Snoring Survivors: the impact of obstructive sleep apnoea and continuous positive airway pressure use on in-hospital mortality, length of stay and costs among patients hospitalised with acute cardiovascular disease – A retrospective analysis of 2016-2019 National Inpatient Sample Data


  • The utilisation of a nationally representative dataset of about 3 million acute cardiovascular hospitalisations taking place over 4 years in the pre-COVID-19 era is one of the strengths of this study.

  • Robust outcomes—in-hospital mortality, length of stay and total hospital charges—were chosen to evaluate the impact of obstructive sleep apnoea (OSA) and continuous positive airway pressure (CPAP) use, while adjusting for a large number of potential confounders, adding to the strengths of this study.

  • Its retrospective design, the selection bias arising from 10th International Classification of Diseases, Clinical Modification and Procedure Coding System-based definitions of OSA and CPAP, and the confounding bias of unaccounted for variables are the major limitations of this study.

  • The study may have not included a sufficiently large number of unstable angina (UA) patients to detect the impact of OSA and CPAP use on UA outcomes.

  • The study also suffers from generalisability to healthcare settings outside of the USA.


Obstructive sleep apnoea (OSA) manifests as recurrent partial or complete airway obstruction, resulting in chronic intermittent hypoxaemia and autonomic dysfunction. The chronic intermittent hypoxia associated with OSA can result in hormonal and autonomic dysregulation, often manifesting as cardiac rhythm disturbances, such as atrial fibrillation, bradycardia and sudden cardiac death. Moreover, OSA is highly prevalent among patients hospitalised for acute cardiovascular disease, such as myocardial infarction, heart failure, atrial fibrillation and stroke.1

Although several studies have observed a significant association between OSA and long-term adverse cardiovascular outcomes,2–9 there are fewer studies addressing the impact of treatment with continuous positive airway pressure (CPAP) among hospitalised patients with OSA and acute cardiovascular disease.1 10 11 The largest randomised controlled trial (RCT) to date, the Sleep Apnea cardioVascular Endpoint [SAVE] trial, randomised 2717 patients with established cardiovascular disease (a minority with HF) and moderate-to-severe OSA to CPAP plus usual care and to usual care alone. CPAP use did not reduce the composite cardiovascular endpoint, although in secondary analysis there was a lower risk of cerebrovascular events among patients using CPAP for at least 4 hours per night.12 In another large single tertiary care centre experience of roughly 94 000 ward admissions, OSA patients had lower odds of inpatient death (aOR=0.70 (95% CI 0.58 to 0.85)), transfer to Intensive Care Unit (ICU) (adjusted Odds Ratio (aOR=0.91 (95% CI 0.84 to 0.99)) and cardiac arrest (OR=0.72 (95% CI 0.55 to 0.95)), all adjusted for important confounders.13

We analysed the Healthcare Cost and Utilization Project Nationwide Inpatient Sample (HCUP NIS)14 to identify patients diagnosed with unstable angina (UA), acute myocardial infarction (AMI), acute decompensated heart failure (ADHF) and atrial fibrillation with rapid ventricular response (AFRVR), who had a concomitant diagnosis of OSA during their index hospitalisation. We looked at CPAP utilisation, in-hospital deaths, hospital length of stay (LOS) and total hospital charges among these patients.


Data source

A retrospective case–control analysis of the deidentified data from the 2016–2019 HCUP NIS14 was carried out. The time frame was chosen to avoid the myriad complex and confounding effects that the SARS-CoV-19 pandemic could have potentially contributed in the years that followed. The HCUP NIS is a comprehensive and nationally representative dataset that provides detailed information on inpatient hospital admissions, offering valuable insights into healthcare utilisation and costs in the USA.15 The 2016–2019 dataset contains data from over 32 million hospitalisations, with diagnoses and procedures encoded with 10th International Classification of Diseases, Clinical Modification (ICD-10-CM) and Procedure Coding System (ICD-10-PCS) codes, respectively.16 Additionally, several important clinical characteristics are further encoded using dataset-specific Agency for Healthcare Research and Quality (AHRQ) codes (eg, diabetes and hypertension with or without complications, alcohol abuse, severe renal failure). The dataset reports the expenses associated with each inpatient admission in US dollars (USD) rounded to the nearest dollar.16

Characteristics of the study sample

Using ICD-10-CM codes, we searched 32 355 827 electronic records available from the 2016–2019 NIS dataset to include patients with diagnoses of UA, AMI, ADHF and AFRVR (figure 1). With the help of ICD-10-CM and AHRQ codes, for each included inpatient record, we identified concomitant diagnosis of OSA, age (categorised), gender, race, past or current tobacco use, alcohol abuse, along with the presence of major comorbidities (diabetes, hypertension, heart failure with reduced and preserved ejection fraction (HFrEF, HFpEF), non-severe or non-end-stage and severe or end-stage renal failure, obesity (excluding overweight), obesity hypoventilation syndrome (OHS; Pickwickian), central sleep apnoea, chronic obstructive pulmonary disease (COPD), right ventricular hypertrophy (RVH) or cor pulmonale, types II and III pulmonary hypertension, any non-traumatic cerebrovascular accident (ischaemic or haemorrhagic stroke, intracranial or intracerebral haemorrhage), sepsis (with and without shock)). Last, we identified patients treated with CPAP using ICD-10-PCS codes. A full list of extracted items along with respective query codes can be found in online supplemental table 1. Missing cases were excluded pairwise.

Supplemental material

Figure 1
Figure 1

Flow chart of record selection from the HCUP NIS database (2016–2019). ADHF, acute decompensated heart failure; AFRVR, atrial fibrillation with rapid ventricular response; CPAP, continuous positive airway pressure; ICD-10-CM, 10th International Classification of Diseases, Clinical Modification; HCUP NIS, Healthcare Cost and Utilization Project Nationwide Inpatient Sample; OSA, obstructive sleep apnoea; UA, unstable angina.


Our primary outcome of interest was in-hospital death. Secondary outcomes included LOS and total hospital charges in USD. We compared in-hospital death, LOS and total charges between patients with and without OSA, and between OSA patients who either received or did not receive CPAP treatment during index hospitalisation.

Statistical analysis

We ran binary logistic regression analysis to determine the unadjusted OR of death during hospitalisation with respect to the presence of OSA among all patients with UA, AMI, ADHF, as well as in each diagnostic subgroup, and among patients with OSA with respect to CPAP treatment. To adjust for potential confounders, several covariates were introduced into the logistic regression model: sex, age category, race, smoking, alcohol abuse, obesity, hypertension, diabetes mellitus, any non-traumatic cerebrovascular accident, COPD, OHS, pulmonary hypertension (type II and III), cor pulmonale or RVH, HFpEF, HFrEF, end-stage renal disease or severe renal failure, non-end-stage chronic kidney disease or non-severe renal failure, sepsis with and without shock. The LOS and hospital charges were compared between subgroups of interest using one-way analysis of variance and univariate analysis of covariates was used to adjust for confounders. In all analyses, a two-tailed alpha level of 0.05 was used as a cut-off for statistical significance. All analyses were carried out using SPSS statistical software (IBM, V.29.0).

Patient and public involvement

A deidentified dataset obtained from a third party was used for this research. As such, patients were not directly involved in the design, recruitment or conduct of the study.


Demographic and clinical characteristics

Our sample included 2 959 991 patients, of which 1.5% were diagnosed with UA, 30.3% with AMI, 37.5% with ADHF and 45.8% with AFRVR. OSA was present in 12.3%. Patients with OSA were slightly younger, more likely to be male, obese and report past or current smoking (p<0.001 for all). They were also more often comorbid with diabetes, renal failure, HFpEF, COPD, cor pulmonale or RVH, pulmonary hypertension and treated with CPAP (p<0.001). The complete demographic and clinical characteristics are shown in table 1.

Table 1

Demographic and clinical characteristics of the study population

Death during hospitalisation

Patients with OSA had significantly lower odds of death (crude OR 0.55, 95% CI (0.54 to 0.56)), which remained significant following adjustment for multiple confounders (aOR 0.72, 95% CI (0.70 to 0.73)) and was similarly observed across all subgroups except UA (table 2). The reduced odds of death among OSA patients were similar through years 2016–2019 (figure 2).

Figure 2
Figure 2

The reduced odds of in-hospital death when diagnosed with OSA through years 2016–2019. ADHF, acute decompensated heart failure; AFRVR, atrial fibrillation with rapid ventricular response; AMI, acute myocardial infarction; HCUP NIS, Healthcare Cost and Utilization Project Nationwide Inpatient Sample; OSA, obstructive sleep apnoea.

Table 2

Outcomes for patient subgroups with unstable angina (UA), acute myocardial infarction (AMI), acute decompensated heart failure (ADHF) and atrial fibrillation with rapid ventricular response (AFRVR)

Among patients with OSA, CPAP treatment during hospitalisation was found to significantly increase the odds of death (crude OR 1.74, 95% CI (1.65 to 1.83)), which remained significant across all subgroups except UA, and following adjustment for multiple confounders (aOR 1.51, 95% CI (1.44 to 1.60), table 2).

LOS and total hospital charges

There were no statistically significant differences in the LOS between patients with and without OSA (p=0.60), and the difference was small following adjustment for confounders (adjusted mean difference −0.06 days, 95% CI (−0.08 to –0.03); p<0.001, online supplemental table 2). On the contrary, patients with OSA were charged significantly lower costs (adjusted mean difference: −US$4709, 95% CI (−US$5085 to –US$4332); p<0.001), which was similar across subgroups except UA.

Patients with OSA who were treated with CPAP during their hospitalisation spent significantly more inpatient days (adjusted mean difference 1.49 days, 95% CI (1.43 to 1.55); p<0.001) and were charged significantly higher costs (adjusted mean difference: US$1168, 95% CI (US$273 to US$2062); p=0.011).


In our study of about 3 million hospitalisation records, patients with OSA were found to have significantly lower in-hospital mortality regardless of CPAP use. Additionally, there was no significant difference in LOS between patients with and without OSA, while patients with OSA were found to have lower hospital charges. OSA patients treated with CPAP were at increased odds of in-hospital death, had longer stay duration and were charged larger hospital costs. The small fraction of UA patients included in our sample (1.5%), the relatively lower severity of the UA, along with its more limited in-hospital mortality (1.1%, as opposed to 3.7%–8.1% in the other acute cardiovascular encounters, table 2) could have accounted for the non-significant differences observed with respect to OSA and CPAP use in this subgroup.

Some studies have reported significantly lower in-hospital mortality17 and less severe cardiac injury18 in AMI patients with OSA. Our study included a larger selection of cardiovascular disease, while aiming to capture the acuity of the encounter: AMI, ADHF and AFRVR; and our results were consistent with the findings of these studies. Our results also showed that patients with OSA were more likely to be younger, male, smokers, obese, and have stroke, COPD, renal failure, and heart failure, which was in line with the findings of many studies reporting associations between cardiovascular disease and OSA (table 1).7 8 10 11 13 17–19

While acknowledging the myriad reported negative cardiovascular outcomes associated with OSA,7 9 19–21 we postulate that the observed survival benefit may be due to ischaemic preconditioning,22–24 a potential cardioprotective phenomenon caused by OSA-induced chronic intermittent hypoxia, resulting in miniature ischaemic episodes that confer adaptation and protection from future infarctions and life-threatening arrhythmias.18 22 25–29 Our findings also showed that patients with OSA who were hospitalised with acute cardiovascular disease and were treated with CPAP had higher mortality. Contrary to expectations that aggressively treating hospitalised OSA patients with CPAP would dramatically improve outcomes,30 we postulate that OSA patients might be chronically dependent on intermittent hypoxia, and the routine use of CPAP during the acute cardiovascular encounter (eg, AMI, ADHF, AFRVR) could neutralise the potential benefits of ischaemic preconditioning, much similar to how routine administration of oxygen to normoxic patients with AMI confers no additional benefit and may even cause harm.31 32

A personalised and carefully considered approach to OSA treatment, particularly during acute cardiovascular events, is essential, as routine CPAP therapy may not always be beneficial and could potentially have adverse effects in certain patients. However, the clinical implications of this study remain open. Further research is required to identify the clinical makeup of patients that would and would not benefit from the use CPAP. Furthermore, prospective trials are needed to assess the risks and benefits of routine versus more selective CPAP use among patients with OSA in the context of acute cardiovascular hospitalisations in general, and AMI, ADHF, and AFRVR in particular. Should these findings be confirmed in randomised prospective settings, a more selective approach will be warranted to improve mortality, reduce LOS and associated healthcare costs.

Although our results are thought-provoking, there are several notable limitations to our analysis. First, we used ICD-10-CM codes to identify OSA patients, thereby lacking data on disease severity, as well as the use of and compliance with CPAP treatment prior to hospitalisation,16 and thereby increasing the risk for selection bias. Second, potential issues arising from incomplete or inaccurate database records along with the lack of randomisation are inherent to the retrospective design of our study, collectively contributing to a higher risk of unidentifiable confounders that could have influenced the results. Third, we did not account for many important diagnostic and therapeutic procedures performed during hospitalisations, such as thrombolysis, diagnostic angiography, percutaneous coronary intervention (PCI) and coronary artery bypass grafting, all of which can impact mortality, LOS and hospital charges.33 Fourth, the observed cost discrepancies in acute cardiovascular hospitalisations above, specifically in the context of OSA and CPAP usage, may not be generalisable to healthcare settings beyond the USA. Fifth, our analysis may not have been sufficiently powered to detect significant impacts of OSA and CPAP use on UA outcomes, given the limited number of included UA patients coupled with a generally low in-hospital mortality associated with the latter diagnosis. Last, although we applied multivariable adjustment for demographic and clinical confounders, there is possibility of unaccounted confounding variables. Notwithstanding these limitations, our results were consistent with previous research.13 17 18 22 26 27


Patients with recognised OSA who are hospitalised with AMI, ADHF and AFRVR have significantly lower in-hospital mortality, whereas OSA patients treated with CPAP had increased in-hospital mortality, resource utilisation and LOS. RCTs focusing on severity of sleep apnoea in different high-risk patient groups are needed to clarify the current practice of initiating CPAP in the acute setting, much like how the old practice of routine oxygen supplementation in AMI was revisited.

Data availability statement

Data are available on reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

Given the deidentified nature of the HCUP NIS database, the current study was deemed exempt from ethical approval requirements by the Institutional Review Board of the Mountainview Hospital.

This post was originally published on