiagnosed among older adults and historically minoritized racial groups such as Black adults,²⁹ and consequently it is undertreated among individuals with comorbid CVD.^22,30 Indeed, estimates suggest that nearly 50% of hospitalized patients with CVD have occult undiagnosed, untreated OSA.^31,32

Prior work from our group and others has evaluated the economic impact of undiagnosed OSA among older adults, but few studies have focused on economic aspects of OSA among older adults with comorbid OSA and CVD.^24,33 Undiagnosed OSA may be of particular concern in this population due to OSA-associated increased risk for subsequent cardiovascular events, hospitalizations, and mortality. Given the scarcity of empirical data regarding economic aspects of OSA and CVD among older adults, the primary objective of this study was to determine the association between undiagnosed OSA and subsequent hospitalizations among older adults with preexisting CVD. A secondary objective was to determine the risk of 30-day hospital readmission associated with undiagnosed OSA among older adults with CVD. We hypothesized that undiagnosed OSA is associated with increased risk for (1) subsequent hospitalizations and (2) 30-day readmissions among older adults with CVD. As an exploratory analysis, we examined Black–White differences in the association between undiagnosed OSA and subsequent hospitalizations.

METHODS

Data source and study population

Data for this study were derived from a 5% random sample of Medicare administrative claims data obtained from the Centers for Medicare & Medicaid Services Chronic Conditions Data Warehouse for years 2006–2013. Eligible participants maintained 12 months of continuous Medicare coverage for Parts A, B, and D, excluding Part C (ie, Medicare Advantage), prior to the index date and 24 months of continuous coverage after the index date. The present study included older adults aged 65 years and older with preexisting CVD with or without an OSA diagnosis. This study was approved by the Institutional Review Board at the University of Maryland, Baltimore (HP-00083453).

CVD

The Chronic Conditions Data Warehouse contains information on 27 common chronic conditions based on validated algorithms using a combination of International Classification of Disease, Version 9, Clinical Modification (ICD-9-CM) codes and point-of-service information.³⁴ For each condition, the date of first diagnosis since Medicare enrollment is provided. We used the Chronic Conditions Data Warehouse to define CVD using the following conditions: ischemic heart disease (including myocardial infarction) (410.xx–414.xx), atrial fibrillation (427.31), heart failure (398.91, 402.xx, 404.xx, 428.xx), and stroke/transient ischemic attack (430.xx, 431.xx, 433.xx–436.xx, 997.02). There is an extensive literature linking OSA to an increased risk of each of these conditions.^35,36 The dates of first diagnosis were used to ascertain the presence of CVD prior to the index date.

Exposure

OSA was defined as a recorded inpatient or outpatient claim with at least 1 of the following ICD-9-CM codes: 780.51, 327.23, 780.57, and 780.53. The date of first diagnosis of OSA following a 12-month OSA-free period was the index date. Consistent with prior work, undiagnosed OSA was defined as the 12-month period leading up to the OSA diagnosis.²⁴

Next, we identified beneficiaries meeting continuous enrollment criteria but without an OSA diagnosis or any other sleep-related condition. We randomly assigned an index date to these beneficiaries such that the distribution of index dates across the study period was equal between cohorts.

Outcomes

Our primary outcome was all-cause hospitalization, defined as the first hospital admission for any reason recorded over the 12 months preceding the index date. For the secondary outcome of 30-day hospital readmission, we restricted to beneficiaries with at least 1 hospitalization and analyzed the first hospital admission during the study period.

Covariates

Demographic information was obtained from the Medicare beneficiary summary file. Race information in Medicare claims is obtained from data provided to the Social Security Administration and is self-reported as White, Black, Other, Asian, Hispanic, and North America Native. It is considered highly accurate for White and Black racial designations.^37,38 For this study, we grouped Asian and North America Native with Other. We identified common chronic conditions present at the index date from the Chronic Conditions Data Warehouse as described for CVD. We combined the 5 available cancer diagnoses (ie, breast, colorectal, endometrial, lung, and prostate) into a single cancer variable. In addition, we created a multimorbidity variable by summing the count of the following diagnoses: Alzheimer’s disease and related dementias, anemia, asthma, atrial fibrillation, benign prostatic hyperplasia, cataracts, chronic kidney disease, chronic obstructive pulmonary disease, depression, diabetes, glaucoma, heart failure, hypertension, hyperlipidemia, hypothyroid, ischemic heart disease, osteoporosis, and rheumatoid or osteoarthritis. This variable was categorized based on its distribution.

Data analysis

Demographic and clinical characteristics were summarized using means and standard deviations for continuous variables and frequencies and percentages for categorical variables. We compared the distribution of these demographic and clinical variables between exposure cohorts (ie, undiagnosed OSA vs no OSA) using chi-square goodness of fit for categorical variables and Student’s t tests for continuous variables. Among beneficiaries with at least 1 hospitalization, we displayed the distribution of hospitalizations over the 12 months prior to the index date (1, 2, 3, and ≥ 4) graphically by OSA status and tested differences between exposure groups using chi-square goodness of fit.

To test hypothesis 1 (Medicare beneficiaries with CVD and undiagnosed OSA will be at increased risk of hospitalization compared to Medicare beneficiaries with CVD alone), we modeled the unadjusted effect of undiagnosed OSA on the risk of hospitalization using log binomial regression to estimate risk ratios (RR). To balance the distribution of baseline covariates between exposure groups, we used stabilized inverse probability of treatment weights (IPTW). We evaluated the balance of covariates between weighted exposure groups using a SAS macro.³⁹ Next, we modeled the effect of undiagnosed OSA on the risk of hospitalization in the weighted sample using log binomial regression, adjusting for variables that were not balanced (mean difference > 0.1). Because of racial differences in hospitalization rates, we tested an interaction of race (Black and White) by undiagnosed OSA status in exploratory analyses.^40,41

To test hypothesis 2 (Medicare beneficiaries with CVD and undiagnosed OSA will be at increased risk of 30-day hospital readmission compared to Medicare beneficiaries with CVD alone), we restricted the analysis to beneficiaries with at least 1 hospitalization and among these used only the first hospitalization. We modeled the unadjusted and adjusted association between undiagnosed OSA and risk of 30-day hospital readmission as above. As previously, we also examined race as a potential effect modifier. All analyses were performed with SAS Studio Enterprise Edition (V.3.71, SAS Institute, Cary, North Carolina).

RESULTS

Sample characteristics

Our final sample included 142,893 Medicare beneficiaries with preexisting CVD (Table 1). On average participants were 78.5 years of age (standard deviation = 5.8), female (62.1%), and non-Hispanic White individuals (84.3%). The most common form of CVD was ischemic heart disease (91.6%), followed by heart failure (57.5%), atrial fibrillation (33.2%), and stroke and transient ischemic attack (29.7%).

Table 1.

Characteristics of Medicare beneficiaries ≥ 65 years of age with cardiovascular disease by undiagnosed OSA status between 2007 and 2013 from a 5% sample of Medicare claims at index date (n = 142,893).

Variables	Undiagnosed OSA (n = 19,390)	No OSA (n = 123,503)	P
Age in years, mean (SD)	75.4 (6.6)	79.0 (7.6)	<.001
Sex, n (%)			<.001
Male	16,660 (46.4)	103,849 (36.6)
Female	10,396 (53.6)	78,339 (63.4)
Race/ethnicity, n (%)			<.001
Non-Hispanic White	16,660 (85.9)	103,849 (84.1)
Black	1,586 (8.2)	11,234 (9.1)
Hispanic	575 (3.0)	2,913 (2.4)
Other	569 (2.9)	5,507 (4.5)
Cardiovascular disease type, n (%)
Atrial fibrillation	7,865 (40.6)	39,521 (32.0)	<.001
Heart failure	13,465 (69.44)	68,673 (55.6)	<.001
Ischemic heart disease	18,134 (93.52)	112,691 (91.3)	<.001
Stroke/transient ischemic	6,265 (32.31)	36,115 (29.2)	<.001
Comorbidities, n (%)
Alzheimer’s disease	976 (5.0)	8,672 (7.0)	<.001
All cancers	15,178 (21.7)	97,277 (21.2)	.12
Anemia	11,752 (60.6)	70,166 (56.8)	<.001
Asthma	4,190 (21.6)	12,937 (10.5)	<.001
Cataracts	14,071 (72.6)	92,688 (75.1)	<.001
Chronic kidney disease	5,783 (29.8)	25,296 (20.5)	<.001
Chronic obstructive pulmonary disease	8,439 (43.5)	34,113 (27.6)	<.001
Depression	6,596 (34.0)	24,688 (20.0)	<.001
Diabetes	9,202 (47.5)	74,098 (60.0)	<.001
Glaucoma	4,677 (24.1)	31,350 (25.4)	<.001
Hip fracture	567 (2.9)	5,947 (4.8)	<.001
Hyperlipidemia	17,064 (88.0)	100,846 (81.7)	<.001
Hyperplasia (benign prostatic)	4,366 (22.5)	19,794 (16.0)	<.001
Hypertension	18,167 (93.7)	110,265 (89.3)	<.001
Hypothyroidism	5,317 (27.4)	30,345 (24.6)	<.001
Osteoporosis	6,742 (34.8)	48,761 (39.5)	<.001
Rheumatoid/osteoarthritis	12,690 (65.5)	68,977 (55.9)	<.001
Count of comorbidities, n (%)
≤ 6	3,581 (18.5)	37,055 (30.0)	<.001
7–8	4,427 (22.8)	34,984 (28.3)
9–10	5,194 (26.8)	29,819 (24.1)
> 10	6,188 (31.9)	123,503 (17.5)

OSA = obstructive sleep apnea, SD = standard deviation.

There were 19,390 beneficiaries with undiagnosed OSA and 123,503 without OSA. Relative to the no OSA group, individuals in the undiagnosed OSA group were younger (75.4 [standard deviation 6.6] years vs 79.0 [standard deviation 7.6] years, P < .001) and more likely to be male (46.4% vs 36.6%; P < .001) (Table 1). Beneficiaries with undiagnosed OSA were more likely to be White (85.9% vs 84.1%; P < .001) and less likely to be Black (8.2% vs 9.1%; P < .001) compared to those with no OSA. The undiagnosed OSA group had higher prevalence of all CVD variables (P < .001 for all comparisons) than the no OSA group. The undiagnosed OSA group was also more likely to have > 10 comorbidities (6,188 [31.9%] vs 123,503 [17.5%]; P < .001) compared to the no OSA group. For example, they were more likely to have anemia (60.6% vs 56.8%), chronic kidney disease (29.8% vs 20.5%), chronic obstructive pulmonary disease (43.5% vs 27.6%), and depression (34.2% vs 20.0%) (P < .001 for all).

Hospitalizations

Among the 142,893 beneficiaries in our study sample, 36,074 (25.2%) experienced at least 1 hospitalization 12 months before the index date. Of these, 12,611 (35.0%) experienced more than 1 hospitalization (Figure 1). Among beneficiaries with undiagnosed OSA, 9,047 (46.7%) experienced at least 1 hospitalization whereas 27,027 (21.9%) of those without OSA experienced at least 1 hospitalization. Beneficiaries with undiagnosed OSA were more likely to experience 4 or more hospitalizations (9.4% vs 4.5%; P < .001).

Figure 1. Hospitalizations by OSA status among Medicare beneficiaries ≥ 65 years of age with cardiovascular disease and at least 1 hospitalization over 12 months prior to diagnosis, n = 36,074.

OSA = obstructive sleep apnea.

In the unadjusted model, the likelihood of hospital admission was significantly greater among beneficiaries with undiagnosed OSA (RR 2.13; 95% confidence interval [CI] 2.09, 2.17) relative to those with no OSA (Table 2). Following adjustment with stabilized IPTW, the association was attenuated but strong (RR 1.86; 95% CI 1.83, 1.90). Balance was achieved in the IPTW for all variables except diabetes. Adding this variable to the IPTW model resulted in a small change to the effect estimate (RR 1.85; 95% CI 1.82, 1.89).

Table 2.

Association between undiagnosed OSA and hospitalization among Medicare beneficiaries ≥ 65 years of age with cardiovascular disease over 12 months before OSA diagnosis (n = 142,893).

	Risk Ratio (95% Confidence Interval)
	Unadjusted	Stabilized IPTW	Stabilized IPTW Plus Diabetes
No OSA	Reference	Reference	Reference
Undiagnosed OSA	2.13 (2.09, 2.17)	1.86 (1.83, 1.90)	1.85 (1.82, 1.89)

IPTW = inverse probability of treatment weighted, OSA = obstructive sleep apnea.

When we restricted our sample to include only Black and White beneficiaries, 46.5% of White beneficiaries with undiagnosed OSA experienced at least 1 hospitalization whereas 54.8% of Black beneficiaries experienced at least 1 hospitalization. Following adjustment using IPTW, there were small differences in likelihood of hospitalization as a function of undiagnosed OSA status by Black (RR 2.01; 95% CI 1.90, 2.12) vs White (RR 1.87; 95% CI 1.83, 1.91) race (P = .01 for interaction).

Thirty-day readmissions

Among beneficiaries with at least 1 hospitalization (n = 36,074), 3,348 (9.3%) experienced a 30-day hospital readmission. Among beneficiaries with undiagnosed OSA, 1,001 (11.1%) experienced a 30-day hospital readmission while 2,347 (8.7%) of those without OSA experienced a 30-day hospital readmission. In the unadjusted model, beneficiaries with undiagnosed OSA were more likely to experience a 30-day hospital readmission (RR 1.27; 95% CI 1.19, 1.27) relative to those with no OSA (Table 3). No variables were unbalanced in this model; thus, following adjustment with stabilized IPTW, we still observed greater likelihood of 30-day hospital readmission among beneficiaries with undiagnosed OSA (RR 1.16; 95% CI 1.08, 1.24).

Table 3.

Association between undiagnosed OSA and 30-day hospital readmission among Medicare beneficiaries ≥ 65 years of age with cardiovascular disease over 12 months before OSA diagnosis and at least 1 hospitalization (n = 36,074).

	Risk Ratio (95% Confidence Interval)
	Unadjusted	Stabilized IPTW
No OSA	Reference	Reference
Undiagnosed OSA	1.27 (1.19, 1.37)	1.16 (1.08, 1.24)

IPTW = inverse probability of treatment weighted, OSA = obstructive sleep apnea.

When we restricted our sample to Black and White beneficiaries who experienced at least 1 hospitalization, 10.7% of White beneficiaries with undiagnosed OSA experienced a 30-day hospital readmission, whereas 12.3% of Black beneficiaries experienced a 30-day hospital readmission. In adjusted models, undiagnosed OSA was significantly associated with 30-day readmission among White beneficiaries (RR 1.17; 95% CI 1.08, 1.27), whereas the association was not significant among Black beneficiaries (RR 0.95; 95% CI 0.78, 1.16) (P = .05 for interaction).

DISCUSSION

In this large, nationally representative study of Medicare beneficiaries with CVD, undiagnosed OSA was associated with a significantly increased risk of both hospitalization and 30-day hospital readmission. This finding is consistent with and builds upon prior work suggesting that undiagnosed OSA is associated with increased health care use,²⁴ expanding these findings to older adults with comorbid CVD. This population is recognized as vulnerable and already at higher risk of subsequent cardiovascular events and hospitalization.³ Further, not only were beneficiaries with undiagnosed OSA at higher risk of hospitalization relative to those without OSA, they were also more than twice as likely to experience 4 or more hospitalizations during a single year. From a health system perspective, the high level of hospitalizations observed in this study represents a heavy clinical, financial, and health system burden. From the patient’s perspective, repeated hospitalizations also represent a major impediment to quality of life. Notably, while a significant association between undiagnosed OSA and hospitalization was observed among both Black and White beneficiaries with comorbid CVD, the association between undiagnosed OSA and 30-day hospital readmission was present among White beneficiaries only. In this study and consistent with prior literature, Black beneficiaries with comorbid CVD were more likely to be hospitalized, regardless of OSA status.^42,43 In our study, 25.5% of Black beneficiaries without OSA were hospitalized compared to only 21.7% of White beneficiaries without OSA. Increased risk of hospitalization among Black beneficiaries may be related to greater severity of CVD or lack of access to adequate preventative care.^43–45 Nonetheless, higher overall risk of hospitalization, especially among those hospitalized previously, may have mitigated the impact of undiagnosed OSA.

Given the Centers for Medicare & Medicaid Services’ goals for reducing costly hospitalizations and readmissions,^46,47 screening for and appropriately treating OSA might represent one approach to improving patient outcomes while lowering clinical and economic burden. Preliminary studies have shown that screening for OSA in hospitalized patients with congestive heart failure or obesity leads to improved diagnosis and treatment in-hospital and after discharge.^31,48,49 Furthermore, among older adults, treatment of OSA with continuous positive airway pressure is associated with reduced risk for new cardiovascular events and inpatient use.^25–27

Our study has strengths. It is the first to evaluate the impact of undiagnosed OSA on hospitalizations among older adults with comorbid CVD. These findings add to a growing body of evidence supporting the need for increased OSA screening among older adults with CVD, particularly during hospitalizations. As well, use of IPTW is a robust method to minimize confounding.

At the same time, our findings must be interpreted in light of several limitations. An important limitation is that relative to unhospitalized individuals, older adults hospitalized with CVD could be more likely to be referred for OSA diagnostic testing and thus to receive an OSA diagnosis. The effect of this bias would be away from the null, suggesting increased risk among those with undiagnosed OSA. However, the second analysis we performed (30-day hospital readmission), restricted to beneficiaries who experienced at least 1 hospitalization, eliminates this bias. Even among this hospitalized cohort, undiagnosed OSA was associated with a significantly increased risk of 30-day readmission, albeit with a somewhat attenuated effect size. Thus, we are confident in our conclusion that undiagnosed OSA increases risk of hospitalization among older adults with CVD. Next, it is possible that some beneficiaries in the undiagnosed OSA group were sent for sleep apnea testing and did not have OSA. This would have a “diluting” effect, biasing results toward the null. Third, our dataset did not enable us to determine reasons for hospitalizations in this study. Fourth, the generalizability of our findings to sicker individuals is unknown. Our continuous enrollment criteria, especially the required 24 months postindex, restricted our data to somewhat healthier beneficiaries, because participants in this study could not have died during follow-up. Also, these data are now over 10 years old. Reduction of hospitalizations for ambulatory-care sensitive conditions has been a priority for the Centers for Medicare & Medicaid Services^46,47 and rates of hospitalizations have been tracking downward over the last decade.⁵⁰ Nevertheless, we are not aware of any initiatives that would have differentially targeted Medicare beneficiaries with undiagnosed OSA. Thus, while the absolute rate of hospitalizations and 30-day hospital admissions may be lower currently than it was during our study period, there is no reason to believe that undiagnosed OSA has a different impact on these outcomes than it did in our study population. Finally, IPTW were created based on measured characteristics. Thus, we could not rule out residual confounding by other, unmeasured variables.

In conclusion, undiagnosed OSA was associated with significantly increased risk of hospitalization and 30-day readmissions among older adults with preexisting CVD. These data add to a growing body of literature highlighting the risks of undiagnosed OSA for numerous adverse CVD clinical and economic outcomes for older adults. Especially given the aging US populace, clinicians should be encouraged to screen for OSA in patients with CVD. Finally, future research should examine the impact of OSA treatments in this population.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. Work for this study was performed at the University of Maryland School of Medicine, Baltimore, Maryland. This work is supported by an investigator-initiated grant and a diversity supplement from the American Academy of Sleep Medicine Foundation (J.S.A., Principal Investigator). E.M.W. and J.S.A.’s institution has received research funding from the Department of Defense, Merck, ResMed, and the ResMed Foundation. E.M.W. served as a scientific consultant for DayZz, Eisai, EnsoData, Idorsia, Merck, Primasun, Purdue, and ResMed and is an equity shareholder in WellTap. V.K.S. has served as a consultant for ResMed, Jazz, Apnimed, Zoll, Huxley, Bayer, Lilly, and Wesper and on the Scientific Advisory Board of Sleep Number and receives funding from a grant from Sleep Number to the Mayo Clinic. J.K. and D.A.J. report no conflicts of interest.

ABBREVIATIONS

CI

confidence interval

CVD

cardiovascular disease

ICD

International Classification of Disease

IPTW

inverse probability of treatment weights

OSA

obstructive sleep apnea

RR

risk ratio