“Opt Out” and Access to Anesthesia Care for Elective and Urgent Surgeries among U.S. Medicare Beneficiaries

IN the United States, anesthesia is most commonly provided by anesthesiologists (i.e., physicians with specialty training in anesthesiology), nurse anesthetists, and anesthesiologist assistants. Similar to physician assistants more generally, anesthesiologist assistants only provide care under the supervision of an anesthesiologist. Generally, nurse anesthetists are supervised by an anesthesiologist, with or without the anesthesiologist formally billing. Occasionally, they are supervised by a proceduralist (e.g., the surgeon performing the case). Rarely, anesthesia is provided by the proceduralist alone. Concerns over access to anesthesia care—as well as predicted future shortages of anesthesiologists¹ —have led policymakers to consider loosening the degree to which nurse anesthetists must be supervised, up to and including independent practice.

In 2001, the U.S. Centers for Medicare and Medicaid Services (CMS) issued a rule allowing states to “opt out” of the requirement that nurse anesthetists be supervised by a physician (either an anesthesiologist or a proceduralist) in order for providers to receive payment from Medicare for the anesthetic. By 2013, 17 states had decided to “opt out” of the Medicare regulations requiring supervision of nurse anesthetists; in some states, this allows for independent nurse anesthetist practice.² 

Debate over the merits of “opt out” has largely focused on whether it has affected the quality of anesthesia care.^3–12  Less work has addressed whether it has increased access to care, the normative intent of the administrative rule. Indeed, the rule states that governors must consider the availability of anesthesia care before deciding whether to “opt out.” Moreover, in choosing to “opt out,” governors have typically cited the necessity of improving access to anesthesia care.^13,14  Whether “opt out” has increased access remains an open question with important policy implications not just for surgical care but also for health care more broadly. In an effort to control health care spending, policymakers are considering laws and regulations that would expand the autonomy of other midlevel providers, such as nurse practitioners, in the belief that doing so will reduce costs and increase access to care.¹⁵  Understanding whether “opt out” has increased access to care could provide a useful test case for gauging these laws’ likely effectiveness. If “opt out” has not improved access to care, understanding why may help policymakers improve these laws in order to improve access.

Two previous studies examined the association between “opt out” provisions and the number of surgical procedures performed, particularly urgent procedures; no association was detected.^16,17  While useful, these studies shed light on only one measure of access to care and focused on the use of urgent surgical procedures.

For this article, we used administrative claims data from Medicare to examine whether “opt out” was associated with a different dimension of increased access to care: the distance patients travel to obtain surgical procedures. We examine this for five elective procedures common among the Medicare population: total knee arthroplasty (also known as total knee replacement), total hip arthroplasty (THA; also known as total hip replacement), colonoscopy/sigmoidoscopy, esophagogastroduodenoscopy, and cataract surgery, as well as two common urgent procedures where timely access to surgical care is important: appendectomy and hip fracture repair.

By facilitating independent nurse anesthetist practice, “opt out” could reduce the distances traveled by patients in two ways. First, the increased supply of anesthesia providers might prompt the opening of new facilities (e.g., hospitals or ambulatory surgery centers). Second, the increased supply of anesthesia providers might also allow new services (e.g., an orthopedic service) to be established in existing hospitals. Distance traveled has been used in many other nonsurgical studies to measure a dimension of access to care.^18–20  Increased travel distance has been shown, in some studies, to be associated with poorer surgical outcomes.^21,22  Finally, regardless of its effects on outcomes, increased travel distances represent an inconvenience for patients, particularly since absolute distances between patients and hospitals influence the choice of hospital for surgical care.^23–25 

Our data consisted of health insurance claims for a random 20% sample of U.S. Medicare beneficiaries enrolled in the traditional fee-for-service Medicare plan. In the United States, Medicare is a public insurance program that primarily provides health insurance for the elderly (persons 65 yr or older) although the program also covers some younger persons with significant disabilities and those with end-stage renal disease. In 2010, more than 80% of Medicare beneficiaries consisted of persons aged 65 yr and older.²⁶  As a general rule, Medicare beneficiaries can choose either to be enrolled in the traditional fee-for-service plan, for which the federal agency administering Medicare—the CMS—is the primary payer, or they can choose to be enrolled in a managed healthcare plan. Under the latter, Medicare essentially subcontracts out the provision of health care to private health insurers, who bear all the costs for an individual’s care. Roughly two thirds of Medicare beneficiaries are enrolled in the traditional fee-for-service plan.²⁷  Health insurance claims data for beneficiaries enrolled in the traditional fee-for-service plans are available for researchers upon approval of a Data Use Agreement with CMS and payment of required fees.

The Medicare data are detailed. They include demographic information such as age, race, sex, county of residence, and zip code (also known as postal code) of residence. The data also provide diagnosis codes (International Classification of Disease, Ninth Edition), which can be used to draw inferences about a patient’s comorbidities, as well as procedure codes (e.g., Current Procedural Terminology^® codes used to classify specific procedures) that can be used to identify whether they received a given procedure.

We constructed a sample consisting of all persons who received a total knee arthroplasty, THA, hip fracture repair, appendectomy, cataract surgery, colonoscopy/sigmoidectomy, or esophagogastroduodenoscopy between 1999 and 2011 and who were continuously enrolled in Medicare fee-for-service for the entirety of the calendar year they received their procedure. We identified patients who received these procedures by isolating claims submitted by a healthcare provider (located in the carrier claims file) with a procedure code (i.e., Current Procedural Terminology^® code) associated with the given procedure. A list of the relevant procedure codes and the references used to identify these codes is provided in the supplemental content (appendix table A, https://links.lww.com/ALN/B360). These claims were then linked to a corresponding claim submitted by a healthcare facility (e.g., hospital or ambulatory surgery center). Medicare does not provide any explicit way of linking provider claims to facility claims. Each provider claim reports the dates for which the patient received care at the facility; we assumed that a facility claim was linked to the provider claim if the date on the provider claim fell within the range of dates listed on the facility claim.

After excluding persons who received more than one of the procedures we studied in a given calendar year, we arrived at an initial sample of 3,818,148 cases. We then applied several exclusion criteria. First, we excluded cases in which we identified more than one facility claim associated with the given carrier claim (n = 135,436), which occurred when there were two or more facility claims whose dates overlapped the carrier claim. Second, total knee arthroplasty, THA, cataract surgery, hip fracture repair, and appendectomy are nearly always performed at a surgical facility (i.e., hospital). By contrast, colonoscopy and esophagogastroduodenoscopy can be performed in a physician’s office. Therefore, we excluded claims for total knee arthroplasty, THA, cataract surgery, hip fracture repair, and appendectomy for which we could not identify an associated facility claim (n = 43,395). Third, colonoscopy, esophagogastroduodenoscopy, and cataract surgery do not always require anesthesia care. Since our interest was in cases that require anesthesia care, we excluded colonoscopy, esophagogastroduodenoscopy, and cataract surgery cases without a corresponding anesthesia claim (n = 2,186,866) from our main analysis although we retained these cases for some secondary analyses in the Secondary Analyses subsection. To identify a corresponding anesthesia claim, we isolated carrier claims with an appropriate anesthesia procedure code (supplemental content, appendix table A, https://links.lww.com/ALN/B360) on a date corresponding to the same date as the procedure itself. Fourth, we excluded cases where we could not identify the patient or facility zip code (n = 134,256) or the case was missing other patient demographic information (n = 9,993). Fifth, we excluded cases where the calculated distance was more than 500 km (310 miles; n = 25,536). The rationale for this exclusion was that patients who traveled this far for their procedure were likely doing so for reasons unrelated to the availability of care in their local area (e.g., presence of family or an urgent procedure occurring while on vacation). Finally, we excluded patients under the age of 65 (n = 139,851). Our final sample consisted of 1,142,815 cases. A flow diagram describing how the sample was created is provided in the supplemental content (appendix fig. 1, https://links.lww.com/ALN/B360).

Our outcome of interest was the distance traveled by the patient to receive their procedure. Our basic approach was to identify the zip code of the patient, as well as the zip code where the procedure took place using the methods described in the subsequent paragraph. We then used data from the U.S. Census to identify the latitude and longitude for the central point within each zip code.²⁸  Using this latitude and longitude, the Haversine formula was then used to calculate the distance between two zip codes. This approach has been used in many previous studies to identify the distances between patients and hospitals/nursing facilities.^18–20,25 

Identifying the zip code of each patient’s residence is straightforward, as it is directly provided in the annual demographic files provided by Medicare. For our study, there are two potential zip codes where the procedure may have taken place: the zip code of the facility and that of the physician. The latter, provided by the physician, may represent a mailing address or the location of the physician’s office. For the procedures of total knee arthroplasty, THA, appendectomy, cataract surgery, and hip fracture repair, we used the zip code of the facility, as these procedures almost certainly take place in a surgical facility. Since colonoscopy and esophagoduodenoscopy can take place in a surgical facility or a physician’s office, we used the facility zip code if present but, otherwise, the zip code of the physician.

Our primary independent variable of interest was whether “opt out” was in effect in the given state during the year the procedure was performed. Specifically, we created a dummy variable equaling 1 in the “opt out” states starting the year after the “opt out” year (supplemental content, appendix table B, https://links.lww.com/ALN/B360). For non-“opt out” states and for “opt out” states in the years before and including the “opt out” year, the variable equaled zero. It is important to note that 2011 was the last year for which we could obtain data. Since Kentucky “opted out” in 2012, our analysis, therefore, implicitly treated Kentucky as a non-“opt out” state (i.e., did not account for the effect of “opt out” in Kentucky).

In addition to our “opt out” variable, we also obtained several covariates at the patient and county levels. At the patient level, we incorporated age, sex, and race as controls, as well as the comorbidities listed in table 1. We chose these comorbidities as they are constituents of the Elixhauser set of comorbidities, which is frequently used for risk adjustment.²⁹  To determine the presence of a comorbidity, we extracted all of an individual’s inpatient, outpatient, and carrier claims that were filed during the year of surgery. A patient was then deemed to have the given comorbidity if they had at least one claim with a relevant International Classification of Disease, Ninth Edition, diagnosis code.³⁰ 

At the county level, we incorporated several controls based on the patient’s county of residence: total population, share of the population that was white, share of the population that was male, share of the population aged more than 65 yr, and median income for the county. These variables were obtained from the U.S. Census Bureau. They were chosen as they could potentially affect the market for hospital and healthcare services.³¹ 

A simple cross-sectional comparison of travel distances across zip codes could be subject to confounding since states that choose to “opt out” differ in many ways from states that did not. For example, Iowa was the first state to “opt out,” and it is geographically large compared to Delaware, a state that has not chosen to “opt out.”

While our analysis did include several controls for observable differences between “opt out” and non-“opt out” areas—such as differences in income and patient comorbidities—a straightforward cross-sectional comparison could still be vulnerable to confounding from unobservable differences between states. Therefore, we utilized a difference-in-differences analysis to reduce confounding further. The difference-in-differences approach is frequently used in policy analysis^32–34  and has been used in the previous literature examining the effects of “opt out.”^16,17  Under this approach, zip code–specific controls were used to adjust for unobservable zip code level factors. Therefore, rather than comparing travel distances across areas, the difference-in-differences approach identifies the effect of “opt out” by estimating within zip code changes in travel distance over time among the “opt out” states.

Simple “before-after” comparison can still be confounded by secular time trends, such as general changes in surgical practice over time. Therefore, the second step of a standard difference-in-differences approach involves the use of year effects in order to control for secular time trends at the national level. In addition to these controls, we incorporated additional linear and quadratic state trends to control for unobserved, secular trends occurring at the state level.

We implemented our difference-in-differences analysis by using regression analyses in which the dependent variables were measures of distance traveled, and the key independent variable was the measure of whether opt out is in effect in a given state. The models controlled for the patient and county factors (table 1), as well as zip code controls, year controls, and controls for linear and quadratic state trends.

For roughly 20% of our cases (n = 219,767), the patient’s zip code and the facility’s zip code were the same, leading to an estimated travel distance of 0. When there are many observations for which the dependent variable is 0, a simple regression analysis will tend to be downward biased; in other words, the estimated effect will be lower in magnitude than the true effect. As a result, a simple regression analysis would be biased toward a finding that “opt out” had no effect on access.³⁵  To address this issue, we performed two regressions. The first analysis was a multivariable regression in which the dependent variable was an indicator variable equaling 1 if the patient and the facility shared the same zip code, and the independent variables were the variables outlined in the Independent Variables subsection. Our coefficient of interest was the coefficient associated with our “opt out” variable, which can be interpreted as the effect of “opt out” on the absolute percentage point change on the percentage of patients travelling outside of their zip code for their procedure. Our second analysis was a multivariable linear regression in which the dependent variable was the actual estimated travel distance. The independent variables were the same as those described for the previous analysis. This analysis was restricted to patients with a nonzero travel distance. For this second analysis, our coefficient of interest was the coefficient associated with our “opt out” variable, which demonstrates the change in travel distance associated with “opt out” among patients with a nonzero travel distance (i.e., among patients who traveled outside their zip code). This two-step approach has been used in other studies to obtain nonbiased estimates when a large proportion of observations assume a value of 0.^36,37 

Zip codes do not have the same size (land area),²⁴  so our distance measures are sensitive to the size of a given zip code. Mechanically, patients residing in larger metropolitan zip codes (i.e., larger land area) are less likely to have to travel outside of their zip code to receive care and should they do so are more likely to travel a longer distance to do so (since the land area of the zip code is larger). Our use of zip code fixed effects addresses this issue since these effects would control for unobservable zip code factors that are fixed over time, such as land area. For some descriptive analyses, we did adjust our results for land area (Technical Appendix in the online supplemental content, https://links.lww.com/ALN/B360).

We performed our analyses with STATA 14.0 (StataCorp, USA) and adjusted our SEs and CIs for clustering at the state level.³⁸  Further details of our approach are found in the Technical Appendix in the online supplemental content (https://links.lww.com/ALN/B360).

We conducted several sets of secondary analyses. First, we conducted two sensitivity analyses to examine the robustness of our results to alternative measures of access. Of the seven procedures we examined, three (colonoscopy, esophagastroduodenoscopy, and cataract surgery) do not always require anesthesia, by which we mean here and elsewhere the presence of an anesthesiologist and/or nurse anesthetist. In the baseline analyses described in the Statistical Analyses subsection, we excluded patients undergoing these three procedures who did not receive anesthesia. Therefore, for these three procedures, our estimates reflect the effect of “opt out” on the travel distances for patients undergoing these procedures who ultimately received anesthesia for their procedure. However, “opt out” could also make it more likely that patients undergoing these procedures would receive anesthesia to begin with. To assess this possibility, we conducted a separate set of analyses for these three procedures in which we included patients undergoing these procedures who did not receive anesthesia (n = 1,639,944). For these analyses, we estimated a linear regression similar to the analyses described in the previous subsection, where the dependent variable was an indicator variable assuming the value of 1 if the patient received anesthesia for their procedure and 0 otherwise. Our independent variable of interest was our “opt out” variable, and we incorporated the same set of additional controls that were used for our baseline analyses.

Second, suppose that, as a result of “opt out,” a new surgery center opens, thereby providing the opportunity for patients to receive a total knee arthroplasty closer to home. However, suppose that the vast majority of patients living in the zip code decline to use the surgery center. Our baseline analyses would generally identify no change in access—given the small proportion of patients using the new surgery center, the effect on overall travel distance will also be small. To address this possibility, we, therefore, conducted analyses where the dependent variable was the annual minimum distance traveled by any patient within the given zip code (for a given procedure). In contrast to the previous analyses, where the unit of observation was the individual patients, these analyses were performed at the zip code level using the same covariates described previously, except that we converted patient-specific variables (e.g., age) to average values at the zip code level (e.g., average age at the zip code level). Because using the minimum distance traveled by any patient is susceptible to outliers, we also considered an alternative specification in which we used the fifth and tenth percentiles of distance traveled. For a given zip code, there need to be at least 19 patients receiving for a given procedure in order to estimate the fifth percentile without bias and at least nine patients for the tenth percentile. Therefore, for these additional analyses, we used the minimum distance traveled for those zip codes that did not have sufficient patients to calculate the given percentile.

We also conducted two subanalyses in which we examined whether the effects of “opt out” were different between two populations most likely to see a benefit: patients residing in zip codes with long travel distances to begin with and low-risk patients. Hypothetically, low-risk patients may be more likely to benefit from “opt out” since sicker patients may be more likely to choose or require care at a tertiary referral center, regardless of travel distance. For the first subanalysis, we restricted our sample to zip codes whose initial average travel distance was greater than 50 km (roughly 30 miles). For the second subanalysis, we used a previously developed empirical weighting of the Elixhauser comorbidities.²⁹  As a score of 10 on this index broadly corresponds to a score of 3 on the Charlson index, we restricted our sample to patients with a score of 10 or less.

The construction of our study sample (such as the inclusion and exclusion criteria), as well the elements of our baseline analysis (such as the choice of statistical model, choice of outcome, and choice of additional covariates), was determined before initiating our analysis. Our additional analyses were conducted post hoc based on the finding of our baseline analysis, as well as comments made during the revision process.

Table 1 presents descriptive statistics for our sample, which included 954,789 cases taking place in non-“opt out” states (i.e., states that never chose to “opt out”) and 188,026 cases that occurred in “opt out” states (i.e., states that ultimately chose to “opt out”). For the cases occurring in “opt out” states, 112,221 occurred before “opt out’ and 75,085 occurred afterward. Patients undergoing a procedure in an “opt out” state were more likely to be males (38.0% vs. 37.5%; P < 0.001) and whites (90.1% vs. 88.5%) and resided in counties with higher median incomes ($48,362 vs. $47,627; P < 0.001). They were also more likely to reside in rural counties, as defined using the 2006 Office of Management and Budget delineation of rural counties³⁹  (32.7% vs. 23.1%; P < 0.001). The presence of every comorbidity we included in our analysis was higher among patients in non-“opt out” states (P < 0.001 for every comorbidity).

Table 2 presents descriptive statistics concerning travel distances in our sample. The first set of columns report the percentage of patients who traveled outside of their home zip code for each of the procedures, adjusted for zip code land area. Across all the procedures, a solid majority of patients traveled outside of their home zip code (ranging from 77 to 88% of patients). The table also reports the average distances traveled, ranging from roughly 20 km for patients undergoing colonoscopy/sigmoidoscopy to roughly 40 km for patients undergoing THA. There were no significant differences between the “opt out” and non-“opt out” states for either of these distance measures.

Figure 1 provides descriptive trends of the average travel distances in our sample, stratified by patients located in “opt out” and non-“opt out” sates. To calculate these average travel distances, we first calculated the average travel distance for each procedure and then took the average (of these averages) across all seven procedures. Qualitatively, the figure suggests that “opt out” may not have had an effect on travel distances. Although travel distance fell in the “opt out” states fell relative to travel distances in non-“opt out” states between 1999 and 2003, this trend predated “opt out,” which at the earliest took place in Iowa in December 2001. Between 2003 and 2007—when eight (of a total of 17) states implemented “opt out,” travel distances in the “opt out” states remained fairly level, similar to travel distances in the non-“opt out” states.

For the most part, “opt out” was not associated with a decrease in the percentage of patients travelling outside of their home zip code (table 3), with point estimates ranging from a 0.973% point (absolute) decrease in the case of esophagogastroduodenoscopy (95% CI, −1.97 to 0.0329; P = 0.058) to a 1.02% point (absolute) increase in the case of appendectomy (95% CI, −6.68 to 8.73; P = 0.791). The sole potential exception was THA, where we observed a statistically significant decrease of 2.18% points (95% CI, −3.72 to −0.636; P = 0.007, without correction for multiple comparisons).

For patients travelling outside of their zip code, “opt out” had no significant effect on the distance traveled across any of the procedures we examined, with point estimates ranging from a decrease of 7.9 km for appendectomy (95% CI, −19.4 to 3.59; P = 0.173) to an increase of 1.55 km (95% CI, −5.07 to 8.17; P = 0.641) for THA. Crucially, for all procedures (except appendectomy), even the lower bounds of our 95% CI suggest decreases that would be of negligible practical significance compared with the distances that influence patient decision making,^23,25  suggesting that the lack of estimated effect represents a true null effect, as opposed to imprecision in our estimates.

For our robustness analyses, we first examined whether “opt out” was associated with an increase in the percentage of patients received anesthesia for colonoscopy/sigmoidoscopy, esophagogastroduodenoscopy, or cataract surgery (table 4). Overall, we found that “opt out” was not associated with a significant change in the percentage of patients receiving anesthesia for colonoscopy/sigmoidoscopy (0.391% point decrease; 95% CI, −2.70 to 1.91; P = 0.952) or esophagogastroduodenoscopy (0.903% point increase; 95% CI, −1.48 to 3.29; P = 0.451). We may have observed a small decrease for cataract surgery (1.49% point decrease; 95% CI, −2.92 to −0.0538; P = 0.042, without correction for multiple comparisons).

We also considered alternative specifications in which we considered whether “opt out” was associated with a decrease in the minimum distance traveled by patients in a given zip code (table 5). Similar to the results shown in table 3, we found that “opt out” was not associated with a significant change in travel distance using these alternative measures for any of the procedures we considered. Alternative specification examining the effect of “opt out” on the fifth percentile and tenth percentile distances traveled by patients in a given zip code also showed no significant effect (supplemental content, appendix table C, https://links.lww.com/ALN/B360).

Finally, we examined the effect of “opt out” between two subpopulations where it might be expected to have a larger effect: patients residing in zip codes with long travel distances initially and low-risk patients (table 5). For both groups of patients, our results suggest that “opt out” was not associated with reduction in travel distance.

In this article, we examined the effect of “opt out” on access to anesthesia care using a commonly used measure of access: patient travel distances. Overall, we found that “opt out” was not associated with an increase in access as measured by the percentage of patients who avoided travel outside of their home zip code to receive their procedure or as measured by the travel distances themselves. We did find that “opt out” was associated with a 2% point reduction in the proportion of patients traveling outsider their home zip code for THA; however, given that well more than 80% of THA patients traveled outside their home zip code (table 3), this effect is of little practical significance. In addition, we found that “opt out” had no effect on the use of anesthesia for patients undergoing colonoscopy/sigmoidoscopy or esophagogastroduodenoscopy and was actually associated with a slight decrease in the use of anesthesia for cataract surgery. Finally, we did not observe any significant effect on travel distances between two groups of patients where “opt out” might have larger effects: low-risk patients and patients residing in zip codes with initially long travel distances. In general, even the lower bounds of our 95% CIs suggested effects that were of negligible magnitude from a practical standpoint, suggesting that our results are not due to lack of statistical power/precision, with the lone exception being our analysis of patients residing in zip codes with initially long travel distances.

There are many reasons why “opt out” could fail to increase access. First, “opt out” only means that a federal insurer (CMS) will pay for cases where a nurse anesthetist is unsupervised by a physician; a whole host of other parties—such as private insurers, hospitals, surgeons, and patients—may have their own preferences and impose their own requirements. To the degree that these other parties require physician supervision, “opt out” may not achieve its intended effect. Second, the implicit assumption of “opt out” is that availability of anesthesia care is the factor that limits access to surgical procedures. However, other factors, such as the availability of surgeons and hospitals, could affect access and would not be addressed by “opt out.” Third, “opt out” would have no effect for those instances where the nurse anesthetist is directly supervised by the proceduralist.

Our results should be viewed in light of their limitations. First, while using the distance between zip codes is an accepted method in the literature, one shortcoming is that we are unable to observe changes in distance within a given zip code, so that if “opt out” shortened travel distances within a zip code (e.g., if a second, closer surgery center opened within a given zip code), we would not observe this effect. However, it does seem unlikely that the effect of “opt out” would be only to shorten distances within a zip code, without affecting the broader measures of distance that we examined. Second, since our study was conducted among the Medicare population, we were unable to ascertain the effect of “opt out” on travel distances for procedures that were rarely performed in this population but may be important from a policy standpoint. Third, while our study considered several commonly performed procedures, it is possible that “opt out” may have reduced travel distances for procedures that we did not examine. Crucially, our study does not address whether “opt out” may have improved access to care for obstetrics procedures. Fourth, we cannot exclude the possibility that the lack of effect could be explained by other confounding variables, such as unobserved factors occurring at the zip code or patient level. Finally, we note that our study did not examine whether “opt out” may have affected the level (i.e., quality or effort) of supervision.

In conclusion, our article suggests that “opt out” did not significantly shorten travel distances for the procedures we examined. In combination with previous work suggesting that “opt out” did not increase the total number of surgeries performed,^40,41  our work suggests that “opt out” has not been effective in increasing two important dimensions of access to anesthesia care: number of procedures performed or decreased travel distances for surgery. In sum, policymakers should understand that “opt out” is unlikely to be a “silver bullet” when it comes to increasing access to anesthesia care. Understanding the reasons why “opt out” has not achieved its intended goal remains an area for further research.

Support was provided solely from institutional and/or departmental sources.

Dr. Sun received funding from a Mentored Research Training Grant from the Foundation for Anesthesia Education and Research, Schaumburg, Illinois. Dr. Miller is employed by the American Society of Anesthesiologists, Schaumburg, Illinois. The other authors declare no competing interests.