Guimaraes, Emily S. MD*†; Campbell, Emily J. MPH‡; Richter, James M. MD, MA†‡
Anesthesia services use for endoscopic procedures have been projected to continue its recent very rapid increase in the coming years if current factors persist.1 Anesthesia care may be of most value (with respect to both outcome and cost) where patients are the most ill, or where the deepest sedation levels are required. The American Society of Anesthesiologists recommends, in its 2002 guideline, that “those intending to administer deep sedation should be able to rescue patients who enter a state of general anesthesia” and that, for planned deep sedation, there be a person in the room skilled in endotracheal intubation and advanced life support.2 This guideline also, however, states that the literature is silent regarding the effectiveness of training on patient outcomes. The Standards of Practice Committee of the American Society of Gastrointestinal Endoscopy states, in its 2008 guideline, that “typically, diagnostic and uncomplicated therapeutic upper endoscopy and colonoscopy are successfully performed with moderate sedation. Deeper levels of sedation may be considered for longer and more complex procedures, including, but not limited to, endoscopic retrograde cholangiopancreatography (ERCP) and endoscopic ultrasound (EUS).3
On December 11, 2009, the Centers for Medicare and Medicaid Services (CMS) revised its interpretive guidelines requiring the administration of deep sedation as an anesthesia service overseen by a phyisician director of anesthesia.4 This CMS policy statement on personnel and level of sedation has yet to be supported by comparative data on safety and outcomes in gastroenterology.5 In January 2010, our hospital made the decision that advanced endoscopy procedures would no longer receive gastroenterologist-supervised nursing sedation. As of April 6, 2010, all patients were cared for by an anesthesiologist, or a certified registered nurse anesthetist (CRNA) under anesthesiologist supervision, to ensure appropriate sedation and full compliance.
Before this policy change, patients received sedation by a skilled and experienced certified registered nurse (RN) supervised by the gastroenterologist, usually with midazolam/meperidine or midazolam/fentanyl, occasionally supplemented with diphenhydramine. All patients receiving nurse-administered procedural sedation also received supplemental nasal oxygen and were monitored with continuous oximetry. The endoscopy nurses administering procedural sedation had at least 5 years experience, usually in an intensive care unit (ICU) or emergency department. They also completed a tutorial on procedural sedation designed by our Department of Anesthesia and Critical Care, as well as a period of 8 weeks of supervised practice. Patients could be referred for anesthesia care by the endoscopist or sedation nurse, usually for severe comorbid illness, previous sedation failure, or substance abuse. Before the policy change, 17% of these procedures used anesthesia care.
The division of gastroenterology and department of anesthesia jointly organized to provide anesthesia care teams for 3 endoscopy rooms, and as of April 6, 2010, all patients scheduled for advanced endoscopy received anesthesia care. The usual model was 3 CRNAs supervised by 1 anesthesiologist. The anesthetic technique and airway management were at the discretion of the anesthesiologist. In a prospectively agreed on exclusion to the policy, patients having isolated distal rectal EUS could request RN-administered minimal or moderate sedation.
The purpose of this study was to establish whether patient safety during advanced endoscopy, as measured by quality assurance (QA) database complication reports, was affected by the policy change. We compared a historical cohort of complications from nursing sedation and anesthesia care for advanced endoscopy for a total time period of 5 years encompassing this change in our practice, using information from 2 QA databases for serious adverse outcomes. Propofol titrated to moderate to deep sedation by RNs under gastroenterologist supervision has been shown in a 2009 meta-analysis to have a low rate of serious adverse outcomes, with 3 deaths in 460,391 reported patients. All 3 deaths were in upper endoscopy in patients ASA physical status III or greater.6 Patient age, higher ASA physical status, and advanced endoscopic procedures are all shown to increase the risk of adverse outcomes in gastroenterologic sedation.7 We hypothesized that our patient population of relatively older and ill patients requiring prolonged and complex procedures and deeper planes of sedation should serve as a high-risk cohort in which to demonstrate any safety advantage for anesthesia care compared with nursing sedation.
We conducted a single-center cohort study using historical controls. All adult patients undergoing ERCP and EUS in our hospital from October 6, 2007 through April 6, 2010 were compared with those performed after a practice change on April 6, 2010, and through October 5, 2012. This study was approved by our IRB, with a waiver for obtaining informed consent.
Patient demographics, procedure data, and sedation type were gathered primarily through the endoscopic report writing software used by the division of gastroenterology (Provation MD), which contains 15 years of endoscopic report data. It is routinely used for complication and performance measurement and as well as operational and administrative reporting. Endoscopic care was provided by the same 5 interventional gastroenterologists, and 5 advanced fellows. These interventional gastroenterologists had between 5 and 38 years of posttraining local experience at the beginning of the study. The growth in advanced endoscopy was proportionate to the growth in the use of endoscopy procedures as a whole. There was no systematic change in patient selection or referral patterns except for small additional increases in a rectal EUS program as a result of a new surgical program in the institution.
Study definitions of interventions and outcomes for this retrospective analysis can be found in Figure 1. Our primary outcome was incidence of sedation, endoscopic, and total complications, reported either to the Gastrointestinal Provation database or to the hospital quality and safety database; we followed all complications in the full medical record to assess for major morbidity and mortality. We compared instances and rates of reported complications between the 2 time periods: prepolicy change (October 6, 2007–April 5, 2010) and postpolicy change (April 6, 2010–October 5, 2012) and performed secondary analyses by actual sedation provider (anesthesia care versus nursing sedation)
Our gastrointestinal division policy requires endoscopists to record procedural complications within the Provation procedure report before each report is finalized. A complication is defined precisely and prospectively by the division, and all gastroenterologists have at least 5 years experience with use of this definition. All procedures are also reviewed by physicians within 2 weeks of the procedure to report any postprocedure (“late”) complications. Complications may also be reported by all hospital staff within the institution’s quality and safety reporting system. During the 5-year study period, there were no policy changes regarding the definition or reporting of complications, nor were there any changes made to either of the databases.
To ensure complete reporting, we queried both of these independent databases collecting the following datapoints: type of procedure, the attending endoscopist, a brief description of the complication, type of sedation used (RN sedation or anesthesia care), as well as identifying information such as medical record number, patient gender, age, and ASA physical status. When a datapoint for sedation type was unavailable in the database, (<1% of cases) the patient’s medical record was reviewed. We used ASA physical status assessed by the uniform group of gastroenterologists for all patients: anesthesiologist assessments were not available for all patients and were thus not used for the purposes of the study. Some studies suggest that there are systematic differences in ASA physical status assessment based on scorers’ training background.8,9 All complications were then reviewed in the medical record for verification and complete documentation of morbidity and mortality. Each complication was discussed by both an endoscopist (JMR) and an anesthesiologist (ESG) to assign a complication category. All complications are described below (Appendix 1.)
Our primary study outcome was the rate of complications between the 2 time periods, which are represented as a proportion of the entire examination volume in each cohort. All the proportions between the 2 time periods were tested for statistical significance using the Fisher exact test. Ninety-five percent (95%) confidence intervals (CIs) were also calculated to test the difference between the 2 independent proportions, using the simple asymptotic method without continuity correction.10 Significance was set at P < 0.05. Assuming a 2-sided α of 0.05 with 80% power, we calculated a study sample size of at least 4645 examinations per group to demonstrate a statistically and clinically significant difference of 0.30% in the rate of complications between the 2 groups.11 Based on this sample size and our procedural volume, we defined our study period as the 30 months both before and after the sedation policy change (total study period of 60 months or 5 years). All data were collected and analyzed using Microsoft Excel software (Microsoft Office 2010, Microsoft Corp., Redmond, WA).
Patient demographics for our entire study cohort are summarized in Table 1. Gender and mean age were the same in both groups, though the “pre” cohort had a larger percentage of patients aged 71 years and older (30.5%) than the “post” cohort (28.4%, P = 0.025). There was a significant decrease in the percent of patients rated as ASA physical status I between the pre- (10.9%) and postgroups (2.0%, P < 0.0001). Likewise, there was also a significant decrease in the rate of ASA physical status II patients (77.7% vs 74.5%, P = 0.017). However, we also found an increase in the rate of ASA physical status III patients between the pre- (10.2%) and postgroups (22.6%), which was also statistically significant (P < 0.0001).
The procedure volume and sedation data we collected are summarized in Table 2. During the prephase of our study, 4514 advanced endoscopy cases were performed; 762 (16.9%) of these cases used anesthesia services. During the post policy period, 5084 advanced endoscopy cases were performed, of which 5053 (99.4%) of these cases used anesthesia services. An increase of 12.6% in advanced endoscopy case volume occurred in the 2.5 years immediately after the introduction of anesthesia care for all patients. Our policy change was successfully executed; only 6 of 4982 (0.1%) upper EUS and ERCP cases received RN sedation after implementation. Twelve ERCP cases (5 of which were repeat procedures in a single patient), no upper EUS patients, and 19 lower EUS patients elected to have procedures without sedation during the entire 5-year study period. After the policy change 15 of 102 lower EUS patients received moderate sedation by an RN in an agreed-upon exclusion to the policy. The lighter plane of sedation required is also reflected in the much lower number of lower EUS patients in the precohort requiring anesthesia referral (1.6% vs 17.3% for upper EUS and ERCP).
We reviewed our complications data between the pre- and postphases, as well as the demographics of the patients who experienced complications. Among these patients who experienced complications, there was no difference in the gender distribution between those in the precohort versus those in the postcohort (Table 3). Similarly, there was no difference in the percent of patients aged 71 years and older, though the mean age of the postcomplications group was 6 years younger (58 years) than the pregroup (64 years). We did not find a statistically significant difference in the distribution of ASA physical status classes among the pre- and postcomplication patient groups.
Total complication rates and major morbidity and mortality were not statistically different between the 2 periods (Table 4). Endoscopic complications comprised the largest proportion of complications in both groups with the most common endoscopic complications being perforation, hemorrhage, pancreatitis, cholangitis, and postprocedure pain. While the rate of endoscopic complications increased before and after the policy change (0.66% and 0.87%, respectively), this difference was not found to be statistically significant (P = 0.293, diff = 0.2, 95% CI, −0.16% to 0.56%). To detect a statistically significant difference between these 2 proportions, would require a sample size of 28,000 procedures in both the pre- and the postcohorts, which we were unable to obtain in this single-center study. Similarly, endoscopic complications comprised the largest percentage of those complications classified as major morbidity/death. Within this subanalysis, we found a similar nonsignificant difference in the rate of endoscopic complications pre- (0.18%) and post (0.30%) (P = 0.297, diff = 0.1, 95% CI, −0.11% to 0.34%).
Reported sedation complications comprised a smaller proportion of our total complications, most were not considered morbid per our study, and not all were respiratory complications (Appendix 1). In the pregroup, there were 17 sedation complications (0.38%) of which 15 were RN sedation cases and 2 were anesthesia care. In the postcohort, there were only 4 sedation-related complications (0.08%), which was statistically significant (P = 0.002, diff = 0.3, 95% CI, 0.11%–0.53%). We did have sufficient power to demonstrate this difference as statistically significant (4645 procedures per cohort). However, in the major morbidity/mortality subanalysis, there were only 3 reported major sedation complications during the entire study period, which did not differ significantly before and after the policy change (P = 0.604) (Table 4).
When we analyzed the rate of complications by the type of sedation actually received rather than the time period when the procedure occurred, we found a similar statistically significant difference in the rate of sedation complications (Table 5). When RN sedation patients were compared with patients who received anesthesia carea sedation-related complications occurred more frequently in the RN sedation group (0.40%) versus the anesthesia care group (0.10%), a difference that was significant (P = 0.003). However, endoscopic complications were not significantly different when analyzed in this manner (P = 0.720).
We conducted additional analyses comparing the sedation complications pre and post with respect to the staff members who provided sedation services. There were 5 interventional gastroenterologists in our study who were each responsible for a similar case volume, who supervised procedural sedation by 40 different staff endoscopy nurses during the prephase of our study. The sedation complication rate among the endoscopists varied from 0% to 0.56%, and there were no significant differences among them (all P > 0.095). Of the 40 total staff nurses, 20 administered procedural sedation to 75% of the patients, and we found no significant differences among these high-volume providers (all P > 0.069). The rate of observed sedation complications in this group was 0.40%, while the observed rate of the other 20 nurses who provided care less frequently was 0.61% (P = 0.322).
During the postphase of our study, 1 anesthesiologist provided or supervised anesthesia services for 69% of the cases, while 15 additional anesthesiologists rotated within the unit for the remaining 31% of the cases. The sedation complication rate of the single anesthesiologist who provided most of the services was 0.03% whereas that of the combined additional anesthesiologists was 0.19%, though this was not statistically significant (P = 0.91). Similarly, 28 CRNAs rotated within the unit, though only 3 sedation complications occurred in cases staffed by CRNAs. The sedation complication rates among the CRNAs ranged from 0% to 0.55%, and the differences among these rates were not significant (all P > 0.533).
We present a comparison of reported complications between primarily a nurse-administered sedation practice and a required anesthesia care policy for patients undergoing EUS and ERCP. In this cohort of 9598 procedures over 5 years, there was a significant difference in the rate of sedation complications, but no significant difference in the rates of total complications, major morbidity, or mortality. We analyzed these data both on an intention-to-treat (before and after the policy change) and on the actual provider administering sedation (RN-administered versus anesthesia care). We found that both approaches did not impact the rate of endoscopic complications.
Our data demonstrate that anesthesia care prevented the reported respiratory complications of nurse-administered benzodiazepine-opioid deep sedation in an ill population for advanced endoscopy in a tertiary care institution. Eight of 3718 (0.2%) of these relatively elderly and ill patients undergoing deep sedation by RNs had the reported hypoxic and hypoventilation complications that the American Society of Anesthesiologists guidelines and CMS policies seek to prevent. Seven of the 8 were successfully rescued by the personnel in the room and the remaining 1 by the hospital code team. All were discharged home immediately after the procedure without reported sequelae. This retrospective analysis of QA data cannot examine whether these events can be considered minor. Our data do suggest, however, that in our institution the nursing and physician personnel present for these cases have skills to rescue patients and prompt adequate backup. This is consistent with other endoscopic RN sedation data in various procedure and sedation types.7,12 Drug errors, arrhythmias, and major morbidity were similar between the 2 groups.
Anesthesia care may have exposed patients to rare risks, some of which appear here as “other” complications. Three dental injuries were reported in our anesthesia care group and none in the RN sedation group. One was clearly attributable to anesthesia instrumentation; for the other 2 (one a swallowed dental appliance and the other documented to have intact teeth after video laryngoscopy), it was not clear which instrumentation was responsible for injury to fragile dentition. Anesthesia could increase patients’ risk of dental injury with any instrument, even if not with a laryngoscope. Two anaphylaxis events occurred in the anesthesia group, one to an anesthetic drug (cisatracurium) and one to an endoscopic drug (fluorescein). Nursing care by its nature is relatively protocolized; in our institution, it exposes patients to many fewer drugs. Anesthesia care may be more likely to involve allergenic drugs, including neuromuscular blockers, that are the most common cause of intraoperative anaphylaxis.13 There was not sufficient power to demonstrate a difference in these rarer outcomes.
In our population of advanced endoscopy patients, we found fewer sedation complications than endoscopic complications. Thus, improving sedation safety with anesthesia care did not result in a significant difference in total complications or morbidity and mortality in our large cohort. Endoscopic complications increased from 0.66% to 0.87% before and after the policy change (P = 0.293); however, our study was not adequately powered to demonstrate this difference as statistically signficant, requiring nearly 60,000 procedures for adequate power. However, these changes may still be clinically significant, as anesthesia care and increased compliance with endoscopic interventions could also permit more invasive approaches that result in more complications. Ootaki et al.14 demonstrated an improvement in cytopathologic outcomes with general anesthesia for EUS, presumably due to enhanced patient stillness. Our study, with historical controls, can only raise this possibility for further investigation.
Because the change in sedation complications pre and post was significant, we also tested these rates for significance with respect to the sedation providers. These analyses demonstrated no significant differences among the provider groups. No 1 gastroenterologist or CRNA was responsible for a significant rate of sedation complications. We tested the gastroenterology nurses and anesthesiologists by volume, as a single provider or small number of providers staffed a large proportion of the cases. The sedation complication rates of these high-volume providers were also not statistically significant from those clinicians who staffed fewer cases.
We used QA data from 2 sources: the gastroenterology database and the hospital quality and safety reporting system. The gastroenterology database has significant advantages: its collection is mandated for each procedure before the clinical record can be completed, and it is prospectively collected in real time. Because our data depend on self-reporting and use historical controls, they are vulnerable to changes in provider reporting practices over time; however, all procedures and QA data acquisition were performed by the same 5 experienced gastroenterologists for the duration of the study.
We also rely on cross-disciplinary reporting of adverse sedation events by gastroenterologists, which could be biased in either direction. Gastroenterologists might have been inclined toward under-reporting of sedation complications by RNs under their supervision, or conversely, gastroenterologists might not have been aware of intra-anesthetic events without postprocedure morbidity. We made a prospective decision not to use the anesthesia department QA database, because this source would not be available for the nursing sedation patients and thus would not provide useful comparative data. The hospital quality and safety database, by comparison, gather reported safety events from any source, was available for both groups of patients, should be sensitive to late complications, and should help correct for failures of gastroenterologist observation. Berzin et al.15 report a prospective assessment of sedation-related intra-anesthetic events that may or may not have had clinical sequelae. We sought only complications serious enough for quality improvement reporting.12
Anesthesia care has been established to be safe and effective for high-volume endoscopic care, including advanced endoscopy and for patients with cirrhosis.16–19 It has also, however, been asserted to result in high cost (particularly for screening colonoscopy) without clearly established value for patients.1 Three groups have recently examined whether anesthesia care enhances endoscopic outcomes: this was not proven for adenoma detection in colonoscopy or for cannulation rates in ERCP but was recently established to increase the cytopathologic diagnostic yield of EUS, presumably by enhancing procedure quality and patient stillness.14,20,21 We believe, based on our data, that anesthesia care adds sedation safety value as well.
We did not address the cost of implementation of this expanded anesthesia program, which includes daily staffing of 3 dedicated anesthetizing locations in the endoscopy unit, a telephone preanesthesia assessment program, and the associated administrative, engineering, and equipment costs. However, this replaced an existing model of anesthesia care in which an unpredictable 17% of patients were considered inappropriate for nursing sedation and awaited the availability of an anesthesia team, introducing delays and inefficiencies for both services. The policy change permitted a 13% increase in advanced endoscopy case volume.
There are multiple axes of debate in anesthesia care for endoscopy: by whom, for which procedure, for which patients, at what level of sedation or general anesthesia, using which drugs, and producing what value, both outcome and cost. Each contribution to this rapidly expanding literature addresses some permutation of these questions. Our study, by its nature, can address only a few. We did not protocolize our care or address the propofol question. We here describe the sedation and total safety effect of a policy altering the personnel assigned, in an ill tertiary care population, for ERCP and EUS.
This study demonstrates that anesthesia care improves sedation safety, measured by quality improvement reporting, in advanced endoscopy. We studied a large, high-risk cohort in a tertiary care institution requiring the most invasive of procedures, requiring deeper levels of sedation and demonstrated a significant improvement in the rates of QA-reported sedation-related complications but no decrease in total complications, in major morbidity, or in mortality. Additional studies are warranted to demonstrate whether this finding is generalizable to other endoscopy centers and for elective anesthesia for other endoscopic procedures such as colonoscopy.
APPENDIX 1: Complications: Descriptive
Prepolicy change, the 17 sedation complications (15 RN sedation, 2 anesthesia care) were as follows:
* 9 hypoxemic events (8 RN sedation, 1 anesthesia care)
* 1 resulted in unsuccessful conservative airway maneuvers and then intubation (major morbidity)
* 2 drug errors with no clinical sequelae (1 anesthesia care, and 1 RN sedation)
* 4 dysrhythmias (bradycardia, SVT, and AF)
* 1 failed RN sedation requiring rescheduling for anesthesia
* 1 mortality: a patient with metastatic cancer and terminal hepatic failure died 3 days postprocedure. She was delirious preprocedure and never regained consciousness after sedation with midazolam/meperidine. She was not aggressively treated at the time. The cause of death was not determined. Procedural sedation may have played a role.
Postpolicy change, all with anesthesia care, the 4 sedation complications were as follows:
* 2 dysrhythmias, both atrial fibrillation
* 1 dental injury clearly attributable to anesthesia care
* 1 aspiration event resulting in pneumonia and ICU admission (major morbidity).
“Other” complications included the following:
* 1 fatal autopsy-diagnosed STEMI
* 2 episodes of anaphylaxis
* 2 dental injuries not clearly attributable to either endoscopic or anesthetic instrumentation
* 1 episode of chest pain diagnosed in the emergency room as esophageal spasm.
* There were 8 morbid nonfatal events “pre” and 12 “post.”
* Two of the highly morbid events were related to sedation:
° a code call for a respiratory arrest in an RN sedation patient who made a rapid recovery in the “pre” phase,
° and an ICU admission for aspiration pneumonia in an anesthesia care patient in the “post” phase, eventually fully recovered after several days of critical illness.
* There were 16 morbid nonfatal endoscopic complications (6 “pre” and 10 “post”).
* The most common morbid endoscopic complications were perforations, pancreatitis, and cholangitis.
* There were 2 anaphylaxis events, 1 to cisatracurium (under anesthesia care in “pre”) with uncomplicated recovery and 1 to fluorescein which resulted in stroke (under anesthesia care “post”).
* 9 of our 9598 patients died immediately following or as the result of the procedure, 4 “pre” and 5 “post.”
* 2 patients had fatal air embolisms during ERCP (both “post”).
* 3 moribund patients with terminal cancer died within a few days, of their underlying illness:
° sedation was thought to have contributed to 1 patient’s continuing deterioration in mental status (“pre”);
° another had biliary and gastric outlet obstruction by tumor and was unresponsive on presentation, and had both aspiration and possible perforation apparent after his procedure (“pre”).
° a third had cholangitis prior to ERCP and fulminant sepsis immediately following (“post”).
* 1 patient had a fatal STEMI 24 hours after an uncomplicated EUS, with full postmortem (“pre”).
* 2 patients died of late complications of post-ERCP pancreatitis (1 “pre” and 1 “post”).
* 1 patient, with metastatic cancer and poor prognosis, died after referral to hospice when pneumoperitoneum was found after ERCP (“post”).
Name: Emily S. Guimaraes, MD.
Contribution: This author helped design the study, assisted with the data analysis, and prepared the manuscript.
Attestation: Emily S. Guimaraes approved the final manuscript.
Conflicts of Interest: The author has no conflicts of interest to declare.
Name: Emily J. Campbell, MPH.
Contribution: This author helped design the study, provided data collection and analysis, and assisted with the preparation of the manuscript.
Attestation: Emily J. Campbell approved the final manuscript and attests to the integrity of the original data and the analysis reported in this manuscript and is the archival author.
Conflicts of Interest: The author has no conflicts of interest to declare.
Name: James M. Richter, MD, MA.
Contribution: This author helped design the study and assisted with the data analysis and the preparation of the manuscript.
Attestation: James M. Richter approved the final manuscript and attests to the integrity of the original data and the analysis reported in this manuscript.
Conflicts of Interest: Dr. Richter works as a consultant for Policy Analysis, Inc., and the United States Department of Justice.
This manuscript was handled by: Sorin J. Brull, MD, FCARCSI (Hon).
The authors would like to thank Hui Zheng, PhD (Massachusetts General Hospital, Biostatistics Center and Harvard Catalyst Program), for his statistical consultation on this manuscript.
a Patients who had no sedation were excluded from this analysis. Cited Here...