HSS Journal

, Volume 7, Issue 1, pp 2–8

Postoperative Hypoxemia in Orthopedic Patients with Obstructive Sleep Apnea

Authors

    • Department of AnesthesiologyHospital for Special Surgery
    • Department of AnesthesiologyWeill Cornell Medical College
  • Mary F. Chisholm
    • Department of AnesthesiologyWeill Cornell Medical College
    • Department of AnesthesiologyHospital for Special Surgery
  • Justin Ngeow
    • Department of AnesthesiologyHospital for Special Surgery
  • Raymond S. John
    • Department of AnesthesiologyHospital for Special Surgery
  • Pamela Shaw
    • Department of AnesthesiologyHospital for Special Surgery
  • Yan Ma
    • Department of Public HealthWeill Cornell Medical College
    • Department of Epidemiology and BiostatisticsHospital for Special Surgery
  • Stavros G. Memtsoudis
    • Department of AnesthesiologyWeill Cornell Medical College
    • Department of AnesthesiologyHospital for Special Surgery
Original Article

DOI: 10.1007/s11420-010-9165-0

Cite this article as:
Liu, S.S., Chisholm, M.F., Ngeow, J. et al. HSS Jrnl (2011) 7: 2. doi:10.1007/s11420-010-9165-0

Abstract

Criteria to determine which patients with obstructive sleep apnea (OSA) require intensive postoperative monitoring are lacking. Our postoperative OSA patients are all intensively monitored in the PACU and can provide such data. Thus, we reviewed patient records to determine incidence and risk factors for postoperative hypoxemia in OSA patients and subsequent association with postoperative complications. Five hundred twenty-seven charts of patients with OSA based on preoperative ICD-9 codes were reviewed for outcomes including episodes of hypoxemia and hypercarbia. Univariate analysis, logistic regression, and propensity analysis were performed to determine independent risk factors for hypoxemia and association with adverse outcomes. Thirty-three and 11 percent of these patients developed hypoxemia or hypercarbia. Risk factors for hypoxemia were hypercarbia, home bronchodilator use, BMI ≥35, and estimated blood loss ≥250 ml. Patients with hypoxemia had significantly more respiratory interventions and increased intensity of care. Patients with hypoxemia had significantly increased length of stay and risk of wound infections. Severe hypoxemia was associated with significantly more interventions than mild hypoxemia. Propensity analysis confirmed significant association of hypoxemia with adverse outcomes after adjustment for pre-existing risk factors. We conclude that postoperative hypoxemia in OSA patients is associated with adverse outcomes. Risk factors for hypoxemia were identified to guide allocation of monitoring resources to high-risk patients.

Keywords

orthopedic surgeryobstructive sleep apneahypoxemiapostoperative complications

Introduction

Obstructive sleep apnea (OSA) is a common sleep disorder associated with increased morbidity such as hypertension, myocardial ischemia, stroke, heart failure, pulmonary hypertension, and sudden death [11]. In the general medical population, estimates for prevalence of mild OSA range from 17 to 26% of men and 9 to 28% of women, whereas estimates for prevalence for severe OSA ranges from 9 to 14% of men and 2 to 7% of women [4, 14]. In the surgical population, excluding bariatric surgery, estimates for prevalence of OSA range from 10 to 64% [4]. Not only is OSA prevalent among surgical patients, but episodes of apnea may be more severe due to documented perioperative disturbances in sleep architecture [6] and respiratory depressant effects of postoperative analgesics [3]. Practice guidelines for preoperative management of OSA have been created [7]. However, as noted in the guidelines and recent reviews [4, 11], recommendations were primarily based on consultant opinion, as there was insufficient data for evidence-based findings. Thus, incidence and severity of postoperative hypoxemia in OSA patients has not been determined. Furthermore, association between hypoxemia and adverse outcomes is unclear.

The primary purpose of this study is to determine the frequency and severity of postoperative hypoxemia in patients with OSA. Our secondary purpose was to determine the risk factors for developing postoperative hypoxemia so as to identify patient groups that would selectively benefit from intensive monitoring. Finally, we wished to examine the association of postoperative hypoxemia with the need for respiratory interventions, increased intensity of care, incidence of wound infection, and length of hospital stay.

Patients and methods

Our institution has had a policy since December, 2005 that all patients with a preoperative diagnosis of OSA spend at least the night in the post-anesthesia care unit (PACU) for continual monitoring including pulse oximetry and frequent arterial blood gas sampling. We retrospectively reviewed charts of patients with a preoperative diagnosis of OSA to determine incidence and severity of postoperative hypoxemia, subsequent outcomes, and to determine risk factors for hypoxemia to aid in allocation of monitoring resources.

After obtaining Institutional Review Board approval, patient records from date December, 2005 to November, 2008 were identified by using ICD-9 codes 327.23 and 780.57 for OSA. Records were reviewed from 527 consecutive patients with a preoperative diagnosis of OSA. As expected in a population with diagnosis of OSA [11], our patients were predominantly male (69%), older (mean age 61), and overweight (mean BMI of 35). All patients spent at least the first postoperative night in the PACU with at least 2 l/min oxygen delivery via nasal cannula. Continual monitoring included standard non-invasive monitors and an arterial line for frequent arterial blood gas sampling. Outcomes were prospectively defined (Table 1) prior to chart review. Collected data included patient characteristics (Table 2), type of procedure, type of anesthetic, intraoperative data, postoperative course, and type of postoperative analgesia.
Table 1

Definitions for outcomes

Outcome

Definition

Mild hypoxemia

Spo2 <95% on pulse oximeter. All readings over time were considered to be the same continuous episode until interrupted by a SpO2 ≥95%. Thus, readings of continued hypoxemia over 5 min or over 5 h were all considered a single episode

Severe hypoxemia

SpO2 <90% on pulse oximeter. Same definition as above for an episode

Hypoxemia

Both mild and severe combined

Hypercarbia

PCO2 >50 mm Hg on arterial blood gas. No definition of duration due to prn nature of blood gas sampling

Increased intensity of care

Transfer from SDU to ICU

Change to 1:1 nursing

Addition of bedside observer

Additional diagnostic testing for respiratory function (e.g., chest radiograph or chest CT scan)

Consultation with other services regarding hypoxemia

Wound infection

Positive wound culture

Table 2

Selected patient characteristics

Characteristic

No hypoxemia (n = 353)

Any hypoxemia (n = 174)

Mild (n = 127)

Severe (n = 47)

Age (years, mean/SD)

60 ± 12

61 ± 12

60 ± 12

64 ± 12

Gender (M/F, %)

71/29

66/34

67/33

62/38

BMI (mean/SD)a

33 ± 7

36 ± 7

36 ± 7

35 ± 7

ASA class (%)a

1

2

0

0

0

2

46

36

35

37

3

51

63

64

61

4

1

1

1

2

Past medical history

CPAP at home (%)

36

37

40

28

Hypertension (%)

58

60

60

60

COPD (%)

5

6

6

6

Asthma (%)

11

14

14

15

Home medications

Anti-hypertensive (%)

55

60

58

68

Bronchodilatorsa (%)

13

22

21

23

Physical exam

Airway class (%) 1/2/3/4

20/55/24/2

15/52/29/4

14/52/31/3

18/52/25/5

aIdentified as potential risk factor by univariate analysis

ASA class ≥3 identified as potential risk factor by univariate analysis

The primary outcome analyzed was the occurrence of postoperative hypoxemia (mild SpO2 <95% or severe SpO2 <90%). Descriptive analysis was initially performed to determine if any differences in perioperative characteristics (Tables 2, 3, and 4) were apparent between groups that had at least one episode of hypoxemia (mild or severe) versus those who did not. Perioperative characteristics that appeared to differ between groups were then compared with univariate analysis (t test or Chi square test) to screen for potential risk factors for hypoxemia. Potential risk factors that had a significance level of p < 0.05 on univariate analysis were then used for multivariate logistic regression to determine independent risk factors for hypoxemia (mild or severe). Logistic regression was performed with both discrete and continuous variables and was adjusted for type of procedure (Appendix). BMI is a continuous variable but a cutoff value of 35 was used to create a binary variable for logistic regression. This value is the World Health Organization cutoff for mild (Class I, BMI <35) vs moderate/severe obesity (Class II, III) [21]. Our patients with a diagnosis of OSA would be expected to have some degree of obesity, thus this cutoff value was selected for external clinical relevance.
Table 3

Intraoperative characteristics

Characteristic

No hypoxemia (n = 353)

Any hypoxemia (n = 174)

Mild (n = 127)

Severe (n = 47)

Procedure

Spine (%)

20

26

26

26

Hip replacement (%)

31

21

23

17

Knee replacement (%)

35

39

36

47

Foot/ankle (%)

8

8

8

8

Other lower extremity (%)

3

2

3

0

Shoulder replacement (%)

3

2

2

2

Other upper extremity (%)

0

2

2

0

Surgeons (N)

61

44

44

30

Anesthesiaa

 General

General only (%)

19

29

30

28

General + epidural only (%)

1

1

0

2

General + nerve block (%)

1

1

2

0

Regional

Combined spinal epidural (%)

53

48

48

49

Combined spinal epidural + nerve block (%)

8

7

6

11

Spinal (%)

4

4

6

0

Spinal + nerve block (%)

1

1

0

2

Epidural (%)

8

5

4

6

Nerve block (%)

5

4

4

2

Anesthesiologists (N)

36

35

35

24

Intraoperative

IV fluidsa (ml, mean/SD)

1,855 ± 692

2,022 ± 1,024

2,010 ± 1,064

2,054 ± 916

Estimated blood lossa (ml, mean/SD)

253 ± 294

382 ± 525

397 ± 582

340 ± 322

Length of procedurea (min, mean/SD)

130 ± 66

154 ± 109

154 ± 112

152 ± 102

a Identified as potential risk factor by univariate analysis

Table 4

Postoperative analgesia characteristics

Characteristic

No hypoxemia (n = 353)

Any hypoxemia (n = 174)

Mild (n = 127)

Severe (n = 47)

Primary analgesia technique

Intravenous PCA (%)

28

32

35

26

Intravenous PCA + nerve block (%)

4

3

3

4

Epidural PCA (%)

44

39

42

30

Epidural PCA + nerve block (%)

18

17

13

28

Nerve block (%)

2

3

3

4

Continuous nerve block (%)

2

4

2

8

Oral opioid only (%)

2

2

2

0

Analgesic adjuncts

Oral opioids (%)

52

44

49

30

NSAIDs (%)

24

20

22

13

IV Opioids (%)

13

12

12

11

PCA patient controlled analgesia

Unadjusted association between hypoxemia (mild or severe) with adverse outcomes such as increased respiratory interventions, need for increased intensity of care, and major morbidities was tested with either Chi square or t test with hypoxemia as the independent variable. Then, we tested whether severity of hypoxemia was associated with increased severity or incidence of adverse outcomes by performing pairwise comparisons (no hypoxemia vs. mild, none vs. severe, and mild vs. severe) with Chi square or t test. A p < 0.05 after Bonferonni correction was considered significant.

Propensity analysis was used to examine association between hypoxemia (mild or severe) with adverse outcomes after adjustment for underlying risk factors identified above. Propensity scores were created based on risk factors identified in Tables 5 and 6 [17]. Logistic regression was performed for most dichotomous outcomes in Table 7 with propensity scores and hypoxemia as independent variables [5, 15]. However, “Administration of naloxone” and “Wound infection” had too few occurrences for analysis. Linear regression was used for “Hospital stay” as a continuous outcome variable again with propensity scores and hypoxemia as independent variables. Then, severity of hypoxemia was tested against increased severity or incidence of adverse outcomes by performing logistic or linear regression for the same outcomes with hypoxemia categorized as none, mild, or severe and controlling for propensity scores.
Table 5

Final risk factors for episodes of hypoxemia

Final independent risk factors for hypoxemia

Odds ratio (95% CI)

Hypercarbia episode

2.9 (1.6–5.2)

Home bronchodilators

2 (1.2–3.3)

BMI ≥35

1.9 (1.3–2.9)

Estimated blood loss >250 ml

1.2 (1.1–1.4)

Table 6

Incidence and risk of hypoxemia for individual and combinations of risk factors

Number of risk factors from Table 5

Risk factors combinations

Incidence of hypoxemia (% and number of patients)

Odds

0

 

No observations

0.21

1

BMI ≥35

40% (96)

0.41

EBL ≥250 ml

38.89% (77)

0.26

Home bronchodilators

46.34% (38)

0.42

Hypercarbia

57.89% (33)

0.61

2

BMI ≥35

42.86% (39)

0.5

+EBL ≥250 ml

BMI ≥35

47.73% (21)

0.81

+Home bronchodilators

BMI ≥35

60% (15)

1.16

+Hypercarbia

EBL ≥250 ml

50% (16)

0.52

+Home bronchodilators

EBL ≥250 ml

57.14% (16)

0.75

+Hypercarbia

Home bronchodilators

84.62% (11)

1.21

+Hypercarbia

3

BMI ≥35

56.25% (9)

1

+EBL ≥250 ml

+Home bronchodilators

BMI ≥35

50% (4)

1.43

+EBL ≥250 ml

+Hypercarbia

BMI ≥35

100% (6)

2.32

+Home bronchodilators + hypercarbia

EBL ≥250 ml

71.43% (5)

1.48

+Home bronchodilators + hypercarbia

4

BMI ≥35

100% (2)

2.85

+ EBL ≥250 ml

+Home bronchodilators + hypercarbia

Odds were calculated from same logistic regression as Table 5. As there were no patients with zero risk factors, odds were estimated from the zero intercept. Individual odds ratios can be calculated for any pair. For example, the odds ratio for risk factor combination (BMI + Home bronchodilators + Hypercarbia) vs. risk factor (BMI ≥35) is 2.32/0.41 = 5.66.

Table 7

Adverse outcomes associated with hypoxemia

Outcome

Total number of patients with adverse outcome

No hypoxemia (N = 353)

Hypoxemia (N = 174)

Mild (N = 127)

Severe (N = 47)

Need for CPAP

158

27%

36%

35%

38%

CPAP at home + subsequent routine PACU use

128

24%

25%

26%

21%

Emergent CPAP

15

2%

21%

14%

37%*

Change in oxygen delivery

121

9%

51%

39%

85%*

Administration of naloxone

7

1%

3%

1%

9%*

Increased intensity of care

131

15%

45%

36%

70%*

Wound infection

7

0%

3%

4%

2%

Hospital stay (days, mean/SD)

n/a

4.4 ± 2.3

6.0 ± 5.2

6.0 ± 5.9

5.9 ± 2.3

All values are significantly different (p < 0.05) between No hypoxemia vs Hypoxemia groups. Mild hypoxemia = SpO2 < 95%, Severe hypoxemia = SpO2 < 90%

*p < 0.05, Severe hypoxemia is significantly different from No hypoxemia and Mild hypoxemia

Results

Thirty-three percent of patients had at least one episode of hypoxemia, 11% of patients had at least one episode of hypercarbia, and 6% had both. In patients with hypoxemia, the median (mode) number of episodes of hypoxemia was 1 (1). The mean and standard deviation for duration of each episode was 143 ± 178 min. In patients with hypercarbia, the median (mode) number of episodes of hypercarbia was 1 (1). Twenty-three percent of patients with hypoxemia had their initial episode on POD 0, and 56% had their initial episode by POD1 (Fig. 1). Each postoperative day began at 12:01 am.
https://static-content.springer.com/image/art%3A10.1007%2Fs11420-010-9165-0/MediaObjects/11420_2010_9165_Fig1_HTML.gif
Fig. 1

This graph plots the time course of occurrence of hypoxemia episodes. POD postoperative day. The total number of patients evaluated per postoperative day is shown on bottom line (e.g., there were 527 patients on postoperative day 0)

Univariate analysis indicated that initial patient characteristic risk factors for the development of hypoxemia included BMI ≥35, ASA class ≥3, and home use of bronchodilators. Univariate analysis indicated that initial intraoperative risk factors (Table 3) included use of general anesthesia, quantity of intravenous fluids, estimated blood loss, and length of procedure. None of the postoperative analgesia modalities qualified as an initial risk factor (Table 4). The final independent risk factors for hypoxemia after logistic regression were the occurrence of hypercarbia episodes, a history of home bronchodilator use, BMI ≥35, and estimated blood loss ≥250 ml (Table 5). Table 6 displays the prevalence and risk for hypoxemia for all combinations of identified risk factors.

Patients with episodes of hypoxemia had significant association with increased adverse outcomes including need for respiratory interventions (emergent CPAP application, change in oxygen delivery, and intravenous naloxone), increased intensity of care, wound infection rates, and increased hospital length of stay (Table 7). Severe hypoxemia was associated with greater incidences of changes in oxygen delivery, intravenous naloxone, and increased intensity of care than mild hypoxemia (Table 7). Propensity analysis also indicated that hypoxemia was associated with adverse outcomes including emergent CPAP application, change in oxygen delivery, increased intensity of care, and increased hospital length of stay (Table 8) after adjustment for underlying risk factors. Outcomes of “Administration of naloxone” and “Wound infection” were not analyzed due to small number of occurrences. Propensity analysis also indicated that severe hypoxemia was associated with greater risk of need for emergent CPAP, change in oxygen delivery, and increased intensity of care than mild hypoxemia (Table 8).
Table 8

Adverse outcomes associated with hypoxemia after adjustment for underlying risk factors in Table 5 with propensity analysis

Outcome (odds ratio with 95% CI)

Hypoxemia

Mild

Severe

Need for emergent CPAP

8 (1.6–39.2)

5.6 (1.02–30.5)

17.4 (2.8–110)

Change in oxygen delivery

8.4 (5.1–13.6)

5.3 (3.1–9)

55.7 (20.1–154)

Increased intensity of care

3.9 (2.5–6.1)

2.9 (1.8–4.7)

10.7 (5–22.9)

Increased hospital stay (analyzed as continuous variable with linear regression)

p < 0.001

ND

ND

Mild hypoxemia = SpO2 < 95%. Severe hypoxemia = SpO2 < 90%. Outcomes of Administration of naloxone and Wound infection not analyzed due to small number of occurrences

ND not significantly different with increased severity of hypoxemia

Discussion

Our purpose was to determine incidence and severity of postoperative hypoxemia in patients with a diagnosis of OSA, adverse outcomes associated with hypoxemia, and to determine risk factors for hypoxemia to aid in allocation of monitoring resources.

There are several limitations to our study. The data collection was retrospective and has typical limitations in that data may have been missed and that patients may have been treated differently due to pre-existing knowledge of the diagnosis of OSA. The diagnosis of OSA was based on ICD9 codes, which are commonly used as an efficient and effective means to identify patients with OSA [19, 20]. Disadvantages of using ICD-9 codes include the possibility of mis-coding [10, 22] and that severity of OSA could not be ascertained. Our primary measure of oxygen saturation via pulse oximetry was monitored continually but recorded intermittently. Finally, our case series does not include a control group, and all associations are inferential.

Incidences of hypoxemia was 33% and hypercarbia was 11% in OSA patients after elective orthopedic surgery. Hypoxemia was significantly associated with adverse outcomes including more frequent respiratory interventions, need for more intensive monitoring, length of hospital stay, and postoperative wound infections. Our findings are in agreement with a previous smaller retrospective survey of OSA patients diagnosed with polysomnography undergoing orthopedic surgery that reported a 21% incidence of hypoxemia and 7% incidence of hypercarbia [8]. Risk of postoperative hypoxemia is likely much greater in our patients with OSA versus those without the diagnosis. The previous smaller study in orthopedic patients reported an 8% incidence of hypoxemia in matched controls without diagnosis of OSA. A much larger observational study in 24,000 patients after mixed surgery including orthopedics reported a postoperative incidence of hypoxemia of 0.8% [16] without specifying diagnoses of OSA.

Our findings are clinically relevant in that we identified combinations of risk factors to determine which patients may benefit most from allocating resources for intensive postoperative monitoring. Ability to select patients is important, as most institutions do not have the resources to intensively monitor all postoperative OSA patients. It has been estimated that each institution would need to spend at least $25,000 annually to be able to provide such monitoring [7]. Direct cause and effect between risk factors and hypoxemia are speculative, but all risk factors (episode of hypercarbia, increased BMI, home use of bronchodilators, and increased blood loss) can be reasonably linked to increased risk of respiratory compromise and resultant hypoxemia. Our results suggest that the peak period of monitoring for hypoxemia may extend through POD1. This is consistent with previous studies reporting that major disturbances in sleep architecture in normal patients occur the night prior to and immediately after surgery [6]. Sleep disturbances over the preceding two nights could easily lead to somnolence and hypoxemia by POD1. Unfortunately, episodes of hypoxemia were not limited to POD1, as many patients continued to have episodes through POD5. This finding is also consistent with previous studies in normal postoperative patients reporting continued disturbance in sleep architecture with REM rebound through POD3 [18]. Thus, sleep disturbances may routinely continue for several days after surgery, and prolonged monitoring for hypoxemia may be required in high-risk patients.

Within our study population, hypoxemia was associated with several adverse outcomes. Every hypoxemic episode is severe enough to deserve clinical interventions, and our patients with hypoxemia did have an increased need for respiratory interventions ranging from the simple such as changing oxygen delivery to the more complex such as emergent application of CPAP. Although we did not determine a direct association, it is likely that clinically significant postoperative respiratory compromise from OSA led to these interventions, and multiple smaller retrospective surveys have also reported increased need for respiratory interventions in OSA vs. non-OSA patients after a variety of surgical procedures [4]. There was an association between hypoxemia and increased intensity of care. This would be expected to increase cost of care and was likely prompted by clinically significant compromised respiratory function due to OSA. Hypoxemia was associated with more serious adverse outcomes including increased rate of wound infection and prolonged length of hospital stay. It would be reasonable to propose that hypoxemia impaired the immune system and perhaps created a more anaerobic environment promoting wound infection [12]. Alternatively, the typical dose of preoperative antibiotic may have been insufficient for our heavier patients. Hospital length of stay is multifactorial and may have been increased due to an increase in all of the above complications. Although this direct association is unclear, previous smaller retrospective surveys have also reported an increased length of hospital stay in postoperative OSA patients compared to non-OSA [4].

In conclusion, hypoxemia in patients with a diagnosis of OSA is associated with increased need for respiratory interventions, increased intensity of care, length of stay, and risk of wound infection. Severe hypoxemia was associated with increased incidences of complications. Our findings are in agreement with previous smaller surveys that also reported high incidence of hypoxemia and complications in OSA patients after orthopedic surgery. In our study characterization and capture of hypoxemia was more complete due to use of continual monitoring in the PACU. Our findings have enhanced clinical relevance in that we identified risk factors for hypoxemia to identify OSA patients that would most benefit from intensive postoperative monitoring.

Copyright information

© Hospital for Special Surgery 2010