- Email: [email protected]
Incidence and risk factors for intensive care unit admission after bariatric surgery: a multicentre population-based cohort study
British Journal of Anaesthesia, 115 (6): 873–82 (2015) doi: 10.1093/bja/aev364 Clinical Practice
C L I N I CA L P R AC T I C E
D. J. R. Morgan1, *, K. M. Ho1,3,4, J. Armstrong2 and S. Baker5 1
Department of Intensive Care Medicine, 2Department of Surgery, St John of God Hospital Subiaco, 12 Salvado Road, Subiaco, WA 6008, Australia, 3School of Population Health, University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia, 4School of Veterinary and Life Sciences, Murdoch University, 90 South Street, Murdoch, WA 6150, Australia, and 5Department of Intensive Care Medicine, Sir Charles Gairdner Hospital, Hospital Ave, Nedlands, WA 6009, Australia *Corresponding author. E-mail: [email protected]
Abstract Background: With increasing rates of bariatric surgery and the consequential involvement of increasingly complex patients, uncertainty remains regarding the use of intensive care unit (ICU) services after bariatric surgery. Our objective was to define the incidence, indications, and outcomes of patients requiring ICU admission after bariatric surgery and assess whether unplanned ICU admission could be predicted using preoperative factors. Methods: All adult bariatric surgery patients between 2007 and 2011 in Western Australia were identified from the Department of Health Data Linkage Unit database and merged with a separate database encompassing all subsequent ICU admissions pertaining to bariatric surgery. The minimal and mean follow-up periods were 12 months and 3.4 yr, respectively. Results: Of the 12 062 patients who underwent bariatric surgery during the study period, 590 patients (4.9%; 650 ICU admissions) were admitted to an ICU after their bariatric surgery. Patients admitted to the ICU were older (48 vs 43 yr, P<0.001), more likely to be male (49.7 vs 20.2%, P<0.001), and more likely to require revisional bariatric surgery (14.4 vs 7.1%, P<0.001). One hundred and seventy-six patients required an emergent unplanned ICU admission, with 51 requiring multiple ICU admissions. Revisional or open surgery, diabetes mellitus, chronic respiratory disease, and obstructive apnoea were the strongest preoperative factors associated with unplanned ICU admission. Conclusions: Intensive care unit admission after bariatric surgery was uncommon (4.9% of all patients), with 30.9% of all referrals being unplanned. A nomogram and smartphone application based on five important preoperative factors may assist anaesthetists to conduct preoperative planning for high-risk bariatric surgical patients. Key words: bariatric surgery; critical care; obesity; postoperative complications
Obesity is an increasing health burden from both an individual and societal perspective in many developed countries.1–5 The prevalence of obesity in Australia has increased to 28% of the adult population.6 As a result, many countries have seen a
dramatic upsurge in the rates of bariatric surgery, with published data now establishing it as the most effective form of treatment in achieving sustained weight loss.7 8 Although there are emerging short- and long-term data to support both its efficacy and
Accepted: August 20, 2015 © The Author 2015. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved. For Permissions, please email: [email protected]
873
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
Incidence and risk factors for intensive care unit admission after bariatric surgery: a multicentre population-based cohort study
874
| Morgan et al.
Editor’s key points • The predictive factors for and utilization of intensive care after bariatric surgery are poorly defined. • In a retrospective analysis of a large database from Western Australia, all bariatric surgery patients and those requiring ICU admission were identified and analysed. • Of 12 062 bariatric surgery patients, 4.9% required ICU admission, of which 31% were unplanned. • A predictive nomogram and smartphone application based on five preoperative risk factors may help to identify highrisk patients.
Methods Ethics approval Prior ethics committee approval was granted by the Western Australian Department of Health (#2012/21) and all eight participating hospitals, including a waiver of patient consent owing to the large cohort size in addition to the observational, non-interventional, low-risk study design.
Patient population and conduct of study This was a retrospective, multicentre, population-based cohort study including all adult patients who underwent elective bariatric surgery in the state of Western Australia throughout a 5 yr period between 2007 and 2011. Patients admitted to any intensivist-led ICU after bariatric surgery formed a distinct subgroup. All outcomes were censored on December 31, 2012 unless the patients died beforehand, allowing for a minimum of 12 months follow-up after bariatric surgery. In 2011, Western Australia had a population of 2.39 million, comprising 10.4% of the total Australian population.15
Data handling Bariatric surgery hospitalizations Patients who had undergone bariatric surgery and their associated co-morbidity data were identified using the Western Australian Department of Health Data Linkage Unit (DLU) databases that systematically connect and update all available health
Bariatric surgery intensive care admissions Study data for patients admitted to an ICU were sourced initially by the Western Australian Department of Health DLU and then crossreferenced by screening all local ICU databases in Western Australia manually to ensure that every ICU admission pertaining to bariatric surgery or its complications was included. Only bariatric patients admitted to one of eight hospitals with a specialist intensivistrun ‘closed’ ICU where considered as requiring critical care support.
Clinical end points The incidence of ICU admission after bariatric surgery and differences in characteristics between those who were admitted to the ICU and those solely managed in the ward were the primary
12 264 patients screened for bariatric surgery between 2007 and 2011 in the whole state of Western Australia 202 patients excluded who had comorbid ICD-10 diagnostic codes for gastric cancer, peptic ulcer disease, or oesophageal varices
12 062 patients with an index bariatric procedure between 2007 and 2011 (mean follow-up 3.4 years)
Linkage with Western Australian death registry
11 472 patients (95.1%) admitted to regular hospital wards following bariatric surgery
35 180 intensive care unit (ICU)admissions for all Western Australian public & private ICUs between 2007 and 2011 merged with bariatric data
590 bariatric patients (4.9% of all bariatric surgery cases, 1.7% of all ICU admissions) admitted to one of eight ICUs in Western Australia following bariatric surgery electively or as a result of a complication after bariatric surgery
414 electively admitted (3.4%) to an ICU after bariatric surgery
176 bariatric patients (1.5%) who had an unplanned ICU admission
Fig 1 Flow chart showing the inclusion of patients and their disposition in relationship to the index bariatric operation and intensive care admission during the study period. ICD, International Diagnostic Code; ICU, intensive care unit.
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
its safety,9 10 the sustained increase in the incidence of bariatric surgery and improving surgical techniques have inevitably led to more patients with multiple co-morbidities. This inescapably means that there will be a growing number of patients who may require either elective or emergent critical care support after bariatric surgery. Despite the wealth of published research relating to bariatric surgery during the last 20 yr, there is a paucity of data describing the relationship between levels of postoperative care [ postanaesthesia, high-dependency, or intensive care unit (ICU)] and bariatric surgery. Currently, data are limited to a handful of older single-centre studies, with admission rates for a higher level of postoperative care after bariatric surgery varying widely between 4 and 21%.11–14 To date, there are no robust epidemiological data to assist clinicians in planning for critical care service requirements after bariatric surgery. Using the well-established population-based data linkage system in Western Australia, this study aimed to define the incidence, risk factors, and outcomes of patients who were admitted to an ICU after bariatric surgery. Specifically, we also hypothesized that unplanned ICU admissions after bariatric surgery could be predicted using preoperative factors.
data for every individual within the entire state of Western Australia using the 10th edition International Classification of Diseases (ICD-10) diagnostic and procedure codes.16 The accuracy of this database has been previously validated and formed the basis of a large number of population-based epidemiological studies in Western Australia.17–19 In addition, the techniques used by the DLU to obtain linkage of health data across different jurisdictions has also been applied to other states in Australia and was found to be capable of providing high-quality data for research purposes (see Fig. 1 and Supplementary Online Information 01: supplemental Methodology).20
Admission to ICU after bariatric surgery
outcomes. Secondary outcomes included identification of preoperative risk factors that were associated with unplanned ICU admissions and the surgical, anaesthetic, and medical complications that necessitated or involved ICU care.
Statistical analysis
was considered as a revisional surgery when it was performed for unresolved bariatric issues (e.g. failed to lose weight) that were not directly related to surgical complications from previous bariatric surgical procedures. During the multivariate modelling process, predictors were removed by a backward stepwise process according to the Design Library syntax recommended by Harrell Jr21 to improve the precision and utility of the final parsimony predictive model. The discrimination and calibration of the predictive model were assessed by area under the receiver operating characteristic curve and a calibration plot with bias corrected by a bootstrapping technique (n=200), respectively. The relative importance of each predictor in explaining the variability in the risk of unplanned ICU admission was assessed using the χ2 statistic minus the degrees of freedom. A nomogram based on the final parsimony predictive model was constructed to assist clinical decision-making.21 22 Missing data for height, and thus BMI, occurred in 4.1% of the ICU bariatric patients, and these patients were excluded when the differences in BMI between emergency and elective ICU admissions were analysed. All analyses were performed using SPSS for Windows (IBM Corp. Released 2013. IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY: IBM Corp.) or S-Plus (version 8.0; Insightful Corp., Seattle, WA, USA). A two-tailed P<0.05 was considered statistically significant. See Supplementary Online Information 01: supplemental Methodology for further details about an expanded participating hospital list; manuscript definitions; Western Australian Department of Health Data Linkage Unit; participants; and the use of International Classification of Diseases (ICD-10) coding.
Table 1 Baseline comparison of 12 062 bariatric patients stratified by need for intensive care unit admission after bariatric surgery in Western Australia between 2007 and 2011. ICU, intensive care unit; IQR, interquartile range; LOS, length of stay. aPatients re-admitted to hospital with unresolved bariatric issues Characteristic Patient numbers per yr [n (%)] 2007 2008 2009 2010 2011 Characteristics [, median, IQR] Age (mean; yr) Gender (male) [n (%)] Weight (mean; kg) BMI (mean; kg m−2) Funding for surgery [n (%)] Private Public Type of surgery [n (%)] Laparoscopic Open Original Revisional Bariatric re-hospitalizationsa Hospital LOS (mean; days) [, median, IQR] Deaths to December 2012 [n (%)]
Bariatric patients requiring ICU (n=590)
Bariatric patients not requiring ICU (n=11 472)
80 (4.0) 168 (6.8) 138 (5.1) 101 (4.2) 104 (4.2)
1901 (96.0) 2293 (93.2) 2574 (94.9) 2316 (95.8) 2388 (95.8)
48 [11.3, 49, 40–57] 294 (49.7) 136 [31.6, 133, 114–156] 46.6 [9.2, 46, 39.8–52.2]
43 [11.65, 43, 34–52] 2021 (20.2) Not available
515 (87.3) 75 (12.7)
10 966 (95.6) 506 (4.4)
555 (94.1) 35 (5.9) 505 (85.6) 85 (14.4) 216 (36.6) 8.6 [14.7, 4, 3–8] 14 (2.4)
11 122 (96.9) 350 (3.1) 10 665 (92.9) 817 (7.1) 686 (6.0) 2.6 [1.5, 2, 2–3] 8 (0.1)
P-value
<0.001
Not available
<0.001
<0.001 <0.001 <0.001 <0.001 <0.001
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
Categorical variables were described in absolute numbers and percentages, with dichotic comparisons performed using the χ2 test. Continuous variables were described as mean, standard deviation (), median, and interquartile range (IQR). Comparisons between normally distributed data were performed using Student’s unpaired t-test, whereas non-parametric data were compared using the Mann–Whitney U-test. Multivariate logistic regression was used to assess associations between patient characteristics, premorbid or perioperative factors, and different types of complications, all reported as odds ratios (ORs) with 95% confidence intervals (CIs). Clustering of unplanned ICU admissions within certain study centres was assessed in a separate sensitivity analysis using the generalized estimating equation. In assessing preoperative risk factors for unplanned ICU admissions, a multivariate analysis was undertaken comparing all unplanned ICU admissions with all bariatric surgical patients not requiring ICU across Western Australia. Our intention was to identify errors in postoperative pathway planning using only biologically plausible preoperative factors. These factors included age, gender, revisional vs primary surgery, open vs closed surgery, and chronic medical conditions before surgery. Bariatric surgery
| 875
876
| Morgan et al.
Results Differences between bariatric patients requiring and not requiring admission to ICU
Of the 650 ICU admissions after bariatric surgery, 449 admissions (69.1%) were electively referred to the ICU after surgery, whereas 201 (30.9%) were for emergent indications. The predominant clinical indication for ICU admission in most patients was surgically related (elective or emergent ; 89.0%), whereas anaesthetic (8.8%) and medical complications (2.2%) accounted for a small proportion of the main reason for ICU admission. Significant differences in patient characteristics existed between the two groups, with emergently admitted bariatric ICU patients more likely to be female and to have a lower BMI than elective admissions (Table 2). Emergent ICU admissions were more frequently associated with surgical complications, particularly when the problems arose in a delayed fashion after hospital discharge. Anaesthetic and medical complications were likewise more common among emergent ICU admissions (Table 3). Of the 298 recorded surgical complications, 21.5% occurred within the operating theatre and another 9.1% within 48 h of surgery. Significantly, 69.4% of all surgical complications occurred beyond the usual time frame of an elective ICU stay (<48 h) after bariatric surgery.
Table 2 Patient characteristics, co-morbidities, preoperative consultation and initial bariatric operation for 590 bariatric patients undergoing 650 admissions to an ICU between 2007 and 2011. CPAP, continuous positive airway pressure; COPD, chronic obstructive pulmonary disease; ICU, intensive care unit; IQR, interquartile range; LAGB, laparoscopic adjustable gastric band. aSpecialist review before day of surgery. b Includes internal physician, haematologist, nephrologist, immunologist, endocrinologist, clinical microbiologist, psychiatrist, or vascular surgeon. cBased on 650 ICU admissions in 590 patients. dRoux-en-Y, mini gastric bypass, pancreatobiliary bypass, gastroplasty, and gastric balloon procedures. eIndex hospitalization was the hospitalization related to the first bariatric surgery and excludes patients admitted to the ICU upon hospital re-admission for ongoing bariatric complications Characteristics Patient characteristics [, median, IQR] Age (mean; yr) Gender (male) [n (%)] Weight (mean; kg) BMI (mean; kg m−2) Premorbid co-morbidities [n (%)] Obstructive sleep apnoea CPAP Asthma/COPD Smoker Ischaemic heart disease Hypertension ≥3 Antihypertensive drugs Diabetes mellitus Preoperative consultationa [n (%)] Cardiologist review Pulmonologist review Anaesthetist review Other specialist reviewb ASA classificationc [, median, IQR] Primary bariatric operation [n (%)] LAGB Sleeve gastrectomy Other bariatric surgeryd Revisional surgery Open surgery Total ICU admissions (n) ICU during index hospitalizatione [n (%)]
P-value
Elective ICU patients (n=414)
Emergent ICU patients (n=176)
48.2 [11.4, 49, 40–57] 231 (55.9) 142 [30, 140, 122–160] 48.2 [9.1, 47.6, 41–54]
47.1 [11.1, 48, 39–55] 62 (35.2) 120 [30, 118, 98–139] 42.3 [8.1, 41.4, 36–49]
151 (36.6) 162 (39.2) 83 (20.1) 51 (12.3) 44 (10.7) 227 (55.0) 54 (13.0) 140 (33.9)
41 (23.3) 28 (15.9) 30 (17.0) 33 (18.8) 9 (5.1) 67 (38.1) 11 (6.2) 36 (20.5)
0.002 <0.001 0.43 0.05 0.04 <0.001 0.01 0.001
23 (5.1) 44 (9.8) 113 (25.2) 10 (2.2) 3.2 [0.7, 3, 3–4]
24 (11.9) 5 (2.5) 33 (16.4) 3 (1.5) 3.8 [2.2, 3, 3–4]
0.002 0.001 0.02 0.76 <0.001
231 (55.9) 174 (42.1) 8 (2.0) 43 (10.4) 17 (4.1) 449 393 (87.5)
95 (54.0) 71 (40.3) 10 (5.7) 42 (23.9) 18 (10.2) 201 92 (45.8)
0.28 <0.001 <0.001 <0.001
0.72 0.72 0.03 0.001 0.007 <0.001
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
In the 5 yr between 2007 and 2011, there were 12 062 index cases of bariatric surgery in Western Australia (mean follow-up 3.4 yr after index bariatric surgery). Of these, 590 (4.9%) patients required 650 ICU admissions (Fig. 1), giving a rate of 22.6 ICU admissions per 1000 patient-yr after bariatric surgery (95% CI 20.9–24.5). During the same period in Western Australia, there were 5.8 bariatric ICU admissions per 100 000 person-yr, 107 bariatric operations per 100 000 person-yr, and 309 overall ICU admissions per 100 000 person-yr. All-cause long-term mortality after bariatric surgery was significantly higher in patients who required ICU admission (6.2 vs 0.2 deaths per 1000 patient-yr, P<0.001). When compared with bariatric surgery patients not requiring critical care support, patients admitted to ICU were an average of 5 yr older, more likely to be male, and more likely to have diabetes mellitus. In addition, those patients requiring admission to ICU were more likely to have undergone revisional or open bariatric surgery, resulting in a longer hospital length of stay and increased rehospitalizations for unresolved bariatric issues (Table 1).
Differences between elective and emergent ICU admissions
Admission to ICU after bariatric surgery
Bariatric surgery patients admitted to the ICU in an emergent manner were significantly more likely to require a raft of ICU interventions (Table 4), including higher rates of intubation, vasopressor infusions, and blood transfusions. Likewise, total parental nutrition, prolonged durations of antibiotics, and increased radiological investigations were all more common in the emergent ICU admission group. As expected, both ICU and hospital length of stay were significantly longer for patients admitted to the ICU for emergent reasons compared with elective procedures.
Risk factors for complications and recurrent ICU admissions
an increased BMI was associated with fewer surgical complications (OR=0.94, 95% CI 0.92–0.96). In the subgroup of 51 patients requiring multiple ICU admissions (Supplementary Online Information 01: supplemental Table S1), gastrointestinal leak, abscess, or both was the only predictor associated with an increased risk of recurrent ICU admissions (OR=7.46, 95% CI 5.7–17.0).
Predictors for unplanned ICU admissions Of the 12 062 bariatric surgery patients included, emergent ICU admissions occurred in 176 (1.5%) patients. Multivariate analysis showed that revisional or open surgery, diabetes mellitus, chronic respiratory disease, chronic cardiovascular disease, chronic renal disease, and obstructive sleep apnoea (OSA) were all associated with emergent ICU admission (Supplementary Online Information 01: supplemental Table S2). Of these seven significant variables, revisional surgery was most important, explaining ∼36.5% of the variability in the risk of requiring unplanned ICU
Table 3 Bariatric complications associated with 650 ICU admissions between 2007 and 2011 stratified by urgency of ICU admission.a AKIN, acute kidney injury network; DVT, deep vein thrombosis; GIT, gastrointestinal tract; ICU, intensive care unit; PE, pulmonary embolism. a590 bariatric patients, of whom 51 patients had multiple ICU admissions. bFifteen patients admitted emergently to the ICU had no defined complications but had unpredictably prolonged or technically difficult operations, or unappreciated co-morbidities before surgery. c Complications occurring after theatre and before initial hospital discharge. dComplications requiring subsequent re-hospitalization. e Complications including gastric band malfunction, port infection, or failure to lose weight. fUnplanned additional surgical interventions. g Complications are as follows: obstructed airway, difficult intubation, hypoventilation, hypoxia, anaphylaxis, bronchospasm, gas embolism, cardiac arrest, respiratory arrest, poorly controlled hypertension, and hypotension. hComplications are as follows: non-cardiac chest pain, neurological, metabolic, hepatic, hypertension, congestive cardiac failure, disseminated intravascular coagulopathy, diabetic ketoacidosis, pancreatitis, severe eczema, diarrhoea, narcotization, undiagnosed sleep apnoea, and undiagnosed cancer Characteristic Surgical complications (timing) [n (%)] Intraoperative complication Postoperative complicationsc Delayed complicationsd Surgical complications (type) [n (%)] Admission with any complication GIT obstruction GIT perforation, leak, abscess Haemorrhage Other surgical complicationse Positive microbiology cultures Surgical interventionsf [n (%)] Drain (unplanned) Endoscopy Laparoscopy Laparotomy Thoracotomy Anaesthetic complications [n (%)] Any anaesthetic complicationg Medical complications [n (%)] Admissions with any complication Cardiac arrhythmia Left pleural effusion Lower respiratory tract infection Aspiration pneumonitis Empyema Asthma Septic shock Acute coronary syndrome AKIN stage 3 DVT, PE, or both Other medical complicationsh
Elective ICU admission (n=449)
Emergent ICU admission (n=201)b
P-value
42 (9.4) 19 (4.2) 55 (12.2)
22 (10.9) 63 (31.3) 97 (48.5)
0.57 <0.001 <0.001
95 (21.2) 19 (4.2) 23 (5.1) 21 (4.7) 29 (6.5) 8 (1.8)
149 (74.1) 21 (10.4) 94 (46.8) 43 (21.4) 30 (14.9) 66 (32.8)
<0.001 0.004 <0.001 <0.001 0.001 <0.001
8 (1.8) 22 (4.9) 61 (13.6) 34 (7.6) 2 (0.4)
35 (17.4) 34 (16.9) 89 (44.3) 85 (42.3) 7 (3.5)
<0.001 <0.001 <0.001 <0.001 0.005
8 (1.8)
55 (27.4)
<0.001
64 (14.3) 13 (2.9) 2 (0.4) 4 (0.9) 0 (0) 0 (0) 6 (1.3) 2 (0.4) 3 (0.7) 0 (0) 2 (0.4) 34 (7.6)
128 (63.7) 19 (9.5) 27 (13.4) 3 (1.5) 2 (1.0) 8 (4.0) 8 (4.0) 30 (14.9) 5 (2.5) 16 (8.0) 7 (3.5) 47 (23.4)
<0.001 <0.001 <0.001 0.68 0.10 <0.001 0.04 <0.001 0.11 <0.001 0.005 <0.001
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
In the multivariate analysis of ICU admissions (Table 5), all-cause surgical complications were associated with female gender (OR=1.53, 95% CI 1.06–2.21), revisional surgery (OR=4.76, 95% CI 2.79–8.12) and open surgery (OR=2.78, 95% CI 1.06–7.27), whereas
| 877
878
| Morgan et al.
Table 4 Intensive care unit referral origins, admission APACHE II scores, interventions, and outcomes for 650 ICU admissions after bariatric surgery between 2007 and 2011. APACHE, acute physiology and chronic health evaluation; CRRT, continuous renal replacement therapy; CT, computerized tomography; ED, emergency department; ICU, intensive care unit; IQR, interquartile range; LOS, length of stay; NIV, noninvasive ventilation; TPN, total parental nutrition. aTotal intubation time divided by total number of intubated patients. bBased on 590 ICU patients. cPatients re-admitted to hospital for unresolved bariatric issues Characteristics
Emergent ICU admission (n=201)
P-value
4.6 [2.9, 4, 2–6] 0 450 (100) 2 (0.4) 14 (3.1) 4.0 [8.1, 2, 1–3] 12 (2.7) 9 (2.0) 7 (1.6) 1 (0.2) 12 (2.7) 5 (1.1) 2 (0.4) 17 (3.8) 5 (1.1) 26 (5.8) 31 (6.9)
6.2 [4.4, 5, 3–9] 54 (26.9) 147 (73.1) 42 (21.0) 97 (48.3) 6.0 [12.4, 2, 1–5] 16 (8.0) 46 (22.9) 7 (3.5) 7 (3.5) 71 (35.3) 47 (23.4) 17 (8.5) 107 (53.2) 54 (26.9) 57 (28.4) 126 (62.7)
<0.001
2.2 [2.0, 2, 2–2] 4.8 [7.1, 4, 2–5] 0 0 106 (23.6)
5.5 [9.7, 3, 2–4] 17.2 [22, 10, 4–21] 3 (1.7) 3 (1.7) 109 (54.2)
admission (Supplementary Online Information 02: supplemental Fig. S1). In the sensitivity analysis using a generalized estimating equation, significant clustering of unplanned ICU admissions by individual hospital sites was not observed. The ability of the final parsimony predictive model to discriminate between patients requiring and not requiring unplanned ICU admission was reasonable (area under the receiver operating characteristic curve=0.820, 95% CI 0.78–0.86). The calibration of the model was, however, not perfect when the predicted risks of requiring unplanned ICU were greater than 30–40% (Supplementary Online Information 03: supplemental Fig. S2). A nomogram based on the five most important preoperative factors to estimate the probability of unplanned ICU admission after bariatric surgery is presented in Fig. 2 and Supplementary Online Information 04: supplemental Fig. S3. The score assigned to revision surgery is 100 points, chronic respiratory disease is 75 points, OSA is 60 points, open surgery is 55 points, and diabetes mellitus is 35 points. A total of 110 points (e.g. diabetes mellitus +respiratory disease), 135 points (e.g. OSA+respiratory disease), and >160 points (e.g. revision surgery+either respiratory disease or OSA) predicts 10, 20, and >30% risk of unplanned ICU admission, respectively. An Android application (http://www. appsgeyser.com/1947332) for smart phones and a Windows PC application (http://tinyurl.com/ps7gptv) have been created to aid with preoperative risk prediction.
<0.001 <0.001 <0.001 <0.001 0.50 0.005 <0.001 0.14 0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.03 0.03 <0.001
Discussion This population-based multicentre study showed that a small (4.9%) but important proportion of patients required ICU support after bariatric surgery. Such patients tended to be older and were more likely to be male, have diabetes mellitus, and have undergone either revisional or open bariatric procedures. Revisional or open surgery, diabetes mellitus, chronic respiratory disease, and OSA were the strongest predictors for emergent unplanned ICU admission. Overall long-term mortality after bariatric surgery was very low. This is the first large study to describe the relationship between bariatric surgery and intensive care use and offers a significantly longer follow-up period than most other studies documenting complications. While other bariatric studies have defined factors contributing towards surgical complications,23 our study uniquely describes the increased complexity, resource utilization, cost, and breadth of complications associated with ICU care above that of complications managed at a ward level. Data regarding clinical care pathways that encompass intensive care after bariatric surgery are sparse.12 13 The previously published Obesity Surgery Mortality Risk Score (OS-MRS) refers only to 90 day mortality, a very rare event after bariatric surgery, with no reference to the more common occurrence of ICU utilization.24 Furthermore, the OS-MRS was conceived before revisional bariatric surgery became commonplace and thus does not
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
Intensive care admission Admission APACHE II score [, median, IQR] Origin (ward or ED) [n (%)] Origin (theatre) [n (%)] Interhospital transfer [n (%)] Intubation [n (%)] Intubation timea (days) [, median, IQR] Unplanned NIV [n (%)] Vasopressor infusion [n (%)] Vasodilator infusion [n (%)] CRRT [n (%)] Blood transfusion [n (%)] TPN [n (%)] Tracheostomy [n (%)] CT abdomen [n (%)] CT chest [n (%)] Barium swallow [n (%)] Antibiotics >72 h [n (%)] Outcomes ICU LOS (days) [, median, IQR] Hospital LOS (days) [, median, IQR] ICU mortalityb [n (%)] Hospital mortalityb [n (%)] Bariatric re-hospitalizationsb,c [n (%)]
Elective ICU admission (n=449)
Admission to ICU after bariatric surgery
| 879
Table 5 Multivariate analysis showing factors associated with surgical, anaesthetic, and medical complications occurring in 650 ICU adminssions after bariatric surgery. CI, confidence interval; OR, odds ratio. aP-value for adjusted OR. bComplications as listed in Table 3. c Specialist review before day of elective bariatric surgery Characteristics
Unadjusted OR [95% CI]
0
10
20
30
40
50
60
P-valuea
0.98 [0.96–1.00] 1.53 [1.06–2.21] 0.94 [0.92–0.96] 0.84 [0.55–1.28] 4.76 [2.79–8.12] 2.78 [1.06–7.27]
0.04 0.02 <0.001 0.42 <0.001 0.04
1.01 [0.98–1.03] 1.10 [0.65–1.85] 0.99 [0.96–1.02] 0.65 [0.38–1.11] 1.82 [0.97–3.40] 0.95 [0.35–2.56] 0.95 [0.51–1.77] 0.54 [0.29–1.02]
0.65 0.72 0.46 0.11 0.06 0.91 0.87 0.06
1.01 [0.99–1.02] 2.08 [1.45–3.03] 0.95 [0.93–0.97] 0.73 [0.48–1.12] 1.80 [0.99–2.68] 0.86 [0.41–1.83] 0.65 [0.41–1.05] 1.48 [0.87–2.50] 0.67 [0.29–1.55]
0.56 <0.001 <0.001 0.15 0.05 0.70 0.08 0.14 0.34
70
80
90
100
Points Yes RESP No Yes DM No Yes OSA No Yes Revision No Yes Open No Total points 0
40
80
120
160
200
240
280
Linear predictor –5
–4
–3
–2
–1
0
0.5
1
1.5
2
2.5
Probability 0.1
0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.9
Fig 2 Nomogram using the five most important preoperative predictors to estimate the probability of emergent, unplanned intensive care unit admission after bariatric surgery. DM, diabetes mellitus; Open, open surgery; OSA, obstructive sleep apnoea; RESP, chronic respiratory conditions other than OS; Revision, revisional surgery.
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
Potential risk factors associated with all surgical complicationsb Age 0.99 [0.98–1.01] Gender (female) 1.97 [1.42–2.74] BMI 0.94 [0.92–0.96] Diabetes mellitus 0.77 [0.54–1.10] Revisional surgery 3.77 [2.43–5.84] Open surgery 3.96 [1.94–8.08] Potential risk factors associated with all anaesthetic complicationsb Age 1.00 [0.98–1.02] Gender (male) 0.95 [0.59–1.54] BMI 0.98 [0.96–1.01] Obstructive sleep apnoea 0.62 [0.38–1.01] Smoking 1.67 [0.91–3.05] Pulmonologist consultationc 0.99 [0.41–2.42] Anaesthetist consultationc 0.85 [0.48–1.53] ASA grade IV classification 0.50 [0.09–2.73] Potential risk factors associated with all medical complicationsb Age 1.01 [0.99–1.03] Gender (female) 1.44 [0.96–2.14] BMI 0.97 [0.95–0.99] Diabetes mellitus 0.79 [0.50–1.23] Smoking 1.51 [0.90–2.54] Pulmonologist consultationc 1.35 [0.67–2.74] Anaesthetist consultationc 0.83 [0.51–1.35] Revisional surgery 1.41 [0.78–2.54] Open surgery 0.69 [0.26–1.82]
Adjusted OR [95% CI]
880
| Morgan et al.
was still lower than the previously reported range of 14.4–28.2% in two smaller and older ICU studies.12 33 Hellings and colleagues reported pulmonary co-morbidity was associated with increased risk of prolonged mechanical ventilation >24 h in 60 bariatric patients admitted to ICU.11 Importantly, amongst our ICU cohort were 24 patients initially diagnosed as lower respiratory tract infections, of whom 14 in reality had gastric leaks with atelectasis, delaying definitive surgical management by as much as 48 h. Hence, it may be prudent for ICU physicians to consider intra-abdominal sepsis from gastric leaks when patients present with clinical signs of sepsis after bariatric surgery, even when there are radiological changes consistent with lower respiratory tract infection. This study is limited by its retrospective, observational nature and possible selection bias and confounding. The use of an administrative database is constrained by the coding conventions pertinent to the collection of large administrative data sets not specifically designed for individual studies. As such, we could not confirm whether one particular subtype of bariatric surgical procedure was more likely to require ICU admission than the others, nor collect heights and weights for the overall bariatric cohort. Additionally, because there is limited public health-care funding for bariatric surgery in Australia, most of our cohort were admitted to private health-care facilities where ICU accessibility was less constrained than in the public health-care system. This may limit the generalizability of our results outside Australia or on patient populations predominately using government-funded health organizations. Our analysis was also constrained by the difficulty of reliably anticipating intraoperative technical complications. Nevertheless, the strength of revisional surgery as a strongest predictive factor suggests that this is a good surrogate for intraoperative technical and surgical difficulties and complications, probably owing to adhesions, altered anatomy, and previous tissue injury.
Conclusions In a multidisciplinary approach to bariatric surgery, intensive care plays a succinct but important role, with one in 20 Western Australians found to require ICU admission after bariatric surgery. This is a lower rate than most published literature to date. Revisional or open surgery, diabetes mellitus, chronic respiratory disease, and OSA were the strongest predictors associated with subsequent unplanned ICU admission after bariatric surgery. A nomogram smart phone application based on these five most important preoperative factors may assist clinicians in identifying bariatric patients who should be operated on in a facility that has an ICU or electively admitted to an ICU after a procedure.
Authors’ contributions Study concept: D.J.R.M., K.M.H. Study design: D.J.R.M., K.M.H., J.A. Data collection: D.J.R.M., K.M.H., S.B. Data analysis: D.J.R.M. Statistical analysis: K.M.H. Drafting the manuscript: D.J.R.M., K.M.H. Manuscript review: J.A., S.B. Final approval: D.J.R.M., K.M.H., J.A., S.B.
Supplementary material Supplementary material is available at British Journal of Anaesthesia online.
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
acknowledge this factor as an important element in postoperative outcomes. In a recent guide to perioperative management of the obese surgical patient, information pertaining to postoperative planning is unsupported by detailed references for ICU clinical pathways.25 Similar guidelines from other major international bariatric bodies are constrained to discussing surgical volume and generic descriptions of supportive staff requirements, with no discussion of critical care support.26 27 Our data on the differences between elective and emergent ICU admission after bariatric surgery improve our understanding of the requirements of critical care services related to bariatric surgery. We chose to assess preoperative factors in predicting unplanned ICU admissions because such a predictive model would be more clinically useful than models that rely on postoperative factors or complications, when unplanned ICU admission would be imminent and inevitable. We anticipated that a predictive model based on preoperative factors only will be imperfect because unexpected events that may require ICU admission do occur during the intraoperative and postoperative period. Nevertheless, our model appeared to have a reasonable calibration up to a predicted risk of 30–40%. If our model is further validated by other centres, the nomogram based on these five most important preoperative factors may be useful in assisting clinicians to decide which bariatric patients should be operated on in a facility that has an ICU or electively admitted to an ICU after a bariatric procedure. Importantly, both revisional and open surgery were risk factors for increased risk of unplanned ICU admission and surgical complications, suggesting that these surgical techniques are likely to have an impact on the risk of requiring critical care services. Comparisons within the literature describing overall ICU utilization after bariatric surgery are constrained by the limited number of small, single-centre publications that often report older surgical techniques, including a much higher proportion of open procedures. Our data highlight some important differences from older studies. The incidence of ICU admission is much lower in our cohort than in many other reports (8–24%),11–13 28 despite a dramatic increase in bariatric procedures in Western Australia throughout the past decade. Additionally, bariatric surgery has remained largely the domain of private hospitals in Western Australia, where ICU beds are more accessible, ranging from 62 to 84% of bariatric procedures between 1988 and 2004 and increasing to 95% of our cohort.29 Our lower incidence of ICU utilization possibly reflects the improved surgical and anaesthetic techniques combined with an increased trend towards laparoscopic surgery, resulting in only 3.2% of our state-wide cohort requiring open procedures. This explains why the rate of complications was lower in our patients than in other reports even with a relatively long follow-up period for complications in this study.30 31 Despite the differences in surgical techniques, our study confirms that older male patients with multiple co-morbidities remain more likely to be admitted to an ICU after bariatric surgery.11–13 The apparent protective effect of a higher BMI in both surgical and medical complications appears counterintuitive. This possibly reflects a greater preponderance to admit patients with a higher BMI electively to the ICU and thus avert complications developing. It may also relate to the extended follow-up interval in our study, reflecting more the effects of long-term complications and repeated bariatric-related re-hospitalizations from surgical complications, resulting in both a decreased caloric intake and increased catabolism of ongoing illness. Respiratory complications are one of the main concerns in obese perioperative patients, with major adverse respiratory events reported to occur in ∼2% of large overall bariatric cohorts.32 In our cohort, the respiratory complication rate (12.5%)
Admission to ICU after bariatric surgery
Acknowledgements The authors thank the Obesity Surgery Society of Australia and New Zealand for their support in obtaining ethics approval from the Western Australian Department of Health (http://www. ossanz.com.au/).
Declaration of interest None declared.
Funding
References 1. Belle SH, Chapman W, Courcoulas AP, et al. LABS Writing Group for the LABS Consortium. Relationship of body mass index with demographic and clinical characteristics in the Longitudinal Assessment of Bariatric Surgery (LABS). Surg Obes Relat Dis 2008; 4: 474–80 2. Dunstan DW, Zimmet PZ, Welborn TA, et al. The rising prevalence of diabetes and impaired glucose tolerance: the Australian Diabetes, Obesity and Lifestyle Study. Diabetes Care 2002; 25: 829–34 3. Colagiuri S, Lee CMY, Colagiuri R, et al. The cost of overweight and obesity in Australia. Med J Aust 2010; 192: 260–4 4. Haslam DW, James WP. Obesity. Lancet 2005; 366: 1197–209 5. Inge TH, King WC, Jenkins TM, et al. The effect of obesity in adolescence on adult health status. Pediatrics 2013; 132: 1098–104 6. The Australian Bureau of Statistics. Profiles of Health, Australia, 2011–13. Overweight and obesity. Available from http:// www.abs.gov.au/ausstats/[email protected]/Lookup/by%20Subject/ 4338.0~2011-13~Main%20Features~Overweight%20and%20 obesity~10007 (accessed 16 November 2014) 7. Chang SH, Stoll CRT, Song J, Varela JE, Eagon CJ, Colditz GA. The effectiveness and risks of bariatric surgery: an updated systematic review and meta-analysis, 2003-2012. JAMA Surg 2014; 149: 275–87 8. Picot J, Jones J, Colquitt JL, et al. The clinical effectiveness and cost effectiveness of bariatric (weight loss) surgery for obesity: a systematic review and economic evaluation. Health Technol Assess 2009; 13: 1–190, 215–357 9. Flum DR, Belle SH, King WC, et al. The Longitudinal Assessment of Bariatric Surgery (LABS) Consortium. Perioperative safety in the longitudinal assessment of bariatric surgery. N Engl J Med 2009; 361: 445–54 10. Morgan DJR, Ho KM, Armstrong J, Litton E. Long-term clinical outcomes and health care utilization after bariatric surgery. A population-based study. Ann Surg 2015; 262: 86–92 11. Helling TS, Willoughby TL, Maxfield DM, Ryan P. Determinants of the need for intensive care and prolonged mechanical ventilation in patients undergoing bariatric surgery. Obes Surg 2004; 14: 1036–41 12. Cendán JC, Abu-aouf D, Gabrielli A, et al. Utilization of intensive care resources in bariatric surgery. Obes Surg 2005; 15: 1247–51
13. Van den Broek RJ, Buise MP, van Dielen FM, Bindels AJ, van Zundert AA, Smulders JF. Characteristics and outcome of patients admitted to the ICU following bariatric surgery. Obes Surg 2009; 19: 560–4 14. Leykin Y, Pellis T, Del Mestro E, Fanti G, Marzano B. Perioperative management of 195 consecutive bariatric patients. Eur J Anaesthesiol 2008; 25: 168–70 15. The Australian Bureau of Statistics. Historical population statistics. Available from http://agencysearch.australia.gov.au/s/ search.html?clickedfluster=historical+population+statistics &form=simple&cluster0=population&query=%60Historical +Population+Statistics%60&collection=agencies&clive=0& profile=abs (accessed 16 November 2014) 16. Data Linkage Western Australia. Available from http://www. datalinkage-wa.org.au/ (accessed 16 November 2014) 17. Holman CDJ, Bass AJ, Rouse IL, Hobbs MS. Population-based linkage of health records in Western Australia: development of a health services research linked database. Aust N Z J Public Health 1999; 23: 453–9 18. Haggar FA, Pereira G, Preen DD, Holman CD, Einarsdottir K. Cancer survival and excess mortality estimates among adolescents and young adults in Western Australia, 1982–2004: a population-based study. PLoS ONE 2013; 8: e55630 19. Hung J, Teng TH, Finn J, et al. Trends from 1996 to 2007 in incidence and mortality outcomes of heart failure after acute myocardial infarction: a population-based study of 20,812 patients with first acute myocardial infarction in Western Australia. J Am Heart Assoc 2013; 2: e000172 20. Boyd JH, Ferrante AM, O’Keefe CM, et al. Data linkage infrastructure for cross-jurisdictional health-related research in Australia. BMC Health Serv Res 2012; 12: 480 21. Harrell FE Jr. Regression Modeling Strategies. New York: Springer, 2001 22. Ho KM, Knuiman M, Finn J, Webb SA. Estimating long-term survival of critically ill patients: the PREDICT model. PLoS One 2008; 3: e3226 23. Finks JF, Kole KL, Yenumula PR, et al. Predicting risk for serious complications with bariatric surgery: results from the Michigan Bariatric Surgery Collaborative. Ann Surg 2011; 254: 633–40 24. DeMaria EJ, Portenier D, Wolfe L. Obesity surgery mortality risk score: proposal for a clinically useful score to predict mortality risk in patients undergoing gastric bypass. Surg Obes Relat Dis 2007; 3: 134–40 25. Nightingale CE, Margarson MP, Shearer E, et al. Peri-operative management of the obese surgical patient 2015. Anaesthesia 2015; 70: 859–76 26. Obesity Surgery Society of Australia & New Zealand. http:// www.ossanz.com.au/ (accessed 17 January 2015) 27. Fried M, Yumuk V, Oppert JM, et al. International Federation for Surgery of Obesity and Metabolic Disorders-European Chapter (IFSO-EC); European Association for the Study of Obesity (EASO); European Association for the Study of Obesity Obesity Management Task Force (EASO OMTF). Interdisciplinary European guidelines on metabolic and bariatric surgery. Obes Surg 2014; 24: 42–55 28. Brolin RE. Gastric bypass. Surg Clin North Am 2001; 81: 1077–95 29. Smith FJ, Holman CD, Moorin RE, Fletcher DR. Incidence of bariatric surgery and postoperative outcomes: a populationbased analysis in Western Australia. Med J Aust 2008; 189: 198–202
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
Raine Foundation clinical research fellowship (to K.M.H.); St John of God Hospital Subiaco Research Grant (to the Western Australian Department of Health, covering the Data Linkage Branch extraction fees for the population-based data for all cases of bariatric surgery in Western Australia between 2007 and 2011).
| 881
882
| Morgan et al.
30. Lidor AO, Moran-Atkin E, Stem M, et al. Hospital-acquired conditions after bariatric surgery: we can predict, but can we prevent? Surg Endosc 2014; 28: 3285–92 31. Reoch J, Mottillo S, Shimony A, et al. Safety of laparoscopic vs open bariatric surgery: a systematic review and metaanalysis. Arch Surg 2011; 146: 1314–22
32. Avriel A, Warner E, Avinoach E, et al. Major respiratory adverse events after laparascopic gastric banding surgery for morbid obesity. Respir Med 2012; 106: 1192–8 33. El Shobary H, Backman S, Christou N, Schricker T. Use of critical care resources after laparoscopic gastric bypass: effect on respiratory complications. Surg Obes Relat Dis 2008; 4: 698–702 Handling editor: H. C. Hemmings
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
C L I N I CA L P R AC T I C E
D. J. R. Morgan1, *, K. M. Ho1,3,4, J. Armstrong2 and S. Baker5 1
Department of Intensive Care Medicine, 2Department of Surgery, St John of God Hospital Subiaco, 12 Salvado Road, Subiaco, WA 6008, Australia, 3School of Population Health, University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia, 4School of Veterinary and Life Sciences, Murdoch University, 90 South Street, Murdoch, WA 6150, Australia, and 5Department of Intensive Care Medicine, Sir Charles Gairdner Hospital, Hospital Ave, Nedlands, WA 6009, Australia *Corresponding author. E-mail: [email protected]
Abstract Background: With increasing rates of bariatric surgery and the consequential involvement of increasingly complex patients, uncertainty remains regarding the use of intensive care unit (ICU) services after bariatric surgery. Our objective was to define the incidence, indications, and outcomes of patients requiring ICU admission after bariatric surgery and assess whether unplanned ICU admission could be predicted using preoperative factors. Methods: All adult bariatric surgery patients between 2007 and 2011 in Western Australia were identified from the Department of Health Data Linkage Unit database and merged with a separate database encompassing all subsequent ICU admissions pertaining to bariatric surgery. The minimal and mean follow-up periods were 12 months and 3.4 yr, respectively. Results: Of the 12 062 patients who underwent bariatric surgery during the study period, 590 patients (4.9%; 650 ICU admissions) were admitted to an ICU after their bariatric surgery. Patients admitted to the ICU were older (48 vs 43 yr, P<0.001), more likely to be male (49.7 vs 20.2%, P<0.001), and more likely to require revisional bariatric surgery (14.4 vs 7.1%, P<0.001). One hundred and seventy-six patients required an emergent unplanned ICU admission, with 51 requiring multiple ICU admissions. Revisional or open surgery, diabetes mellitus, chronic respiratory disease, and obstructive apnoea were the strongest preoperative factors associated with unplanned ICU admission. Conclusions: Intensive care unit admission after bariatric surgery was uncommon (4.9% of all patients), with 30.9% of all referrals being unplanned. A nomogram and smartphone application based on five important preoperative factors may assist anaesthetists to conduct preoperative planning for high-risk bariatric surgical patients. Key words: bariatric surgery; critical care; obesity; postoperative complications
Obesity is an increasing health burden from both an individual and societal perspective in many developed countries.1–5 The prevalence of obesity in Australia has increased to 28% of the adult population.6 As a result, many countries have seen a
dramatic upsurge in the rates of bariatric surgery, with published data now establishing it as the most effective form of treatment in achieving sustained weight loss.7 8 Although there are emerging short- and long-term data to support both its efficacy and
Accepted: August 20, 2015 © The Author 2015. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved. For Permissions, please email: [email protected]
873
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
Incidence and risk factors for intensive care unit admission after bariatric surgery: a multicentre population-based cohort study
874
| Morgan et al.
Editor’s key points • The predictive factors for and utilization of intensive care after bariatric surgery are poorly defined. • In a retrospective analysis of a large database from Western Australia, all bariatric surgery patients and those requiring ICU admission were identified and analysed. • Of 12 062 bariatric surgery patients, 4.9% required ICU admission, of which 31% were unplanned. • A predictive nomogram and smartphone application based on five preoperative risk factors may help to identify highrisk patients.
Methods Ethics approval Prior ethics committee approval was granted by the Western Australian Department of Health (#2012/21) and all eight participating hospitals, including a waiver of patient consent owing to the large cohort size in addition to the observational, non-interventional, low-risk study design.
Patient population and conduct of study This was a retrospective, multicentre, population-based cohort study including all adult patients who underwent elective bariatric surgery in the state of Western Australia throughout a 5 yr period between 2007 and 2011. Patients admitted to any intensivist-led ICU after bariatric surgery formed a distinct subgroup. All outcomes were censored on December 31, 2012 unless the patients died beforehand, allowing for a minimum of 12 months follow-up after bariatric surgery. In 2011, Western Australia had a population of 2.39 million, comprising 10.4% of the total Australian population.15
Data handling Bariatric surgery hospitalizations Patients who had undergone bariatric surgery and their associated co-morbidity data were identified using the Western Australian Department of Health Data Linkage Unit (DLU) databases that systematically connect and update all available health
Bariatric surgery intensive care admissions Study data for patients admitted to an ICU were sourced initially by the Western Australian Department of Health DLU and then crossreferenced by screening all local ICU databases in Western Australia manually to ensure that every ICU admission pertaining to bariatric surgery or its complications was included. Only bariatric patients admitted to one of eight hospitals with a specialist intensivistrun ‘closed’ ICU where considered as requiring critical care support.
Clinical end points The incidence of ICU admission after bariatric surgery and differences in characteristics between those who were admitted to the ICU and those solely managed in the ward were the primary
12 264 patients screened for bariatric surgery between 2007 and 2011 in the whole state of Western Australia 202 patients excluded who had comorbid ICD-10 diagnostic codes for gastric cancer, peptic ulcer disease, or oesophageal varices
12 062 patients with an index bariatric procedure between 2007 and 2011 (mean follow-up 3.4 years)
Linkage with Western Australian death registry
11 472 patients (95.1%) admitted to regular hospital wards following bariatric surgery
35 180 intensive care unit (ICU)admissions for all Western Australian public & private ICUs between 2007 and 2011 merged with bariatric data
590 bariatric patients (4.9% of all bariatric surgery cases, 1.7% of all ICU admissions) admitted to one of eight ICUs in Western Australia following bariatric surgery electively or as a result of a complication after bariatric surgery
414 electively admitted (3.4%) to an ICU after bariatric surgery
176 bariatric patients (1.5%) who had an unplanned ICU admission
Fig 1 Flow chart showing the inclusion of patients and their disposition in relationship to the index bariatric operation and intensive care admission during the study period. ICD, International Diagnostic Code; ICU, intensive care unit.
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
its safety,9 10 the sustained increase in the incidence of bariatric surgery and improving surgical techniques have inevitably led to more patients with multiple co-morbidities. This inescapably means that there will be a growing number of patients who may require either elective or emergent critical care support after bariatric surgery. Despite the wealth of published research relating to bariatric surgery during the last 20 yr, there is a paucity of data describing the relationship between levels of postoperative care [ postanaesthesia, high-dependency, or intensive care unit (ICU)] and bariatric surgery. Currently, data are limited to a handful of older single-centre studies, with admission rates for a higher level of postoperative care after bariatric surgery varying widely between 4 and 21%.11–14 To date, there are no robust epidemiological data to assist clinicians in planning for critical care service requirements after bariatric surgery. Using the well-established population-based data linkage system in Western Australia, this study aimed to define the incidence, risk factors, and outcomes of patients who were admitted to an ICU after bariatric surgery. Specifically, we also hypothesized that unplanned ICU admissions after bariatric surgery could be predicted using preoperative factors.
data for every individual within the entire state of Western Australia using the 10th edition International Classification of Diseases (ICD-10) diagnostic and procedure codes.16 The accuracy of this database has been previously validated and formed the basis of a large number of population-based epidemiological studies in Western Australia.17–19 In addition, the techniques used by the DLU to obtain linkage of health data across different jurisdictions has also been applied to other states in Australia and was found to be capable of providing high-quality data for research purposes (see Fig. 1 and Supplementary Online Information 01: supplemental Methodology).20
Admission to ICU after bariatric surgery
outcomes. Secondary outcomes included identification of preoperative risk factors that were associated with unplanned ICU admissions and the surgical, anaesthetic, and medical complications that necessitated or involved ICU care.
Statistical analysis
was considered as a revisional surgery when it was performed for unresolved bariatric issues (e.g. failed to lose weight) that were not directly related to surgical complications from previous bariatric surgical procedures. During the multivariate modelling process, predictors were removed by a backward stepwise process according to the Design Library syntax recommended by Harrell Jr21 to improve the precision and utility of the final parsimony predictive model. The discrimination and calibration of the predictive model were assessed by area under the receiver operating characteristic curve and a calibration plot with bias corrected by a bootstrapping technique (n=200), respectively. The relative importance of each predictor in explaining the variability in the risk of unplanned ICU admission was assessed using the χ2 statistic minus the degrees of freedom. A nomogram based on the final parsimony predictive model was constructed to assist clinical decision-making.21 22 Missing data for height, and thus BMI, occurred in 4.1% of the ICU bariatric patients, and these patients were excluded when the differences in BMI between emergency and elective ICU admissions were analysed. All analyses were performed using SPSS for Windows (IBM Corp. Released 2013. IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY: IBM Corp.) or S-Plus (version 8.0; Insightful Corp., Seattle, WA, USA). A two-tailed P<0.05 was considered statistically significant. See Supplementary Online Information 01: supplemental Methodology for further details about an expanded participating hospital list; manuscript definitions; Western Australian Department of Health Data Linkage Unit; participants; and the use of International Classification of Diseases (ICD-10) coding.
Table 1 Baseline comparison of 12 062 bariatric patients stratified by need for intensive care unit admission after bariatric surgery in Western Australia between 2007 and 2011. ICU, intensive care unit; IQR, interquartile range; LOS, length of stay. aPatients re-admitted to hospital with unresolved bariatric issues Characteristic Patient numbers per yr [n (%)] 2007 2008 2009 2010 2011 Characteristics [, median, IQR] Age (mean; yr) Gender (male) [n (%)] Weight (mean; kg) BMI (mean; kg m−2) Funding for surgery [n (%)] Private Public Type of surgery [n (%)] Laparoscopic Open Original Revisional Bariatric re-hospitalizationsa Hospital LOS (mean; days) [, median, IQR] Deaths to December 2012 [n (%)]
Bariatric patients requiring ICU (n=590)
Bariatric patients not requiring ICU (n=11 472)
80 (4.0) 168 (6.8) 138 (5.1) 101 (4.2) 104 (4.2)
1901 (96.0) 2293 (93.2) 2574 (94.9) 2316 (95.8) 2388 (95.8)
48 [11.3, 49, 40–57] 294 (49.7) 136 [31.6, 133, 114–156] 46.6 [9.2, 46, 39.8–52.2]
43 [11.65, 43, 34–52] 2021 (20.2) Not available
515 (87.3) 75 (12.7)
10 966 (95.6) 506 (4.4)
555 (94.1) 35 (5.9) 505 (85.6) 85 (14.4) 216 (36.6) 8.6 [14.7, 4, 3–8] 14 (2.4)
11 122 (96.9) 350 (3.1) 10 665 (92.9) 817 (7.1) 686 (6.0) 2.6 [1.5, 2, 2–3] 8 (0.1)
P-value
<0.001
Not available
<0.001
<0.001 <0.001 <0.001 <0.001 <0.001
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
Categorical variables were described in absolute numbers and percentages, with dichotic comparisons performed using the χ2 test. Continuous variables were described as mean, standard deviation (), median, and interquartile range (IQR). Comparisons between normally distributed data were performed using Student’s unpaired t-test, whereas non-parametric data were compared using the Mann–Whitney U-test. Multivariate logistic regression was used to assess associations between patient characteristics, premorbid or perioperative factors, and different types of complications, all reported as odds ratios (ORs) with 95% confidence intervals (CIs). Clustering of unplanned ICU admissions within certain study centres was assessed in a separate sensitivity analysis using the generalized estimating equation. In assessing preoperative risk factors for unplanned ICU admissions, a multivariate analysis was undertaken comparing all unplanned ICU admissions with all bariatric surgical patients not requiring ICU across Western Australia. Our intention was to identify errors in postoperative pathway planning using only biologically plausible preoperative factors. These factors included age, gender, revisional vs primary surgery, open vs closed surgery, and chronic medical conditions before surgery. Bariatric surgery
| 875
876
| Morgan et al.
Results Differences between bariatric patients requiring and not requiring admission to ICU
Of the 650 ICU admissions after bariatric surgery, 449 admissions (69.1%) were electively referred to the ICU after surgery, whereas 201 (30.9%) were for emergent indications. The predominant clinical indication for ICU admission in most patients was surgically related (elective or emergent ; 89.0%), whereas anaesthetic (8.8%) and medical complications (2.2%) accounted for a small proportion of the main reason for ICU admission. Significant differences in patient characteristics existed between the two groups, with emergently admitted bariatric ICU patients more likely to be female and to have a lower BMI than elective admissions (Table 2). Emergent ICU admissions were more frequently associated with surgical complications, particularly when the problems arose in a delayed fashion after hospital discharge. Anaesthetic and medical complications were likewise more common among emergent ICU admissions (Table 3). Of the 298 recorded surgical complications, 21.5% occurred within the operating theatre and another 9.1% within 48 h of surgery. Significantly, 69.4% of all surgical complications occurred beyond the usual time frame of an elective ICU stay (<48 h) after bariatric surgery.
Table 2 Patient characteristics, co-morbidities, preoperative consultation and initial bariatric operation for 590 bariatric patients undergoing 650 admissions to an ICU between 2007 and 2011. CPAP, continuous positive airway pressure; COPD, chronic obstructive pulmonary disease; ICU, intensive care unit; IQR, interquartile range; LAGB, laparoscopic adjustable gastric band. aSpecialist review before day of surgery. b Includes internal physician, haematologist, nephrologist, immunologist, endocrinologist, clinical microbiologist, psychiatrist, or vascular surgeon. cBased on 650 ICU admissions in 590 patients. dRoux-en-Y, mini gastric bypass, pancreatobiliary bypass, gastroplasty, and gastric balloon procedures. eIndex hospitalization was the hospitalization related to the first bariatric surgery and excludes patients admitted to the ICU upon hospital re-admission for ongoing bariatric complications Characteristics Patient characteristics [, median, IQR] Age (mean; yr) Gender (male) [n (%)] Weight (mean; kg) BMI (mean; kg m−2) Premorbid co-morbidities [n (%)] Obstructive sleep apnoea CPAP Asthma/COPD Smoker Ischaemic heart disease Hypertension ≥3 Antihypertensive drugs Diabetes mellitus Preoperative consultationa [n (%)] Cardiologist review Pulmonologist review Anaesthetist review Other specialist reviewb ASA classificationc [, median, IQR] Primary bariatric operation [n (%)] LAGB Sleeve gastrectomy Other bariatric surgeryd Revisional surgery Open surgery Total ICU admissions (n) ICU during index hospitalizatione [n (%)]
P-value
Elective ICU patients (n=414)
Emergent ICU patients (n=176)
48.2 [11.4, 49, 40–57] 231 (55.9) 142 [30, 140, 122–160] 48.2 [9.1, 47.6, 41–54]
47.1 [11.1, 48, 39–55] 62 (35.2) 120 [30, 118, 98–139] 42.3 [8.1, 41.4, 36–49]
151 (36.6) 162 (39.2) 83 (20.1) 51 (12.3) 44 (10.7) 227 (55.0) 54 (13.0) 140 (33.9)
41 (23.3) 28 (15.9) 30 (17.0) 33 (18.8) 9 (5.1) 67 (38.1) 11 (6.2) 36 (20.5)
0.002 <0.001 0.43 0.05 0.04 <0.001 0.01 0.001
23 (5.1) 44 (9.8) 113 (25.2) 10 (2.2) 3.2 [0.7, 3, 3–4]
24 (11.9) 5 (2.5) 33 (16.4) 3 (1.5) 3.8 [2.2, 3, 3–4]
0.002 0.001 0.02 0.76 <0.001
231 (55.9) 174 (42.1) 8 (2.0) 43 (10.4) 17 (4.1) 449 393 (87.5)
95 (54.0) 71 (40.3) 10 (5.7) 42 (23.9) 18 (10.2) 201 92 (45.8)
0.28 <0.001 <0.001 <0.001
0.72 0.72 0.03 0.001 0.007 <0.001
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
In the 5 yr between 2007 and 2011, there were 12 062 index cases of bariatric surgery in Western Australia (mean follow-up 3.4 yr after index bariatric surgery). Of these, 590 (4.9%) patients required 650 ICU admissions (Fig. 1), giving a rate of 22.6 ICU admissions per 1000 patient-yr after bariatric surgery (95% CI 20.9–24.5). During the same period in Western Australia, there were 5.8 bariatric ICU admissions per 100 000 person-yr, 107 bariatric operations per 100 000 person-yr, and 309 overall ICU admissions per 100 000 person-yr. All-cause long-term mortality after bariatric surgery was significantly higher in patients who required ICU admission (6.2 vs 0.2 deaths per 1000 patient-yr, P<0.001). When compared with bariatric surgery patients not requiring critical care support, patients admitted to ICU were an average of 5 yr older, more likely to be male, and more likely to have diabetes mellitus. In addition, those patients requiring admission to ICU were more likely to have undergone revisional or open bariatric surgery, resulting in a longer hospital length of stay and increased rehospitalizations for unresolved bariatric issues (Table 1).
Differences between elective and emergent ICU admissions
Admission to ICU after bariatric surgery
Bariatric surgery patients admitted to the ICU in an emergent manner were significantly more likely to require a raft of ICU interventions (Table 4), including higher rates of intubation, vasopressor infusions, and blood transfusions. Likewise, total parental nutrition, prolonged durations of antibiotics, and increased radiological investigations were all more common in the emergent ICU admission group. As expected, both ICU and hospital length of stay were significantly longer for patients admitted to the ICU for emergent reasons compared with elective procedures.
Risk factors for complications and recurrent ICU admissions
an increased BMI was associated with fewer surgical complications (OR=0.94, 95% CI 0.92–0.96). In the subgroup of 51 patients requiring multiple ICU admissions (Supplementary Online Information 01: supplemental Table S1), gastrointestinal leak, abscess, or both was the only predictor associated with an increased risk of recurrent ICU admissions (OR=7.46, 95% CI 5.7–17.0).
Predictors for unplanned ICU admissions Of the 12 062 bariatric surgery patients included, emergent ICU admissions occurred in 176 (1.5%) patients. Multivariate analysis showed that revisional or open surgery, diabetes mellitus, chronic respiratory disease, chronic cardiovascular disease, chronic renal disease, and obstructive sleep apnoea (OSA) were all associated with emergent ICU admission (Supplementary Online Information 01: supplemental Table S2). Of these seven significant variables, revisional surgery was most important, explaining ∼36.5% of the variability in the risk of requiring unplanned ICU
Table 3 Bariatric complications associated with 650 ICU admissions between 2007 and 2011 stratified by urgency of ICU admission.a AKIN, acute kidney injury network; DVT, deep vein thrombosis; GIT, gastrointestinal tract; ICU, intensive care unit; PE, pulmonary embolism. a590 bariatric patients, of whom 51 patients had multiple ICU admissions. bFifteen patients admitted emergently to the ICU had no defined complications but had unpredictably prolonged or technically difficult operations, or unappreciated co-morbidities before surgery. c Complications occurring after theatre and before initial hospital discharge. dComplications requiring subsequent re-hospitalization. e Complications including gastric band malfunction, port infection, or failure to lose weight. fUnplanned additional surgical interventions. g Complications are as follows: obstructed airway, difficult intubation, hypoventilation, hypoxia, anaphylaxis, bronchospasm, gas embolism, cardiac arrest, respiratory arrest, poorly controlled hypertension, and hypotension. hComplications are as follows: non-cardiac chest pain, neurological, metabolic, hepatic, hypertension, congestive cardiac failure, disseminated intravascular coagulopathy, diabetic ketoacidosis, pancreatitis, severe eczema, diarrhoea, narcotization, undiagnosed sleep apnoea, and undiagnosed cancer Characteristic Surgical complications (timing) [n (%)] Intraoperative complication Postoperative complicationsc Delayed complicationsd Surgical complications (type) [n (%)] Admission with any complication GIT obstruction GIT perforation, leak, abscess Haemorrhage Other surgical complicationse Positive microbiology cultures Surgical interventionsf [n (%)] Drain (unplanned) Endoscopy Laparoscopy Laparotomy Thoracotomy Anaesthetic complications [n (%)] Any anaesthetic complicationg Medical complications [n (%)] Admissions with any complication Cardiac arrhythmia Left pleural effusion Lower respiratory tract infection Aspiration pneumonitis Empyema Asthma Septic shock Acute coronary syndrome AKIN stage 3 DVT, PE, or both Other medical complicationsh
Elective ICU admission (n=449)
Emergent ICU admission (n=201)b
P-value
42 (9.4) 19 (4.2) 55 (12.2)
22 (10.9) 63 (31.3) 97 (48.5)
0.57 <0.001 <0.001
95 (21.2) 19 (4.2) 23 (5.1) 21 (4.7) 29 (6.5) 8 (1.8)
149 (74.1) 21 (10.4) 94 (46.8) 43 (21.4) 30 (14.9) 66 (32.8)
<0.001 0.004 <0.001 <0.001 0.001 <0.001
8 (1.8) 22 (4.9) 61 (13.6) 34 (7.6) 2 (0.4)
35 (17.4) 34 (16.9) 89 (44.3) 85 (42.3) 7 (3.5)
<0.001 <0.001 <0.001 <0.001 0.005
8 (1.8)
55 (27.4)
<0.001
64 (14.3) 13 (2.9) 2 (0.4) 4 (0.9) 0 (0) 0 (0) 6 (1.3) 2 (0.4) 3 (0.7) 0 (0) 2 (0.4) 34 (7.6)
128 (63.7) 19 (9.5) 27 (13.4) 3 (1.5) 2 (1.0) 8 (4.0) 8 (4.0) 30 (14.9) 5 (2.5) 16 (8.0) 7 (3.5) 47 (23.4)
<0.001 <0.001 <0.001 0.68 0.10 <0.001 0.04 <0.001 0.11 <0.001 0.005 <0.001
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
In the multivariate analysis of ICU admissions (Table 5), all-cause surgical complications were associated with female gender (OR=1.53, 95% CI 1.06–2.21), revisional surgery (OR=4.76, 95% CI 2.79–8.12) and open surgery (OR=2.78, 95% CI 1.06–7.27), whereas
| 877
878
| Morgan et al.
Table 4 Intensive care unit referral origins, admission APACHE II scores, interventions, and outcomes for 650 ICU admissions after bariatric surgery between 2007 and 2011. APACHE, acute physiology and chronic health evaluation; CRRT, continuous renal replacement therapy; CT, computerized tomography; ED, emergency department; ICU, intensive care unit; IQR, interquartile range; LOS, length of stay; NIV, noninvasive ventilation; TPN, total parental nutrition. aTotal intubation time divided by total number of intubated patients. bBased on 590 ICU patients. cPatients re-admitted to hospital for unresolved bariatric issues Characteristics
Emergent ICU admission (n=201)
P-value
4.6 [2.9, 4, 2–6] 0 450 (100) 2 (0.4) 14 (3.1) 4.0 [8.1, 2, 1–3] 12 (2.7) 9 (2.0) 7 (1.6) 1 (0.2) 12 (2.7) 5 (1.1) 2 (0.4) 17 (3.8) 5 (1.1) 26 (5.8) 31 (6.9)
6.2 [4.4, 5, 3–9] 54 (26.9) 147 (73.1) 42 (21.0) 97 (48.3) 6.0 [12.4, 2, 1–5] 16 (8.0) 46 (22.9) 7 (3.5) 7 (3.5) 71 (35.3) 47 (23.4) 17 (8.5) 107 (53.2) 54 (26.9) 57 (28.4) 126 (62.7)
<0.001
2.2 [2.0, 2, 2–2] 4.8 [7.1, 4, 2–5] 0 0 106 (23.6)
5.5 [9.7, 3, 2–4] 17.2 [22, 10, 4–21] 3 (1.7) 3 (1.7) 109 (54.2)
admission (Supplementary Online Information 02: supplemental Fig. S1). In the sensitivity analysis using a generalized estimating equation, significant clustering of unplanned ICU admissions by individual hospital sites was not observed. The ability of the final parsimony predictive model to discriminate between patients requiring and not requiring unplanned ICU admission was reasonable (area under the receiver operating characteristic curve=0.820, 95% CI 0.78–0.86). The calibration of the model was, however, not perfect when the predicted risks of requiring unplanned ICU were greater than 30–40% (Supplementary Online Information 03: supplemental Fig. S2). A nomogram based on the five most important preoperative factors to estimate the probability of unplanned ICU admission after bariatric surgery is presented in Fig. 2 and Supplementary Online Information 04: supplemental Fig. S3. The score assigned to revision surgery is 100 points, chronic respiratory disease is 75 points, OSA is 60 points, open surgery is 55 points, and diabetes mellitus is 35 points. A total of 110 points (e.g. diabetes mellitus +respiratory disease), 135 points (e.g. OSA+respiratory disease), and >160 points (e.g. revision surgery+either respiratory disease or OSA) predicts 10, 20, and >30% risk of unplanned ICU admission, respectively. An Android application (http://www. appsgeyser.com/1947332) for smart phones and a Windows PC application (http://tinyurl.com/ps7gptv) have been created to aid with preoperative risk prediction.
<0.001 <0.001 <0.001 <0.001 0.50 0.005 <0.001 0.14 0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.03 0.03 <0.001
Discussion This population-based multicentre study showed that a small (4.9%) but important proportion of patients required ICU support after bariatric surgery. Such patients tended to be older and were more likely to be male, have diabetes mellitus, and have undergone either revisional or open bariatric procedures. Revisional or open surgery, diabetes mellitus, chronic respiratory disease, and OSA were the strongest predictors for emergent unplanned ICU admission. Overall long-term mortality after bariatric surgery was very low. This is the first large study to describe the relationship between bariatric surgery and intensive care use and offers a significantly longer follow-up period than most other studies documenting complications. While other bariatric studies have defined factors contributing towards surgical complications,23 our study uniquely describes the increased complexity, resource utilization, cost, and breadth of complications associated with ICU care above that of complications managed at a ward level. Data regarding clinical care pathways that encompass intensive care after bariatric surgery are sparse.12 13 The previously published Obesity Surgery Mortality Risk Score (OS-MRS) refers only to 90 day mortality, a very rare event after bariatric surgery, with no reference to the more common occurrence of ICU utilization.24 Furthermore, the OS-MRS was conceived before revisional bariatric surgery became commonplace and thus does not
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
Intensive care admission Admission APACHE II score [, median, IQR] Origin (ward or ED) [n (%)] Origin (theatre) [n (%)] Interhospital transfer [n (%)] Intubation [n (%)] Intubation timea (days) [, median, IQR] Unplanned NIV [n (%)] Vasopressor infusion [n (%)] Vasodilator infusion [n (%)] CRRT [n (%)] Blood transfusion [n (%)] TPN [n (%)] Tracheostomy [n (%)] CT abdomen [n (%)] CT chest [n (%)] Barium swallow [n (%)] Antibiotics >72 h [n (%)] Outcomes ICU LOS (days) [, median, IQR] Hospital LOS (days) [, median, IQR] ICU mortalityb [n (%)] Hospital mortalityb [n (%)] Bariatric re-hospitalizationsb,c [n (%)]
Elective ICU admission (n=449)
Admission to ICU after bariatric surgery
| 879
Table 5 Multivariate analysis showing factors associated with surgical, anaesthetic, and medical complications occurring in 650 ICU adminssions after bariatric surgery. CI, confidence interval; OR, odds ratio. aP-value for adjusted OR. bComplications as listed in Table 3. c Specialist review before day of elective bariatric surgery Characteristics
Unadjusted OR [95% CI]
0
10
20
30
40
50
60
P-valuea
0.98 [0.96–1.00] 1.53 [1.06–2.21] 0.94 [0.92–0.96] 0.84 [0.55–1.28] 4.76 [2.79–8.12] 2.78 [1.06–7.27]
0.04 0.02 <0.001 0.42 <0.001 0.04
1.01 [0.98–1.03] 1.10 [0.65–1.85] 0.99 [0.96–1.02] 0.65 [0.38–1.11] 1.82 [0.97–3.40] 0.95 [0.35–2.56] 0.95 [0.51–1.77] 0.54 [0.29–1.02]
0.65 0.72 0.46 0.11 0.06 0.91 0.87 0.06
1.01 [0.99–1.02] 2.08 [1.45–3.03] 0.95 [0.93–0.97] 0.73 [0.48–1.12] 1.80 [0.99–2.68] 0.86 [0.41–1.83] 0.65 [0.41–1.05] 1.48 [0.87–2.50] 0.67 [0.29–1.55]
0.56 <0.001 <0.001 0.15 0.05 0.70 0.08 0.14 0.34
70
80
90
100
Points Yes RESP No Yes DM No Yes OSA No Yes Revision No Yes Open No Total points 0
40
80
120
160
200
240
280
Linear predictor –5
–4
–3
–2
–1
0
0.5
1
1.5
2
2.5
Probability 0.1
0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.9
Fig 2 Nomogram using the five most important preoperative predictors to estimate the probability of emergent, unplanned intensive care unit admission after bariatric surgery. DM, diabetes mellitus; Open, open surgery; OSA, obstructive sleep apnoea; RESP, chronic respiratory conditions other than OS; Revision, revisional surgery.
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
Potential risk factors associated with all surgical complicationsb Age 0.99 [0.98–1.01] Gender (female) 1.97 [1.42–2.74] BMI 0.94 [0.92–0.96] Diabetes mellitus 0.77 [0.54–1.10] Revisional surgery 3.77 [2.43–5.84] Open surgery 3.96 [1.94–8.08] Potential risk factors associated with all anaesthetic complicationsb Age 1.00 [0.98–1.02] Gender (male) 0.95 [0.59–1.54] BMI 0.98 [0.96–1.01] Obstructive sleep apnoea 0.62 [0.38–1.01] Smoking 1.67 [0.91–3.05] Pulmonologist consultationc 0.99 [0.41–2.42] Anaesthetist consultationc 0.85 [0.48–1.53] ASA grade IV classification 0.50 [0.09–2.73] Potential risk factors associated with all medical complicationsb Age 1.01 [0.99–1.03] Gender (female) 1.44 [0.96–2.14] BMI 0.97 [0.95–0.99] Diabetes mellitus 0.79 [0.50–1.23] Smoking 1.51 [0.90–2.54] Pulmonologist consultationc 1.35 [0.67–2.74] Anaesthetist consultationc 0.83 [0.51–1.35] Revisional surgery 1.41 [0.78–2.54] Open surgery 0.69 [0.26–1.82]
Adjusted OR [95% CI]
880
| Morgan et al.
was still lower than the previously reported range of 14.4–28.2% in two smaller and older ICU studies.12 33 Hellings and colleagues reported pulmonary co-morbidity was associated with increased risk of prolonged mechanical ventilation >24 h in 60 bariatric patients admitted to ICU.11 Importantly, amongst our ICU cohort were 24 patients initially diagnosed as lower respiratory tract infections, of whom 14 in reality had gastric leaks with atelectasis, delaying definitive surgical management by as much as 48 h. Hence, it may be prudent for ICU physicians to consider intra-abdominal sepsis from gastric leaks when patients present with clinical signs of sepsis after bariatric surgery, even when there are radiological changes consistent with lower respiratory tract infection. This study is limited by its retrospective, observational nature and possible selection bias and confounding. The use of an administrative database is constrained by the coding conventions pertinent to the collection of large administrative data sets not specifically designed for individual studies. As such, we could not confirm whether one particular subtype of bariatric surgical procedure was more likely to require ICU admission than the others, nor collect heights and weights for the overall bariatric cohort. Additionally, because there is limited public health-care funding for bariatric surgery in Australia, most of our cohort were admitted to private health-care facilities where ICU accessibility was less constrained than in the public health-care system. This may limit the generalizability of our results outside Australia or on patient populations predominately using government-funded health organizations. Our analysis was also constrained by the difficulty of reliably anticipating intraoperative technical complications. Nevertheless, the strength of revisional surgery as a strongest predictive factor suggests that this is a good surrogate for intraoperative technical and surgical difficulties and complications, probably owing to adhesions, altered anatomy, and previous tissue injury.
Conclusions In a multidisciplinary approach to bariatric surgery, intensive care plays a succinct but important role, with one in 20 Western Australians found to require ICU admission after bariatric surgery. This is a lower rate than most published literature to date. Revisional or open surgery, diabetes mellitus, chronic respiratory disease, and OSA were the strongest predictors associated with subsequent unplanned ICU admission after bariatric surgery. A nomogram smart phone application based on these five most important preoperative factors may assist clinicians in identifying bariatric patients who should be operated on in a facility that has an ICU or electively admitted to an ICU after a procedure.
Authors’ contributions Study concept: D.J.R.M., K.M.H. Study design: D.J.R.M., K.M.H., J.A. Data collection: D.J.R.M., K.M.H., S.B. Data analysis: D.J.R.M. Statistical analysis: K.M.H. Drafting the manuscript: D.J.R.M., K.M.H. Manuscript review: J.A., S.B. Final approval: D.J.R.M., K.M.H., J.A., S.B.
Supplementary material Supplementary material is available at British Journal of Anaesthesia online.
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
acknowledge this factor as an important element in postoperative outcomes. In a recent guide to perioperative management of the obese surgical patient, information pertaining to postoperative planning is unsupported by detailed references for ICU clinical pathways.25 Similar guidelines from other major international bariatric bodies are constrained to discussing surgical volume and generic descriptions of supportive staff requirements, with no discussion of critical care support.26 27 Our data on the differences between elective and emergent ICU admission after bariatric surgery improve our understanding of the requirements of critical care services related to bariatric surgery. We chose to assess preoperative factors in predicting unplanned ICU admissions because such a predictive model would be more clinically useful than models that rely on postoperative factors or complications, when unplanned ICU admission would be imminent and inevitable. We anticipated that a predictive model based on preoperative factors only will be imperfect because unexpected events that may require ICU admission do occur during the intraoperative and postoperative period. Nevertheless, our model appeared to have a reasonable calibration up to a predicted risk of 30–40%. If our model is further validated by other centres, the nomogram based on these five most important preoperative factors may be useful in assisting clinicians to decide which bariatric patients should be operated on in a facility that has an ICU or electively admitted to an ICU after a bariatric procedure. Importantly, both revisional and open surgery were risk factors for increased risk of unplanned ICU admission and surgical complications, suggesting that these surgical techniques are likely to have an impact on the risk of requiring critical care services. Comparisons within the literature describing overall ICU utilization after bariatric surgery are constrained by the limited number of small, single-centre publications that often report older surgical techniques, including a much higher proportion of open procedures. Our data highlight some important differences from older studies. The incidence of ICU admission is much lower in our cohort than in many other reports (8–24%),11–13 28 despite a dramatic increase in bariatric procedures in Western Australia throughout the past decade. Additionally, bariatric surgery has remained largely the domain of private hospitals in Western Australia, where ICU beds are more accessible, ranging from 62 to 84% of bariatric procedures between 1988 and 2004 and increasing to 95% of our cohort.29 Our lower incidence of ICU utilization possibly reflects the improved surgical and anaesthetic techniques combined with an increased trend towards laparoscopic surgery, resulting in only 3.2% of our state-wide cohort requiring open procedures. This explains why the rate of complications was lower in our patients than in other reports even with a relatively long follow-up period for complications in this study.30 31 Despite the differences in surgical techniques, our study confirms that older male patients with multiple co-morbidities remain more likely to be admitted to an ICU after bariatric surgery.11–13 The apparent protective effect of a higher BMI in both surgical and medical complications appears counterintuitive. This possibly reflects a greater preponderance to admit patients with a higher BMI electively to the ICU and thus avert complications developing. It may also relate to the extended follow-up interval in our study, reflecting more the effects of long-term complications and repeated bariatric-related re-hospitalizations from surgical complications, resulting in both a decreased caloric intake and increased catabolism of ongoing illness. Respiratory complications are one of the main concerns in obese perioperative patients, with major adverse respiratory events reported to occur in ∼2% of large overall bariatric cohorts.32 In our cohort, the respiratory complication rate (12.5%)
Admission to ICU after bariatric surgery
Acknowledgements The authors thank the Obesity Surgery Society of Australia and New Zealand for their support in obtaining ethics approval from the Western Australian Department of Health (http://www. ossanz.com.au/).
Declaration of interest None declared.
Funding
References 1. Belle SH, Chapman W, Courcoulas AP, et al. LABS Writing Group for the LABS Consortium. Relationship of body mass index with demographic and clinical characteristics in the Longitudinal Assessment of Bariatric Surgery (LABS). Surg Obes Relat Dis 2008; 4: 474–80 2. Dunstan DW, Zimmet PZ, Welborn TA, et al. The rising prevalence of diabetes and impaired glucose tolerance: the Australian Diabetes, Obesity and Lifestyle Study. Diabetes Care 2002; 25: 829–34 3. Colagiuri S, Lee CMY, Colagiuri R, et al. The cost of overweight and obesity in Australia. Med J Aust 2010; 192: 260–4 4. Haslam DW, James WP. Obesity. Lancet 2005; 366: 1197–209 5. Inge TH, King WC, Jenkins TM, et al. The effect of obesity in adolescence on adult health status. Pediatrics 2013; 132: 1098–104 6. The Australian Bureau of Statistics. Profiles of Health, Australia, 2011–13. Overweight and obesity. Available from http:// www.abs.gov.au/ausstats/[email protected]/Lookup/by%20Subject/ 4338.0~2011-13~Main%20Features~Overweight%20and%20 obesity~10007 (accessed 16 November 2014) 7. Chang SH, Stoll CRT, Song J, Varela JE, Eagon CJ, Colditz GA. The effectiveness and risks of bariatric surgery: an updated systematic review and meta-analysis, 2003-2012. JAMA Surg 2014; 149: 275–87 8. Picot J, Jones J, Colquitt JL, et al. The clinical effectiveness and cost effectiveness of bariatric (weight loss) surgery for obesity: a systematic review and economic evaluation. Health Technol Assess 2009; 13: 1–190, 215–357 9. Flum DR, Belle SH, King WC, et al. The Longitudinal Assessment of Bariatric Surgery (LABS) Consortium. Perioperative safety in the longitudinal assessment of bariatric surgery. N Engl J Med 2009; 361: 445–54 10. Morgan DJR, Ho KM, Armstrong J, Litton E. Long-term clinical outcomes and health care utilization after bariatric surgery. A population-based study. Ann Surg 2015; 262: 86–92 11. Helling TS, Willoughby TL, Maxfield DM, Ryan P. Determinants of the need for intensive care and prolonged mechanical ventilation in patients undergoing bariatric surgery. Obes Surg 2004; 14: 1036–41 12. Cendán JC, Abu-aouf D, Gabrielli A, et al. Utilization of intensive care resources in bariatric surgery. Obes Surg 2005; 15: 1247–51
13. Van den Broek RJ, Buise MP, van Dielen FM, Bindels AJ, van Zundert AA, Smulders JF. Characteristics and outcome of patients admitted to the ICU following bariatric surgery. Obes Surg 2009; 19: 560–4 14. Leykin Y, Pellis T, Del Mestro E, Fanti G, Marzano B. Perioperative management of 195 consecutive bariatric patients. Eur J Anaesthesiol 2008; 25: 168–70 15. The Australian Bureau of Statistics. Historical population statistics. Available from http://agencysearch.australia.gov.au/s/ search.html?clickedfluster=historical+population+statistics &form=simple&cluster0=population&query=%60Historical +Population+Statistics%60&collection=agencies&clive=0& profile=abs (accessed 16 November 2014) 16. Data Linkage Western Australia. Available from http://www. datalinkage-wa.org.au/ (accessed 16 November 2014) 17. Holman CDJ, Bass AJ, Rouse IL, Hobbs MS. Population-based linkage of health records in Western Australia: development of a health services research linked database. Aust N Z J Public Health 1999; 23: 453–9 18. Haggar FA, Pereira G, Preen DD, Holman CD, Einarsdottir K. Cancer survival and excess mortality estimates among adolescents and young adults in Western Australia, 1982–2004: a population-based study. PLoS ONE 2013; 8: e55630 19. Hung J, Teng TH, Finn J, et al. Trends from 1996 to 2007 in incidence and mortality outcomes of heart failure after acute myocardial infarction: a population-based study of 20,812 patients with first acute myocardial infarction in Western Australia. J Am Heart Assoc 2013; 2: e000172 20. Boyd JH, Ferrante AM, O’Keefe CM, et al. Data linkage infrastructure for cross-jurisdictional health-related research in Australia. BMC Health Serv Res 2012; 12: 480 21. Harrell FE Jr. Regression Modeling Strategies. New York: Springer, 2001 22. Ho KM, Knuiman M, Finn J, Webb SA. Estimating long-term survival of critically ill patients: the PREDICT model. PLoS One 2008; 3: e3226 23. Finks JF, Kole KL, Yenumula PR, et al. Predicting risk for serious complications with bariatric surgery: results from the Michigan Bariatric Surgery Collaborative. Ann Surg 2011; 254: 633–40 24. DeMaria EJ, Portenier D, Wolfe L. Obesity surgery mortality risk score: proposal for a clinically useful score to predict mortality risk in patients undergoing gastric bypass. Surg Obes Relat Dis 2007; 3: 134–40 25. Nightingale CE, Margarson MP, Shearer E, et al. Peri-operative management of the obese surgical patient 2015. Anaesthesia 2015; 70: 859–76 26. Obesity Surgery Society of Australia & New Zealand. http:// www.ossanz.com.au/ (accessed 17 January 2015) 27. Fried M, Yumuk V, Oppert JM, et al. International Federation for Surgery of Obesity and Metabolic Disorders-European Chapter (IFSO-EC); European Association for the Study of Obesity (EASO); European Association for the Study of Obesity Obesity Management Task Force (EASO OMTF). Interdisciplinary European guidelines on metabolic and bariatric surgery. Obes Surg 2014; 24: 42–55 28. Brolin RE. Gastric bypass. Surg Clin North Am 2001; 81: 1077–95 29. Smith FJ, Holman CD, Moorin RE, Fletcher DR. Incidence of bariatric surgery and postoperative outcomes: a populationbased analysis in Western Australia. Med J Aust 2008; 189: 198–202
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015
Raine Foundation clinical research fellowship (to K.M.H.); St John of God Hospital Subiaco Research Grant (to the Western Australian Department of Health, covering the Data Linkage Branch extraction fees for the population-based data for all cases of bariatric surgery in Western Australia between 2007 and 2011).
| 881
882
| Morgan et al.
30. Lidor AO, Moran-Atkin E, Stem M, et al. Hospital-acquired conditions after bariatric surgery: we can predict, but can we prevent? Surg Endosc 2014; 28: 3285–92 31. Reoch J, Mottillo S, Shimony A, et al. Safety of laparoscopic vs open bariatric surgery: a systematic review and metaanalysis. Arch Surg 2011; 146: 1314–22
32. Avriel A, Warner E, Avinoach E, et al. Major respiratory adverse events after laparascopic gastric banding surgery for morbid obesity. Respir Med 2012; 106: 1192–8 33. El Shobary H, Backman S, Christou N, Schricker T. Use of critical care resources after laparoscopic gastric bypass: effect on respiratory complications. Surg Obes Relat Dis 2008; 4: 698–702 Handling editor: H. C. Hemmings
Downloaded from http://bja.oxfordjournals.org/ at Stockholms Universitet on November 22, 2015