Differential Perioperative Outcomes in Patients With Obstructive Sleep Apnea, Obesity, or a Combination of Both Undergoing Open Colectomy: A Population-Based Observational Study : Anesthesia & Analgesia

Secondary Logo

Journal Logo

Original Research Articles: Original Clinical Research Report

Differential Perioperative Outcomes in Patients With Obstructive Sleep Apnea, Obesity, or a Combination of Both Undergoing Open Colectomy: A Population-Based Observational Study

Stundner, Ottokar MD, MBA*,†; Zubizarreta, Nicole MPH; Mazumdar, Madhu PhD; Memtsoudis, Stavros G. MD, PhD, MBA*,§,‖; Wilson, Lauren A. MPH; Ladenhauf, Hannah N. MD; Poeran, Jashvant MD, PhD#

Author Information
Anesthesia & Analgesia 133(3):p 755-764, September 2021. | DOI: 10.1213/ANE.0000000000005638
  • Free
  • SDC
  • CME Test
  • Continuing Medical Education

Abstract

KEY POINTS

  • Question: What are the differential impacts of obesity, obstructive sleep apnea (OSA), or a combination of both on perioperative outcomes in patients undergoing open colectomy?
  • Findings: In this study, we found an incremental effect of OSA alone, obesity alone, or a combination of both conditions on the risk of perioperative complications, and length and cost of hospitalization in patients undergoing open colectomy.
  • Meaning: Obesity and OSA are highly prevalent and likely underreported comorbidities in patients undergoing colectomy; this study helps to better tailor interventions, screening, and precautionary measures to the individual risk.

Obstructive sleep apnea (OSA) is an established risk factor for adverse perioperative outcomes across a number of surgical disciplines,1 including orthopedic surgery,2 gynecology,3 and obstetrics.4 A little is known about the impact of OSA on the risk of perioperative complications after colectomy. This procedure is of particular interest given the high baseline perioperative complication risk, making interventions with the potential to optimize outcome highly sought after.5 Moreover, while OSA is often found in combination with high body mass index (BMI), distinct phenotypes of nonobese OSA patients and obese patients without OSA exist with considerable frequency.6 Importantly, there is a lack of knowledge about the individual impact of OSA or obesity (Ob), and their composite impact on perioperative outcome. Finally, as both OSA and Ob have recently been identified as individual risk factors for the development of colorectal cancer, patients suffering from either condition are more likely to undergo a colectomy at some point in their life.7

Therefore, using a population-based approach, this study seeks to: (1) quantify the impact of OSA on perioperative complications/resource utilization in patients undergoing colectomy and (2) further specify the role of Ob in the association between OSA and complications/resource utilization. We hypothesized that while both Ob and OSA individually increase the likelihood for perioperative complications, the overlap of the 2 conditions will be associated with the highest risk.

METHODS

Study Design, Data Source, and Ethics Approval

This retrospective cohort study used claims data from 2006 to 2016 as recorded in the Premier Healthcare Database (Premier, Inc, Charlotte, NC). This database contains discharge information from approximately 600 acute care hospitals located throughout the United States and representing 20% to 25% of all hospitalizations.8 This study was considered exempt from consent requirements by our Institutional Review Board as data are deidentified and in accordance with the Health Insurance Portability and Accountability Act.9

Study Sample

Data were included in the study cohort if patients had International Classification of Diseases-9th Revision-Clinical Modification (ICD-9-CM) procedure codes for open colectomies, including right hemicolectomy, left hemicolectomy, resection of transverse colon, sigmoidectomy, or other procedures (see Supplemental Digital Content 1, Table 1, https://links.lww.com/AA/D582, for ICD-9-CM codes). Exclusion criteria in order of implementation included trauma cases (n = 4503), patients with no information on opioid utilization (n = 16,209), missing demographic information (n = 87), unknown discharge status or unknown mortality status (n = 289), outpatient procedures (n = 319), and hospitals with <30 procedures (n = 802).

Study Variables

The main effects of interest were Ob (defined based on the definition used in the Elixhauser comorbidity index10) and OSA as defined by the ICD-9-CM codes, all listed in Supplemental Digital Content 2, Table 2, https://links.lww.com/AA/D583. Only preoperatively confirmed diagnoses of OSA and Ob are captured, which is a frequently utilized approach used in numerous previous reports.2 Four separate subgroups were defined based on the presence of diagnosis codes of OSA, Ob, both, or none: [+OSA/+Ob]: both OSA and Ob diagnoses present; [+OSA/−Ob]: OSA diagnosis present but not Ob diagnosis; [−OSA/+Ob]: Ob diagnosis present but not OSA; [−OSA/−Ob]: neither OSA nor Ob diagnoses present. These subgroups allow a distinction between risk estimations of OSA and Ob separately on outcomes. The main outcomes of interest were respiratory and cardiac complications, intensive care unit (ICU) admission, utilization of mechanical ventilation, and inhospital mortality (see Supplemental Digital Content 3, Table 3, https://links.lww.com/AA/D584, for definitions of these outcomes). In addition, cost and length of hospitalization as well as opioid utilization were evaluated. Cost data are submitted to Premier directly from participating hospitals on the individual patients’ level in the form of a chargemaster database, a comprehensive table of billable items including “hospital services, medical procedures, equipment fees, supplies, drugs, and diagnostic evaluations such as imaging and laboratory tests.”8 Using the chargemaster database along with unique identifying codes and set prices, both services utilization and costs can be calculated accurately. Opioid utilization was expressed in oral morphine equivalents computed from billing data using the GlobalRPH “opioid analgesic converter”11 and the Lexicomp “opioid agonist conversion.”12 Patient characteristics were age, gender, and race/ethnicity (Caucasian, African American, Hispanic, and Other). Health care–related variables were insurance type, hospital location (rural versus urban), hospital size, hospital teaching status, and hospital-specific volume of colectomies performed annually. Procedure-related characteristics included admission type (elective versus urgent), indication for colectomy (neoplasm, diverticular disease, inflammatory bowel disease, and others), type, and year of procedure. Anesthesia/analgesia-related variables were type of anesthesia (general, combined general-neuraxial, and other), use of patient-controlled analgesia, and nonopioid analgesic use (intravenous acetaminophen, nonsteroidal antiinflammatory drugs, cyclooxygenase-2-inhibitors, ketamine, pregabalin, and gabapentin). The prevalence of individual comorbidities and overall comorbidity burden was assessed using the Charlson comorbidity index, calculated by using the method described by Deyo et al.13 In addition, history of substance use/abuse, chronic pain conditions, and psychiatric comorbidities were included given their association with opioid use.

Statistical Analysis

All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC). First, univariable analyses were performed to compare the study cohorts in terms of patient demographics, and health care–, procedure-, and anesthesia/analgesia-related characteristics and comorbidities. χ2 and t tests (or nonparametric tests of significance, where applicable) were used for categorical and continuous variables, respectively. Given our sample size (and thus decreased need for parsimony), all available variables were subsequently included in multilevel multivariable models that measured the association between the 4 OSA-Ob subgroups and outcomes. Multilevel models were applied to account for correlation of patients within hospitals; here, a random intercept term was applied that varies at the level of each hospital. The [−OSA, −Ob] group was defined as the reference. For the continuous outcome variables (length of hospital stay and cost of hospitalization), we applied a log-link with gamma distribution. Results from the multilevel logistic regression model are presented as percent change compared to the reference and 95% confidence interval (CI) for continuous outcomes, and as odds ratios (OR) and 95% CI for binary outcomes.

To measure the supraadditive effect of the interaction term between OSA and Ob in relation to the binary outcomes, the relative excess risk due to interaction (RERI) and 95% CI were calculated. RERI can be defined as the proportion of increased risk for the binary outcomes due to the interaction of OSA and Ob, when compared to the sum of both conditions’ individual risk.14 It can be interpreted as the difference between the summation of the separate effects and the effect in the joint exposure category, that is, the additional risk caused by suffering from both conditions at the same time. A positive RERI signifies supraadditive risk (or effect “exceeding the sum of the individual risks”); a RERI of zero signifies no interaction between the individual conditions; and a negative RERI signifies infraadditive risk (or “canceling out” of risks). Concordance statistics (c-statistics) were determined as a measure for goodness-of-fit. Intraclass correlation coefficients (ICCs) were calculated to estimate to what extent variation in outcomes could be explained by between-hospital variation.

Sensitivity Analyses

To determine the influence of Ob severity (BMI subclass) on outcomes, we conducted a sensitivity analysis restricting the definition of Ob to the most severe BMI class in the cohort (V83.4X; BMI >40; see Supplemental Digital Content 2, Table 2, https://links.lww.com/AA/D583). In sensitivity analysis number 2, we repeated our main analysis by including the initially excluded cohort of patients that did not have information on opioid utilization, by assuming that these patients did not receive any opioids.

RESULTS

Three hundred forty thousand forty-seven patients who underwent open colectomy between 2006 and 2016 were identified in the database. The Figure provides a graphical presentation of the 4 subgroups of interest and their sizes. Of the overall cohort, 9028 (2.7%) had both OSA and Ob diagnoses ([+Ob, +OSA]), 10,137 (3.0%) had a diagnosis of OSA but not Ob ([+OSA, −Ob]), 33,692 (9.9%) had a diagnosis of Ob but not OSA ([−OSA, +Ob]), and 287,190 (84.5%) had neither ([−OSA, −Ob]). This signifies that just 47.1% of patients with OSA were identified as being obese.

F1
Figure.:
Group distribution ([+OSA/+Ob]: both OSA and obesity diagnoses present; [+OSA/−Ob]: OSA diagnosis present but not obesity diagnosis; [−OSA/+Ob]: obesity diagnosis present but not OSA; [−OSA/−Ob]: neither OSA nor obesity diagnoses present). Numbers are displayed as group size (n) and percentage of entire cohort. Ob indicates obesity; OSA, obstructive sleep apnea.

Table 1 lists patients’ demographics, and health care–, procedure-, anesthesia/analgesia-related characteristics and comorbidities, subgrouped by the 4 distinct subgroups. While the majority of group comparisons were statistically significant, the differences between the [+OSA, +Ob] and [+OSA, −Ob] groups were slightly more pronounced than those between the [−OSA, +Ob] and [−OSA, −Ob] groups. Overall, some comparisons stood out in particular. Patients in the [−OSA, +Ob] and [−OSA, −Ob] groups were more frequently female (62.7% and 53.7%, respectively), compared to the other groups. Diverticular disease and inflammatory bowel disease were more frequent indications for colectomy in groups [+OSA, +Ob], [+OSA, −Ob], and [−OSA, +Ob] compared to the group with no OSA-Ob diagnoses, while neoplasm was a more frequent indication in [−OSA, −Ob]. The Charlson comorbidity index was highest in [+OSA, +Ob] (2.6 ± 2.8), followed by [+OSA, −Ob] (2.4 ± 2.8), and [−OSA, −Ob] (2.3 ± 3.0); interestingly, it was lowest in the group with a sole Ob diagnosis (2.2 ± 2.8), which was also the group with the lowest mean age. Opioid utilization was highest in the group [+OSA, +Ob] and lowest in [−OSA, −Ob], with the other groups in between.

Table 1. - Study Variables by Groups
Variable Name [+OSA, +Ob] [+OSA, −Ob] [−OSA, +Ob] [−OSA, −Ob] P value
Group size 9028 (2.7%) 10,137 (3.0%) 33,692 (9.9%) 287,190 (84.5%) -
Patient demographics
 Mean age 61.9 ± 11.9 65.6 ± 12.4 60.5 ± 13.9 63.6 ± 16.6 <.0001
 Gender
  Female 4295 (47.6%) 3990 (39.4%) 21,126 (62.7%) 154,263 (53.7%) <.0001
  Male 4733 (52.4%) 6147 (60.6%) 12,566 (37.3%) 132,927 (46.3%)
 Ethnicity
  Caucasian 6883 (76.2%) 8040 (79.3%) 24,040 (71.4%) 207,067 (72.1%) <.0001
  African American 963 (10.7%) 796 (7.9%) 4281 (12.7%) 29,029 (10.1%)
  Hispanic 119 (1.3%) 91 (0.9%) 661 (2.0%) 6385 (2.2%)
  Other 1063 (11.8%) 1210 (11.9%) 4710 (14.0%) 44,709 (15.6%)
Healthcare-related
 Insurance type
  Commercial 3037 (33.6%) 3297 (32.5%) 12,896 (38.3%) 96,559 (33.6%) <.0001
  Medicaid 684 (7.6%) 413 (4.1%) 2848 (8.5%) 19,907 (6.9%)
  Medicare 4822 (53.4%) 6039 (59.6%) 15,445 (45.8%) 151,782 (52.9%)
  Uninsured 231 (2.6%) 146 (1.4%) 1611 (4.8%) 11,941 (4.2%)
  Unknown 254 (2.8%) 242 (2.4%) 892 (2.7%) 7001 (2.4%)
 Hospital location
  Rural 957 (10.6%) 1049 (10.4%) 3684 (10.9%) 31,131 (10.8%) .3425
  Urban 8071 (89.4%) 9088 (89.7%) 30,008 (89.1%) 256,059 (89.2%)
 Hospital size
  <300 beds 2991 (33.1%) 3241 (32.0%) 11,742 (34.9%) 95,705 (33.3%) <.0001
  300–499 beds 3392 (37.6%) 3817 (37.7%) 11,976 (35.6%) 102,973 (35.9%)
  ≥500 beds 2645 (29.3%) 3079 (30.4%) 9974 (29.6%) 88,512 (30.8%)
 Hospital teaching status
  Nonteaching 5243 (58.1%) 5866 (57.9%) 20,182 (59.9%) 166,014 (57.8%) <.0001
  Teaching 3785 (41.9%) 4271 (42.1%) 13,510 (40.1%) 121,176 (42.2%)
 Hospital-specific mean # of annual colectomies 122 ± 67 126 ± 69 118 ± 66 121 ± 70 <.0001
Procedure-related
 Admission type
  Elective 4248 (47.1%) 5628 (55.5%) 16,211 (48.1%) 135,220 (47.1%) <.0001
  Urgent 4780 (52.9%) 4509 (44.5%) 17,481 (51.9%) 151,970 (52.9%)
 Indication for colectomy
  Neoplasm 2137 (23.7%) 2375 (23.4%) 8218 (24.4%) 76,836 (26.8%) <.0001
  Diverticular disease 2598 (28.8%) 2881 (28.4%) 10,656 (31.6%) 67,818 (23.6%) <.0001
  Inflammatory bowel disease 2705 (30.0%) 3072 (30.3%) 11,192 (33.2%) 76,182 (26.5%) <.0001
  Other 4246 (47.0%) 4745 (26.8%) 14,535 (43.1%) 136,192 (47.4%) <.0001
 Type of procedure
  Right hemicolectomy 2868 (31.8%) 3221 (31.8%) 9859 (29.3%) 97,804 (34.1%) <.0001
  Left hemicolectomy 1115 (12.4%) 1192 (11.8%) 4271 (12.7%) 34,001 (11.8%) <.0001
  Resection of transverse colon 656 (7.3%) 602 (5.9%) 2046 (6.1%) 16,659 (5.8%) <.0001
  Sigmoidectomy 2978 (33.0%) 3410 (33.6%) 12,309 (36.5%) 92,574 (32.2%) <.0001
  Other 1760 (19.5%) 2009 (19.8%) 6329 (18.8%) 55,023 (19.2%) .0883
 Year of procedure
  2006–2008 1321 (14.6%) 2208 (21.8%) 6960 (20.7%) 91,163 (31.7%) <.0001
  2009–2011 2120 (23.5%) 2364 (23.3%) 8116 (24.1%) 73,212 (25.5%)
  2012–2014 3065 (34.0%) 3133 (30.9%) 10,745 (31.9%) 75,339 (26.2%)
  2015–2016 2522 (27.9%) 2432 (24.0%) 7871 (23.4%) 47,476 (16.5%)
Anesthesia/analgesia
 Anesthesia type
  General 7381 (81.8%) 8358 (82.5%) 27,780 (82.5%) 233,831 (81.4%) <.0001
  General + neuraxial 291 (3.2%) 358 (3.5%) 1240 (3.7%) 11,033 (3.8%)
  Unknown/missing 1356 (15.0%) 1421 (14.0%) 4672 (13.9%) 42,326 (14.7%)
  Use of patient-controlled analgesia 2223 (24.6%) 2746 (27.1%) 9312 (27.6%) 81,034 (28.2%) <.0001
 Use of nonopioid analgesics
  IV acetaminophen 1695 (18.8%) 1742 (17.2%) 5957 (17.7%) 38,512 (13.4%) <.0001
  NSAID 2649 (29.3%) 3307 (32.6%) 11,391 (33.8%) 97,277 (33.9%) <.0001
  Cox-2 inhibitor 127 (1.4%) 165 (1.6%) 444 (1.3%) 3107 (1.1%) <.0001
  Ketamine 405 (4.5%) 381 (3.8%) 1241 (3.7%) 8305 (2.9%) <.0001
  Pregabalin/gabapentin 1030 (11.4%) 938 (9.3%) 2435 (7.2%) 13,158 (4.6%) <.0001
  Oral morphine equivalents (mg) 720 (33–1580) 573 (278–1161) 641 (300–1360) 520 (244–1095) <.0001
 Comorbidities
  Mean Charlson comorbidity index 2.6 ± 2.8 2.4 ± 2.8 2.2 ± 2.8 2.3 ± 3.0 <.0001
  History of substance use/abuse 1509 (16.7%) 1602 (15.8%) 5824 (17.3%) 50,197 (17.5%) <.0001
  Pain conditions 4041 (44.8%) 4416 (43.6%) 14,454 (42.9%) 126,101 (43.9%) .0011
  Psychiatric comorbidities 2615 (29.0%) 2568 (25.3%) 7515 (22.3%) 48,664 (16.9%) <.0001
Patient- and procedure-related characteristics are listed in this table, subgrouped by presence of OSA and/or obesity diagnosis ([+OSA/+Ob]: both OSA and obesity diagnoses present; [+OSA/−Ob]: OSA diagnosis present, but not obesity diagnosis; [−OSA/+Ob]: obesity diagnosis present, but not OSA; [−OSA/−Ob]: neither OSA nor obesity diagnoses present). Data are presented as mean and standard deviation for continuous variables, or prevalence/incidence and percent for categorical variables, respectively. For univariate significance analysis, t test or χ2 test were used for continuous and categorical variables, respectively.
Abbreviations: Cox-2, cyclooxygenase 2; IV, intravenous; NSAID, nonsteroidal anti-inflammatory drug; Ob, Obesity; OSA, obstructive sleep apnea.

Table 2 lists unadjusted outcomes of interest by study subgroups. Length of hospital stay was longest in patients in [+OSA, +Ob] but similar in the 3 remaining groups. Cost of hospitalization was highest in the group [+OSA, +Ob], followed by [−OSA, +Ob] and [+OSA, −Ob], and lowest in [−OSA, −Ob]. Of the 4 groups, patients in group [+OSA, +Ob] had the highest unadjusted incidences of all complications. Patients with OSA only ([+OSA, −Ob]) had intermediary incidences of respiratory and cardiac complications, ICU utilization, and mechanical ventilation, but lower inhospital mortality than patients with neither OSA nor Ob. Similarly, patients with Ob only ([−OSA, +Ob]) had intermediary incidences for respiratory complications, ICU utilization, and mechanical ventilation, but lower incidences of cardiac complications and inhospital mortality than patients with neither OSA nor Ob.

Table 2. - Outcomes by Groups
[+OSA, +Ob] [+OSA, −Ob] [−OSA, +Ob] [−OSA, −Ob] P value
Group size 9028 (2.7%) 10,137 (3.0%) 33,692 (9.9%) 287,190 (84.5%) -
Length of hospital stay (d) 10 (6–17) 8 (5–13) 8 (6–14) 8 (5–13) <.0001
Cost of hospitalization (USD) 25,370 (15,917–45,061) 19,093 (12,669–32,153) 20,500 (13,560–34,697) 17,998 (11,947–30,446) <.0001
Respiratory complications 1953 (21.6%) 1351 (13.3%) 4511 (13.4%) 29,141 (10.2%) <.0001
Cardiac complications 969 (10.7%) 935 (9.2%) 2683 (8.0%) 23,825 (8.3%) <.0001
ICU utilization 2702 (29.9%) 2039 (20.1%) 6763 (20.1%) 42,367 (14.8%) <.0001
Mechanical ventilation utilization 2241 (24.8%) 1498 (14.8%) 5552 (16.5%) 38,440 (13.4%) <.0001
Mortality 583 (6.5%) 492 (4.9%) 1527 (4.5%) 15,631 (5.4%) <.0001
This table displays opioid utilization, length of hospital stay, cost of hospitalization, and complication incidence, subgrouped by presence of OSA and/or obesity diagnosis ([+OSA/+Ob]: both OSA and obesity diagnoses present; [+OSA/−Ob]: OSA diagnosis present, but not obesity diagnosis; [−OSA/+Ob]: obesity diagnosis present, but not OSA; [−OSA/−Ob]: neither OSA nor obesity diagnoses present). Data are presented as median and interquartile range for continuous variables, or incidence and percent for categorical variables, respectively. For univariate significance analysis, Kruskal-Wallis test or χ2 test were used for continuous and categorical variables, respectively.
Abbreviations: ICU, intensive care unit; Ob, obesity; OSA, obstructive sleep apnea; USD, US dollars.

Table 3 lists results from the multivariable models. In the [+OSA, +Ob], [+OSA, −Ob], and [−OSA, +Ob] groups, length of hospital stay was increased 41.9% (CI, 34.7%–49.6%), 4.6% (CI, −0.5% to 10.0%), and 19.7% (CI, 16.3%–23.2%) when compared to the reference (neither diagnosis present), respectively. Likewise, cost of hospitalization was increased by 43.0% (CI, 39.6%–46.5%), 10.1% (7.6%–12.6%), and 22.8% (21.2%–24.4%).

Table 3. - Results From Multilevel Multivariable Models
[+OSA, +Ob] [+OSA, −Ob] [−OSA, +Ob] [−OSA, −Ob]
Length of hospital stay 41.9% (34.7–49.6%), P < .0001 4.6% (−0.5 to 10.0%), P = .0809 19.7% (16.3–23.2%), P < .0001 Reference
Cost of hospitalization 43.0% (39.6–46.5%), P < .0001 10.1% (7.6–12.6%), P < .0001 22.8% (21.2–24.4%), P < .0001
Respiratory complications 2.41 (2.28–2.56), P < .0001 1.40 (1.31–1.49), P < .0001 1.50 (1.45–1.56), P < .0001
Cardiac complications 1.57 (1.45–1.69), P < .0001 1.17 (1.09–1.26), P < .0001 1.28 (1.23–1.34), P < .0001
ICU utilization 1.90 (1.80–2.01), P < .0001 1.27 (1.19–1.34), P < .0001 1.30 (1.25–1.34), P < .0001
Mechanical ventilation utilization 1.46 (1.37–1.54), P < .0001 1.05 (0.98–1.11), P = .1637 1.15 (1.11–1.20), P < .0001
Mortality 1.21 (1.10–1.33), P < .0001 0.90 (0.81–0.99), P = .0368 1.02 (0.96–1.08), P = .6110
This table details results from the multilevel regression, subgrouped by presence of OSA and/or obesity diagnosis ([+OSA/+Ob]: both OSA and obesity diagnoses present; [+OSA/−Ob]: OSA diagnosis present, but not obesity diagnosis; [−OSA/+Ob]: obesity diagnosis present but not OSA; [−OSA/−Ob]: neither OSA nor obesity diagnoses present). [−OSA, −Ob] is the reference group. The models were adjusted for age, gender, ethnicity, insurance type, hospital location, hospital size, hospital teaching status, hospital annual colectomy volume, length of stay, elective/emergent procedure, indication for colectomy, type of procedure, year of procedure, anesthesia type, use of patient controlled analgesia, opioid utilization (oral morphine equivalents), use of nonopioid analgesics, Charlson comorbidity index, history of substance use/abuse, pain conditions, and psychiatric comorbidities. Displayed are exponentiated coefficients from the log model, which provides percent change comparing OSA/obesity categories for continuous outcomes or ORs, 95% CIs for categorical outcomes, and P values.
Abbreviations: CI, confidence intervals; ICU, intensive care unit; Ob, obesity; OR, odds ratio; OSA, obstructive sleep apnea.

The [+OSA, +Ob] group shows the relative highest risk for respiratory complications (OR [CI], 2.41 [2.28–2.56]), cardiac complications (OR [CI], 1.57 [1.45–1.69]), ICU utilization (OR [CI], 1.90 [1.80–2.01]), mechanical ventilation (OR [CI], 1.46 [1.37–1.54]), and inhospital mortality (1.21 [1.10–1.33]). The [+OSA, −Ob] and [−OSA, +Ob] show intermediary complication ORs between the groups with both diagnoses and neither. Inhospital mortality was not significantly increased in the latter 2 groups, compared to the reference.

RERI analysis revealed a supraadditive effect of 0.51 (95% CI, 0.34–0.68) for respiratory complications, 0.11 (−0.04 to 0.26) for cardiac complications, 0.30 (0.14–0.45) for ICU utilization, 0.34 (0.21–0.47) for mechanical ventilation utilization, and 0.26 (0.15–0.37) for mortality. With the exception of cardiac complications, the additive risk in the [+OSA, +Ob] group significantly exceeds the sum of the individual groups’ risk, [+OSA, −Ob] plus [−OSA, +Ob], signifying a supraadditive effect of the conditions.

Model c-statistics ranged from 0.773 (respiratory complications) to 0.861 (mechanical ventilation). ICCs ranged from 2.7% (cardiac complications) to 17.3% (ICU utilization).

In the first sensitivity analysis, we found that while incidence and odds for complications were higher in the more restricted definition of Ob (using a BMI of 40 kg/m2 as cutoff), these results mirror those from the main analysis (Supplemental Digital Content 4, Table 4, https://links.lww.com/AA/D585). Similarly, in the second sensitivity analysis including patients with no information on opioid utilization, our results did not meaningfully differ from our main analysis, thus emphasizing the robustness of our results.

DISCUSSION

In this population-based study, we found a significant increase in resource utilization and risk for complications after open colectomy when diagnoses of Ob, OSA, or both were present. Although there were overlapping CIs in the binary outcomes, effect estimates were generally smaller in the group with a sole diagnosis of OSA, intermediate in the group with a sole diagnosis of Ob, and highest in the group with both diagnoses. For most outcomes except cardiac complications, RERI analysis suggests that the combined effect of OSA and Ob exceeds the sum of the individual effects. In contrast, there was no overlap of CIs in terms of length of hospital stay and cost of hospitalization, suggesting the presence of a true difference in outcomes between these 2 groups.

Interestingly, virtually all current guidelines on the perioperative treatment of OSA patients suggest caution when prescribing opioids and recommend alternative analgesic routes.15 However, unadjusted opioid doses recorded in our cohort were higher in group [+OSA, −Ob] and almost one-third higher in the group [+OSA, +Ob] compared to the group with neither OSA nor Ob. Previous studies similarly found higher likelihood of opioid prescription among both surgical and nonsurgical OSA patients. Poeran et al3 report a higher opioid dose (334 vs 311 mg oral morphine equivalent [OME]) and higher adjusted percent opioid dose difference (+2.6% [CI, +2.1% to 3.1%]) in patients undergoing hysterectomy with and without a diagnosis of OSA, respectively. In a cohort of veterans (n = 1149,874), Chen et al16 recently found OSA patients more likely to receive opioids both in the acute (relative risk [RR], 2.02 [CI, 1.98–2.06]) and chronic (RR, 2.15 [CI, 2.09–2.22]) settings, compared to non-OSA patients. These results make it clear that against all expectations and recommendations, OSA was not associated with lower opioid use. One possible explanation is the emerging evidence that patients with combined OSA and Ob may have an altered opioid metabolism. Dalesio et al17 recently reported a higher level of inactive-to-active morphine metabolite ratio and faster morphine clearance in patients with both OSA and Ob compared to nonobese, non-OSA counterparts.

Colectomy is a surgical intervention with a high risk for major perioperative complications. While safety benefits were reported through the implementation of laparoscopic and robotic techniques, respiratory, cardiac, and neurologic complications are still commonly reported to occur in as many as 16.9%.18 Lamm et al5 performed a database analysis on 160,792 patients who underwent colectomy from 1995 to 2014. The rate of overall postcolectomy mortality decreased from 6.3% to 3.0% over this time period; however, patients who received open colectomy had a more than 3-fold higher likelihood for mortality (OR, 3.65), with the RR increasing over the years.5 Thus, strategies to improve the perioperative risk profile through identification of potentially modifiable risk factors are highly sought after. In our study, unadjusted inhospital mortality ranged from 5.4% in the group with no diagnoses to 6.5% in those with both OSA and Ob and was thus overall higher than previously reported. While inclusive nationwide data on mortality rates after colectomy are lacking, it has been suggested that numerous factors beyond individual patients’ characteristics may bear influence on it, including hospital safety-net burden19 and procedure volume.20 These higher rates may therefore be a reflection of the inclusion of multiple settings in our sample. Inhospital mortality was not significantly different in the OSA-only and Ob-only groups when compared to the reference. This finding is in keeping with much of the existing literature, where no impact of OSA or Ob could be found on mortality.21,22 It was repeatedly attempted to explain this seemingly paradoxical finding with the increased level of monitoring, preoperative optimization, and clinical attention these patients received to circumvent complications. However, the efficacy of such preventative measures is difficult to measure and predict; therefore, much uncertainty exists on how to provision them most adequately. On the other hand, mortality was increased >20% in patients with diagnoses of both OSA and Ob in our study. This finding could implicate that obese OSA patients are at higher risk for perioperative mortality than previously estimated. Moreover, it highlights the need for stringent patient selection, screening, and preoperative optimization as a strategy to mitigate the complication risk.23

OSA is an established risk factor for perioperative complications throughout numerous surgical subspecialties.1–3,24 Not least because of fear of catastrophic outcomes like hypoxic brain damage and death, as well as risk of subsequent litigation, much uncertainty prevails as to how much additional monitoring and attention are warranted in OSA patients undergoing surgery. Therefore, quantification of the added risk of OSA and associated risk factors is of particular importance.

The prevalence of Ob among the general population and those undergoing surgery is increasing rapidly, especially at the upper BMI extreme.25 Multiple studies have demonstrated a protective effect of moderately increased BMI on mortality and perioperative complication incidence, when compared to normal or low BMI (the so-called “Ob paradox”).26 However, the risk was progressively higher with increasing BMI class. In our dataset, complication risk and resource utilization were significantly higher in all groups with either OSA or Ob diagnoses compared with the control group. For patients with a sole diagnosis of OSA, we found a lower risk profile for respiratory and cardiac complications, ICU admission, and mechanical ventilation than in patients with a sole Ob diagnosis. However, overlapping CIs signify that the difference between the [+OSA, −Ob] and [−OSA, +Ob] groups is not statistically significant. Length of hospital stay and cost of hospitalization, on the other hand, were higher in the OSA group than in the Ob group. Mortality was not different among the [−OSA, −Ob], [+OSA, −Ob], and [−OSA, +Ob] groups.

It is increasingly recognized that OSA and Ob, despite their concomitant appearance in many “stereotypical” patients, are 2 distinct conditions that do not necessarily always coexist. Joosten et al27 reported that Ob accounts for only 50% of the variability in apnea-hypopnea-index (AHI) definition of OSA. Moreover, weight loss was shown to be a comparatively ineffective measure to improve OSA severity, pointing out the weak association between the conditions.28 In a polysomnography study in 803 patients, Alshehri et al29 found a prevalence of OSA of 77.7% among obese subjects and 22.3% among nonobese subjects. In patients with severe OSA, 14.7% were classified as nonobese.29 Four individual endotypes of OSA have been described, each with different underlying etiologies: anatomical disturbance of the upper respiratory tract, impaired function of the pharyngeal dilator apparatus, insufficient ventilatory control (elevated loop gain), and low arousal threshold.6 While some of these factors are independent of BMI and may underlie the development of OSA in nonobese individuals, there is currently insufficient evidence to prove a causal association. Additionally, an overlap between central sleep apnea and OSA may exist.

In addition to posing challenges of their own to the perioperative physician, OSA and Ob offer multiple zones of possible interaction that have not previously been distinguished well. Ob-hypoventilation syndrome is a loosely defined overlap area between OSA and Ob characterized by daytime hypoventilation, sleep-disordered breathing, and Ob, appearing predominantly in the morbidly obese.30 However, to the best of our knowledge, the interaction between OSA and Ob and its impact on perioperative outcomes after surgery have not been defined previously. In our study, the risk of combined OSA and Ob diagnoses exceeded the sum of the conditions’ individual risks, signifying a supraadditional composite effect. This effect was most prominent for respiratory complications, where the OR signified a 128% higher odds in [+OSA, +Ob] patients, compared to the reference.

OSA prevalence in our sample was 5.6%, compared to 17% to 43% (depending on sex and age) assumed prevalence in the general population.31 However, as a large proportion of OSA patients is thought to be undiagnosed, an accurate measure of disease burden is challenging to obtain.32 Additionally, gender aspects may play a role—female patients are even more likely to be underdiagnosed,33 which is also reflected in the lower proportion of females in the OSA groups in our sample. In inpatient populations, sleep apnea prevalence, as reported in the literature, varied wildly from 2.67% in patients undergoing hysterectomy,3 6.8% in patients undergoing spine surgery,34 to almost 48% in those undergoing treatment for cardiovascular disease.35 Similarly, Ob prevalence as evidenced by diagnosis code was 12.6% in our sample. For the US general population, the Centers for Disease Control and Prevention published a self-reported Ob prevalence of at least 20% and up to 36.1%, varying by state.36 The prevalence of morbid Ob in patients undergoing colectomy (n = 2019) was recently reported at 20.5% in a population-based study by Weber et al.37

This discrepancy between prevalence of diagnosis and prevalence of disease likely results from undercoding. Undercoding and therefore an underestimation of the true prevalence of an exposure are thought to only minimally affect population-based outcome analyses.38 However, its presence highlights some key issues with regard to how both OSA and Ob are handled in clinical and administrative practice. OSA is likely often missed due to the high complexity of arriving at a definite diagnosis (involving polysomnography).32 In contrast, Ob is trivial to diagnose by only looking at patients’ body height and weight, which are almost universally recorded at admission. Lacking incentives in terms of reimbursement, perceived unimportance, unclear responsibility, or stigma-associated could all contribute to a medical practitioner’s failure to add Ob as a diagnosis to a patient’s chart.39 Ob is increasingly recognized as a public health emergency, including in, but also far beyond the perioperative arena.40 OSA similarly has important impact in both the acute and chronic settings. Perioperative health care practitioners and surgeons should thus recognize their responsibility in identifying and correctly coding these “secondary diagnoses,” as they are often at a vantage “point of first contact” with these patients.

A number of limitations must be mentioned for this study. First, it is retrospective in nature and follows population-based, pragmatic research principles. While the large sample size, real-world data sampling, data integrity checks, and careful model calibration mitigate this limitation, it must be born in mind that, by design, it only allows for association rather than causation. Second, our analysis relies on correct diagnosis and coding of OSA and Ob. Thus, only diagnosed and coded cases of OSA and/or Ob are captured through the analysis. Third, we were not able to take into account the severity of OSA as this information is not captured in the database. We analyzed the influence of different BMI subgroups in an analysis. These results did not meaningfully differ from our main analysis but do point toward stronger effects on adverse outcomes with a higher BMI threshold.

In conclusion, both OSA and Ob are associated with adverse perioperative outcomes in patients undergoing open colectomies. With regard to the continuous outcomes (length of hospital stay and cost), Ob without OSA appears to exert a stronger risk than OSA without Ob. Excess risk analysis suggests a supraadditive effect if both OSA and Ob are present compared to either diagnosis. Given the high volume of open colectomies, its high baseline complication risk, and the large proportion of undetected OSA, further research is warranted into not only risk stratification but also effectiveness of tailored interventions. Furthermore, both OSA and Ob appear to be underreported in our sample compared to general and surgical populations. This highlights the need for stringent perioperative screening, documentation, and reporting. Awareness for both conditions’ aptitude to result in serious complications, including in but not limited to the perioperative setting, should be raised.

DISCLOSURES

Name: Ottokar Stundner, MD, MBA.

Contribution: This author helped generate the research outline, review the data, write the first manuscript, and draft and revise subsequent manuscript iterations.

Name: Nicole Zubizarreta, MPH.

Contribution: This author helped review the data, perform initial and subsequent statistical analyses, and review the manuscript.

Name: Madhu Mazumdar, PhD.

Contribution: This author helped generate the research outline, review the data, provide statistical supervision, and write and review the manuscript.

Name: Stavros G. Memtsoudis, MD, PhD, MBA.

Contribution: This author helped generate the research outline, review the data, and write and revise the manuscript.

Name: Lauren A. Wilson, MPH.

Contribution: This author helped perform initial and subsequent statistical analyses and write and review the manuscript.

Name: Hannah N. Ladenhauf, MD.

Contribution: This author helped generate the research outline, review the data, provide clinical counseling, write parts of the manuscript, and proofread the manuscript.

Name: Jashvant Poeran, MD, PhD.

Contribution: This author helped generate the research outline, provide statistical supervision, write manuscript drafts, and review the manuscript.

This manuscript was handled by: Toby N. Weingarten, MD.

REFERENCES

1. Bolden N, Posner KL, Domino KB, et al. Postoperative critical events associated with obstructive sleep apnea: results from the Society of Anesthesia and Sleep Medicine Obstructive Sleep Apnea registry. Anesth Analg. 2020;131:1032–1041.
2. Memtsoudis SG, Stundner O, Rasul R, et al. Sleep apnea and total joint arthroplasty under various types of anesthesia: a population-based study of perioperative outcomes. Reg Anesth Pain Med. 2013;38:274–281.
3. Poeran J, Mörwald EE, Zubizarreta N, Cozowicz C, Mazumdar M, Memtsoudis SG. Impact of obstructive sleep apnea on perioperative complications among patients undergoing hysterectomy: a population-based analysis. Sleep Med. 2019;56:117–122.
4. Dominguez JE, Krystal AD, Habib AS. Obstructive sleep apnea in pregnant women: a review of pregnancy outcomes and an approach to management. Anesth Analg. 2018;127:1167–1177.
5. Lamm R, Mathews SN, Yang J, et al. Patient acuity and operative technique associated with post-colectomy mortality across New York state: an analysis of 160,792 patients over 20 years. J Gastrointest Surg. 2017;21:879–884.
6. Malhotra A, Mesarwi O, Pepin JL, Owens RL. Endotypes and phenotypes in obstructive sleep apnea. Curr Opin Pulm Med. 2020;26:609–614.
7. Chen CY, Hu JM, Shen CJ, et al. Increased incidence of colorectal cancer with obstructive sleep apnea: a nationwide population-based cohort study. Sleep Med. 2020;66:15–20.
8. Premier Inc. Premier Healthcare Database White Paper: Data That Informs and Performs. 2020. Accessed February 05, 2021. https://products.premierinc.com/downloads/PremierHealthcareDatabaseWhitepaper.pdf.
9. US Department of Health and Human Services. OCR Privacy Brief: Summary of the HIPAA Privacy Rule. Office for Civil Rights, HIPAA Compliance Assistance. 2003. Accessed February 4, 2021. https://www.hhs.gov/sites/default/files/privacysummary.pdf.
10. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36:8–27.
11. McAuley D. GlobalRPH Opioid Analgesic Converter.2020. Accessed April 25, 2020. http://globalrph.com/narcoticonv.htm.
12. Lexicomp. Opioid Agonist Conversion.2020. Accessed April 20, 2020. http://online.lexi.com/lco/action/calc/calculator/70050.
13. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45:613–619.
14. Knol MJ, VanderWeele TJ, Groenwold RH, Klungel OH, Rovers MM, Grobbee DE. Estimating measures of interaction on an additive scale for preventive exposures. Eur J Epidemiol. 2011;26:433–438.
15. Memtsoudis SG, Cozowicz C, Nagappa M, et al. Society of Anesthesia and Sleep Medicine guideline on intraoperative management of adult patients with obstructive sleep apnea. Anesth Analg. 2018;127:967–987.
16. Chen K, Yaggi HK, Fiellin DA, et al. Associations between obstructive sleep apnea and prescribed opioids among veterans. Pain. 2020;161:2035–2040.
17. Dalesio NM, Lee CKK, Hendrix CW, et al. Effects of obstructive sleep apnea and obesity on morphine pharmacokinetics in children. Anesth Analg. 2020;131:876–884.
18. Sheetz KH, Norton EC, Dimick JB, Regenbogen SE. Perioperative outcomes and trends in the use of robotic colectomy for Medicare beneficiaries from 2010 through 2016. JAMA Surg. 2020;155:41–49.
19. Wang W, Hoyler MM, White RS, Tangel VE, Pryor KO. Hospital safety-net burden is associated with increased inpatient mortality and perioperative complications after colectomy. J Surg Res. 2021;259:24–33.
20. El Amrani M, Lenne X, Clement G, et al. Specificity of procedure volume and its association with postoperative mortality in digestive cancer surgery: a nationwide study of 225,752 patients. Ann Surg. 2019;270:775–782.
21. Lockhart EM, Willingham MD, Abdallah AB, et al. Obstructive sleep apnea screening and postoperative mortality in a large surgical cohort. Sleep Med. 2013;14:407–415.
22. Mullen JT, Moorman DW, Davenport DL. The obesity paradox: body mass index and outcomes in patients undergoing nonbariatric general surgery. Ann Surg. 2009;250:166–172.
23. Chung F, Memtsoudis SG, Ramachandran SK, et al. Society of Anesthesia and Sleep Medicine guidelines on preoperative screening and assessment of adult patients with obstructive sleep apnea. Anesth Analg. 2016;123:452–473.
24. Memtsoudis S, Liu SS, Ma Y, et al. Perioperative pulmonary outcomes in patients with sleep apnea after noncardiac surgery. Anesth Analg. 2011;112:113–121.
25. Hales CM, Carroll MD, Fryar CD, Ogden CL. Prevalence of obesity and severe obesity among adults: United States, 2017-2018. NCHS Data Brief. 2020;360:1–8.
26. Gupta M, Dugan A, Chacon E, et al. Detailed perioperative risk among patients with extreme obesity undergoing nonbariatric general surgery. Surgery. 2020;168:462–470.
27. Joosten SA, Hamilton GS, Naughton MT. Impact of weight loss management in OSA. Chest. 2017;152:194–203.
28. Araghi MH, Chen YF, Jagielski A, et al. Effectiveness of lifestyle interventions on obstructive sleep apnea (OSA): systematic review and meta-analysis. Sleep. 2013;36:1553–1562, 1562A–1562E.
29. Alshehri KA, Bashamakh LF, Alshamrani HM, et al. Pattern and severity of sleep apnea in a Saudi sleep center: the impact of obesity. J Family Community Med. 2019;26:127–132.
30. Mokhlesi B. Obesity hypoventilation syndrome: a state-of-the-art review. Respir Care. 2010;55:1347–1362.
31. Senaratna CV, Perret JL, Lodge CJ, et al. Prevalence of obstructive sleep apnea in the general population: a systematic review. Sleep Med Rev. 2017;34:70–81.
32. Kapur V, Strohl KP, Redline S, Iber C, O’Connor G, Nieto J. Underdiagnosis of sleep apnea syndrome in U.S. communities. Sleep Breath. 2002;6:49–54.
33. Basoglu OK, Tasbakan MS. Gender differences in clinical and polysomnographic features of obstructive sleep apnea: a clinical study of 2827 patients. Sleep Breath. 2018;22:241–249.
34. Chung AS, DiGiovanni R, Tseng S, Hustedt JW, Chutkan N. Obstructive sleep apnea in elective spine surgery: national prevalence and inpatient outcomes. Global Spine J. 2018;8:550–556.
35. Suen C, Wong J, Ryan CM, et al. Prevalence of undiagnosed obstructive sleep apnea among patients hospitalized for cardiovascular disease and associated in-hospital outcomes: a scoping review. J Clin Med. 2020;9:E989.
36. Centers for Disease Control and Prevention (CDC). Adult Obesity Prevalence Maps. 2020. Accessed February 04, 2021. https://www.cdc.gov/obesity/data/prevalence-maps.html.
37. Weber KT, Chung PJ, La Gamma N, et al. Effect of body mass index on outcomes after surgery for perforated diverticulitis. J Surg Res. 2020;247:220–226.
38. Poeran J, Cozowicz C, Chung F, Mokhlesi B, Ramachandran SK, Memtsoudis SG. Suboptimal diagnostic accuracy of obstructive sleep apnea in one database does not invalidate previous observational studies. Anesthesiology. 2016;124:1192–1193.
39. Peng M, Southern DA, Williamson T, Quan H. Under-coding of secondary conditions in coded hospital health data: impact of co-existing conditions, death status and number of codes in a record. Health Informatics J. 2017;23:260–267.
40. Ayton A, Ibrahim A. Obesity is a public health emergency. BMJ. 2019;366:l5463.

Supplemental Digital Content

Copyright © 2021 International Anesthesia Research Society