Sleep-Disordered Breathing and Postoperative Outcomes After Bariatric Surgery: Analysis of the Nationwide Inpatient Sample
- First Online:
- Cite this article as:
- Mokhlesi, B., Hovda, M.D., Vekhter, B. et al. OBES SURG (2013) 23: 1842. doi:10.1007/s11695-013-0991-2
- 553 Views
Sleep-disordered breathing (SDB), also known as obstructive sleep apnea (OSA), has been increasingly recognized as a possible risk factor for adverse perioperative outcomes in non-bariatric surgeries. However, the impact of SDB on postoperative outcomes in patients undergoing bariatric surgery remains less clearly defined. We hypothesized that SDB would be independently associated with worse postoperative outcomes.
Data were obtained from the Nationwide Inpatient Sample database and included a total of 91,028 adult patients undergoing bariatric surgeries from 2004 to 2008. The primary outcomes were in-hospital death, total charges, and length of stay. There were two secondary outcomes of interest: respiratory and cardiac complications. Regression models were fitted to assess the independent association between SDB and the outcomes of interest.
SDB was independently associated with decreased mortality (OR = 0.34, 95% CI = 0.23–0.50, p < 0.001), total charges (−$869, p < 0.001), and length of stay (−0.25 days, p < 0.001). SDB was independently associated with significantly increased odds ratio of emergent endotracheal intubation (OR = 4.35, 95% CI = 3.97–4.77, p < 0.001), noninvasive ventilation (OR = 14.12, 95% CI = 12.09–16.51, p < 0.001), and atrial fibrillation (OR = 1.25, 95% CI = 1.11–1.41, p < 0.001). Emergent intubation occurred significantly earlier in the postoperative course in patients with SDB. Although non-SDB patients had an overall lower risk of emergent intubation compared to SDB patients, their outcomes were significantly worse when they did get emergently intubated.
In this large nationally representative sample, despite the increased association of SDB/OSA with postoperative cardiopulmonary complications, the diagnosis of SDB/OSA was negatively, rather than positively, associated with in-hospital mortality and resource use.
KeywordsSleep-disordered breathingBariatric surgeryObstructive sleep apneaPostoperative complicationsIntubationRespiratory failureDeathLength of stayCost
Charlson comorbidity index
Continuous positive airway pressure/noninvasive ventilation
International Classification of Diseases, 9th Revision, Clinical Modification
Length of stay
Nationwide inpatient sample
Sleep-disordered breathing (SDB) is increasingly recognized as a possible risk factor for adverse perioperative outcomes [1–6]. Several studies have reported worse postoperative outcomes in SDB patients such as increased rates of hypoxemia, endotracheal intubation, respiratory failure, intensive care unit transfers, increased hospital length of stay (LOS), encephalopathy, and postoperative infections [2, 3, 5–9]. Clinicians may expect SDB to be associated with an increased risk of adverse postoperative outcomes after bariatric surgery. However, to the best of our knowledge, the rates of postoperative complications after bariatric surgery have not been systematically compared in patients with and without SDB in a large, nationally representative sample.
To that end, we examined the association of SDB with several postoperative outcomes in patients undergoing bariatric surgery. We analyzed the Nationwide Inpatient Sample (NIS) database to quantify the impact of the diagnosis of SDB on in-hospital death, total charges, LOS, respiratory outcomes, and cardiac outcomes. We hypothesized that the diagnosis of SDB would be independently associated with worse postoperative outcomes after controlling for comorbidities and demographic characteristics.
Data were obtained from the NIS database, which is one of several databases that form the Healthcare Utilization Project. The NIS is the largest all-payer database in the USA and has been used in a variety of research studies [9–11]. The NIS contains information on approximately 8 million hospitalizations per year from 1,050 hospitals in 44 states. The data approximate a 20 % stratified sample of hospitals in the USA. The data have been collected on an annual basis since 1988 . The database includes a record for every hospital discharge, regardless of payer, at included hospitals during a given year.
This study was approved by the University of Chicago's Institutional Review Board (BSD/UCH IRB approval no. 10-567-E).
Our cohort was derived by including all hospital admissions in adults (age 18 years or more) for bariatric surgeries in the NIS database from the years 2004 to 2008. We selected the most recent 5 years in the NIS database to avoid significant changes in practice patterns. At the time of data extraction, 2008 was the most recent year with data available in the NIS database. Patients were stratified based on the diagnosis of SDB. The ICD-9-CM codes used to characterize SDB are described in the Appendix. The ICD-9-CM codes used to identify the bariatric surgery procedures are also described in the Appendix.
Patient demographics included age, sex, self-reported race/ethnicity, Charlson comorbidity index (CCI), income by quartile, health insurance source (i.e., Medicare, Medicaid, private), teaching or non-teaching hospital status, and United States region (Northeast, South, West, Midwest/Central). The information about race is missing in approximately 27 % of cases because some participating states restrict race data. The CCI is a tool to assign severity to a patient’s comorbid conditions. Common comorbid conditions are assigned varying weights, and the sum of the patient’s score indicates their cumulative comorbid condition; higher scores indicate increased comorbidity . Income was divided into quartiles, with 1 being the poorest quartile and 4 being the wealthiest quartile. Income data were obtained from zip codes and demographic data from Nielson online demographic services .
The primary outcomes compared between SDB and non-SDB patients included in-hospital death, cost in total hospital charges, and LOS. Secondary respiratory outcomes included emergent endotracheal intubation and mechanical ventilation, continuous positive airway pressure/noninvasive ventilation (CPAP/NIV) during hospitalization, tracheostomy, pneumonia, and respiratory failure. Secondary cardiac outcomes included atrial fibrillation and percutaneous coronary procedures.
For unadjusted comparisons between SDB and non-SDB patients, continuous variables (age, LOS, and total charges) were presented as the mean and standard deviation and were compared using the Student’s t test. All other categorical variables were summarized as percentages and compared using the chi-square test. Generalized linear models with log-link and gamma-distributed errors were fitted to assess the independent association between SDB and length of stay and total charges in order to adjust the difference over other independent variables. We used the iteratively reweighted least squares method for maximum likelihood estimation of the model parameters. There is a substantial statistical literature that recommends using generalized linear model techniques when modeling health care costs and LOS due to the unique features of such data, namely, that they tend to be differentially dispersed around the mean (heteroskedastic) and prone to large outliers (right-skewed) [16, 17]. Mean-adjusted LOS and total charges for patients with and without SDB were then estimated from the generalized linear models. We constructed logistic regression models for outcomes that were dichotomous to determine the independent association of SDB and the outcomes of interest. Independent variables introduced in the generalized linear models and logistic regression models included age by quartiles, sex, type of health insurance, income by quartiles, CCI, type of surgical procedure, year of the surgical procedure, hospital teaching status, United States region where the hospital is located, and weekend admission. We included weekend admission as an independent variable given its association with worse outcomes in some studies. Race was not included in the models because it was missing for approximately 27 % of admissions. Body mass index is not available in the NIS database.
The percentage of patients that were dropped from the regression models due to a missing variable was quite negligible, between 3.0 and 3.5 %. Although multiple imputation could have been performed for missing variables, we elected not to do it since missing data was small and statistical power was not an issue. In order to measure model discrimination and the goodness of fit of the models, several statistics were examined, including deviance, Pearson’s chi-square, Akaike information criterion, Bayesian information criterion, and pseudo-R2 statistics.
Stata version 11.0 (StataCorp LP) statistical software was used for all analyses.
Demographic data for bariatric surgery patients based on SDB status
Age, mean (SD)
Charlson comorbidity index (%)
CCI ≥ 4
Insurance type (%)
Income quartile (%)
Teaching hospital (%)
Hospital region (%)
Unadjusted outcomes of bariatric surgery patients based on SDB status
In-hospital death (%)
Total charges (US $)
Length of stay (days)
Respiratory complications (%)
Cardiac complications (%)
Multivariable regression modeling to estimate adjusted risk of outcomes associated with SDB
Regression coefficient or OR for SDB
95% Confidence interval
−0.04 to −0.02
Length of stay
−0.10 to −0.07
SDB was independently associated with a significantly increased odds ratio of emergent endotracheal intubation and mechanical ventilation (OR = 4.35, 95% CI = 3.97–4.77, p < 0.001), CPAP/NIV (OR = 14.12, 95% CI = 12.09–16.51, p < 0.001), and atrial fibrillation (OR = 1.25, 95% CI = 1.11–1.41, p < 0.001). In contrast, SDB was associated with a decreased odds ratio of pneumonia and tracheostomy (Table 3).
Outcomes of patients requiring emergent endotracheal intubation and mechanical ventilation
Outcomes based on the diagnosis of SDB and need for emergent endotracheal intubation in bariatric surgery patients
No sleep-disordered breathing
No. of cases
In-hospital death (%)
Total charges (US $)
Length of stay (days)
Respiratory failure (%)
Intubation on day 0 or day 1 (%)
Multivariable regression modeling to estimate adjusted risk of outcomes in patients requiring emergent endotracheal intubation
Outcomes in patients requiring intubation
Regression coefficient or OR for non-SDB
95% Confidence interval
Length of stay
The use of CPAP/NIV during the hospital course was significantly lower in non-SDB patients requiring emergent intubation and mechanical ventilation. Notably, CPAP/NIV use during the hospitalization was not recorded in any patient who did not require emergent intubation and mechanical ventilation regardless of the SDB status (Table 4). A significantly higher proportion of patients with SDB required emergent intubation and mechanical ventilation on the day of surgery or the first postoperative day compared to non-SDB patients (90 vs. 63 %, p < 0.001).
In our analysis of the largest nationally representative sample of patients undergoing bariatric surgery, we found an increased rate of postoperative cardiopulmonary complications in patients with SDB. Specifically, SDB was independently associated with increased risk of emergent endotracheal intubation, CPAP/NIV use, and atrial fibrillation. These findings differ from what has been previously reported [18–20]. In a retrospective single-center study, Weingarten et al.  found that SDB was not associated with an increased likelihood of pulmonary complications in patients undergoing bariatric surgery. However, 93 % of their patients received postoperative positive airway pressure therapy for SDB and were closely monitored in the postoperative setting with continuous pulse oximetry, which likely explains the low complication rates. Although we have no information regarding the perioperative management of SDB in the NIS sample, we suspect that it was suboptimal as CPAP/NIV use was not recorded in the vast majority of patients with SDB during their hospitalization. In an analysis by the Longitudinal Assessment of Bariatric Surgery consortium, SDB was also not significantly associated with their composite end point of death, deep venous thrombosis/pulmonary embolism, surgical reintervention, or delayed discharge in the adjusted model . These studies were performed in tertiary care academic centers, whereas 42 % of our patients had their surgery performed in a non-teaching hospital [18–20]. Academic centers may have used multidisciplinary perioperative processes and pathways to identify and treat SDB more often than non-teaching hospitals.
Despite the fact that SDB was associated with increased rates of emergent endotracheal intubation, CPAP/NIV use, and atrial fibrillation, we were surprised to find that it was negatively associated with rates of in-hospital death, total charges, or LOS. Other studies have demonstrated that SDB is not a risk factor for death in the postoperative bariatric setting [18, 20–22], but why this association was in the opposite direction of what we expected is unclear. The NIS database includes up to 15 diagnostic codes for each hospitalization. One possibility may be that when a patient dies or has a prolonged hospital stay additional diagnoses are added to their hospitalization, which can push SDB off the list. Another possible putative mechanism by which SDB could decrease mortality is ischemia preconditioning. Indeed, studies have reported that in patients with acute myocardial infarction, those with SDB and intermittent hypoxemia during sleep have less severe cardiac injury and better coronary collateral circulation [23–26]. Therefore, it is plausible that SDB does not increase in-hospital cardiovascular mortality in the immediate postoperative setting despite the increased risk of upper airway complications. Although chronically untreated severe SDB is strongly associated with increased cardiovascular mortality [27–30], it remains unclear whether untreated SDB in the acute postoperative setting can lead to increased cardiovascular mortality. It is also possible that the lower mortality rate and lower levels of resource utilization are due to more appropriate care of SDB patients with CPAP/NIV, while the population of patients without a diagnosis of SDB includes patients with SDB experiencing adverse outcomes because of failure to identify and treat SDB. The higher rates of emergent endotracheal intubation in patients with SDB, however, suggest that it is unlikely that patients with SDB received better postoperative care. In addition, the vast majority of patients undergoing bariatric surgery did not require emergent intubation, and in this cohort, CPAP/NIV use during the hospitalization was not recorded in any patient (Table 4). The low rates of CPAP/NIV utilization may be due to underreporting in the NIS database. Another possibility is that postoperative CPAP/NIV may have been underutilized due to the concern that it can increase the risk of aerophagia and anastomotic leak [31–33]. Although postoperative CPAP does not increase the risk of disrupting the anastomotic integrity after upper gastrointestinal tract surgery , its safety during the immediate postoperative period may have not been widely recognized by bariatric surgeons during the years included in our analysis (2004–2008). Our findings that SDB was not associated with increased LOS and in-hospital mortality are in line with a recent single-center study of presurgical patients . The goal of this prospective observational study was to determine whether a prior diagnosis of SDB or a positive screen for SDB was associated with increased mortality in 14,962 presurgical patients. Surprisingly, despite higher rates of ICU admissions, patients with SDB did not have an increase in LOS or in 30-day mortality in the postoperative period .
Our analysis of patient outcomes in relation to emergent endotracheal intubation yielded interesting results. We found that SDB patients who required emergent intubation were intubated much earlier in their postoperative course. As expected, both SDB and non-SDB patients requiring emergent intubation had worsened outcomes compared to their counterparts who did not require emergent intubation. However, non-SDB patients who required intubation had substantially worse outcomes than SDB patients requiring intubation. This finding suggests that upper airway complications may be the main cause of emergent intubation in patients with SDB. We speculate that in patients with SDB, the primary reason for emergent intubation was related to a rapidly reversible upper airway complication in the setting of sedative and opioid administration in the immediate postoperative period. This speculation is supported by the fact that in patients with SDB, emergent intubation occurred earlier in their postoperative course. Additionally, the fact that emergent intubation in patients with SDB led to a significantly less increase in in-hospital death, total charges, LOS, and pneumonias as compared to non-SDB patients requiring intubation supports this speculation. It is well known that patients with SDB have elevated pharyngeal closing pressures as well as increased upper airway instability and collapsibility in the postoperative setting as compared to matched controls, thus predisposing them to upper airway complications [36, 37]. Our data demonstrate that although an increase in postoperative complications should be anticipated in SDB patients undergoing bariatric surgery, the overall prognosis with a complication like emergent endotracheal intubation is better than if they did not have SDB. Our analysis revealed that non-SDB patients requiring emergent intubation were intubated later in their postoperative course and had higher rates of pneumonia, and we suspect that their intubation was indicative of a global decline or complication that took much longer to recover from resulting in worse outcomes. Another possibility is that clinicians may have had a lower threshold for intubating patients with known SDB in the setting of postoperative respiratory complications.
Our study has several limitations that are inherent to the analysis of large administrative databases, such as the inability to confirm the diagnosis and severity of SDB and lack of data on CPAP therapy at home or during the postoperative period. It is also likely that CPAP/NIV was often not coded even when they were used during the postoperative period. We could not ascertain whether the diagnostic codes for SDB were part of the admission diagnoses or discharge diagnoses; therefore, it is conceivable that respiratory complications could have led to a presumptive diagnosis of SDB. We also do not have clinical information on all potential confounders, such as body mass index. Moreover, the NIS database also lacks information about readmissions as well as long-term postoperative outcomes.
We found the prevalence of SDB to be 36 % among patients undergoing elective bariatric surgery in the NIS database. We suspect that the majority of the cases of SDB were of significant severity that warranted clinical attention, which is why they were coded for in the database. Although most recent studies have reported a prevalence of severe obstructive sleep apnea in patients undergoing bariatric surgery ranging from 27 to 36 %, there is likely a component of underreporting in the NIS database since the reported prevalence of moderate and severe SDB in patients undergoing bariatric surgery is 44–48 % [20, 38–40]. The lower than expected prevalence of SDB in patients undergoing bariatric surgery can in part be related to a significantly higher prevalence of premenopausal women undergoing surgery. Moreover, it is important to consider that body mass index explains only 4–26 % of the variance of SDB severity as measured by the apnea–hypopnea index [41–43]. Therefore, we believe that the NIS database does not represent a significant underreporting of clinically significant cases of SDB.
In summary, using a large nationally representative database, we have reported that in patients undergoing bariatric surgery, SDB is independently associated with significant postoperative cardiopulmonary complications, but not with increased in-hospital mortality, total charges, and LOS. Although non-SDB patients had an overall lower risk of emergent intubation compared to SDB patients, their outcomes were significantly worse when they did get emergently intubated. Given the overall low complication rates, prospective large-scale multicenter randomized controlled trials are needed to assess the impact of SDB treatment on patient outcomes during the postoperative period in patients undergoing bariatric surgery.
This study was supported by the University of Chicago Institute for Translational Medicine and the Clinical and Translational Science Awards (CTSA) program (UL1 RR024999). Dr. Arora is supported by the National Institutes on Aging (K23 AG033763). Dr. Meltzer is supported by a Midcareer Career Development Award from the National Institutes of Health (1 K24 AG031326-01).
Conflict of Interest Statement
Babak Mokhlesi has served as a consultant for Philips/Respironics, but has no conflicts of interest relevant to the present study. Margaret D. Hovda, Benjamin Vekhter, Vineet M. Arora, Frances Chung, and David O. Meltzer have no conflicts of interest relevant to the present study to declare.
Role of the Sponsor
The sponsor had no role in the design and conduct of the study, in the collection, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript.