Share this article on:

Anesthesia Care Transitions and Risk of Postoperative Complications

Hyder, Joseph A. MD, PhD*†‡; Bohman, J. Kyle MD*; Kor, Daryl J. MD*‡; Subramanian, Arun MBBS*; Bittner, Edward A. MD, PhD§; Narr, Bradly J. MD; Cima, Robert R. MD, MA‡¶#; Montori, Victor M. MD, MSc**

doi: 10.1213/ANE.0000000000000692
Patient Safety: Research Report

BACKGROUND: A patient undergoing surgery may receive anesthesia care from several anesthesia providers. The safety of anesthesia care transitions has not been evaluated. Using unconditional and conditional multivariable logistic regression models, we tested whether the number of attending anesthesiologists involved in an operation was associated with postoperative complications.

METHODS: In a cohort of patients undergoing elective colorectal surgical in an academic tertiary care center with a stable anesthesia care team model participating in the American College of Surgeons National Surgical Quality Improvement Program, using unconditional and conditional multivariable logistic regression models, we tested adjusted associations between numbers of attending anesthesiologists and occurrence of death or a major complication (acute renal failure, bleeding that required a transfusion of 4 units or more of red blood cells within 72 hours after surgery, cardiac arrest requiring cardiopulmonary resuscitation, coma of 24 hours or longer, myocardial infarction, unplanned intubation, ventilator use for 48 hours or more, pneumonia, stroke, wound disruption, deep or organ-space surgical-site infection, superficial surgical-site infection, sepsis, septic shock, systemic inflammatory response syndrome).

RESULTS: We identified 927 patients who underwent elective colectomy of comparable surgical intensity. In all, 71 (7.7%) patients had major nonfatal complications or death. One anesthesiologist provided care for 530 (57%) patients, 2 anesthesiologists for 287 (31%), and 3 or more for 110 (12%). The number of attending anesthesiologists was associated with increased odds of postoperative complication (unadjusted odds ratio [OR] = 1.52, 95% confidence interval [CI] 1.18–1.96, P = 0.0013; adjusted OR = 1.44, 95% CI 1.09–1.91, P = 0.0106). In sensitivity analyses, occurrence of a complication was significantly associated with the number of in-room providers, defined as anesthesia residents and nurse anesthetists (adjusted OR = 1.39, 95% CI 1.01–1.92, P = 0.0446) and for all anesthesia providers (adjusted OR = 1.58, 95%CI 1.20–2.08, P = 0.0012). Findings persisted across multiple, alternative adjustments, sensitivity analyses, and conditional logistic regression with matching on operative duration.

CONCLUSIONS: In our study, care by additional attending anesthesiologists and in-room providers was independently associated with an increased odds of postoperative complications. These findings challenge the assumption that anesthesia transitions are care neutral and not contributory to surgical outcomes.

Published ahead of print March 19, 2015

From the *Division of Critical Care Medicine, Department of Anesthesiology, Mayo Clinic, Rochester, Minnesota; Center for Surgery and Public Health, Brigham and Women’s Hospital, Boston, Massachusetts; The Robert D. and Patricia E. Kern Center for the Science of Healthcare Delivery, Mayo Clinic, Rochester, Minnesota; §Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Boston, Massachusetts; Department of Anesthesiology, Mayo Clinic, Rochester, Minnesota; Department of Surgery, Mayo Clinic, Rochester, Minnesota; #The Center for the Science of Healthcare Delivery, Mayo Clinic, Rochester, Minnesota; and **Department of Medicine, Mayo Clinic, Rochester, Minnesota.

Accepted for publication December 29, 2014.

Published ahead of print March 19, 2015

Funding: None.

The authors declare no conflicts of interest.

Reprints will not be available from the authors.

Address correspondence to Joseph A. Hyder, MD, PhD, Department of Anesthesiology, Mayo Clinic, 200 1st St. SW, Rochester, MN 55905. Address e-mail to

The delivery of high-value care for hospital patients requires integrated teamwork including effective caregiver handoffs. The Joint Commission,1 the Institutes of Medicine,2 and the National Quality Forum have cited failed teamwork as a common cause of sentinel events and handoffs as the point at which “safety fails first.”3

Although a recent experimental study of medical patients linked the number of handoffs to a decrement in perceived quality of care, investigations of care transitions and outcomes are scarce.4 Thus, the evidence base for component of a handoff is underdeveloped.5,6 This is true for medical as well as surgical patients,7 even though >200 million patients undergo surgery annually.8 Surgical patients in the operating room may be particularly vulnerable to lapses in care because of their rapidly changing clinical needs, critical physiology (including the need for mechanical ventilation), and the involvement of multiple providers from surgery, anesthesia, and nursing during a single operation. This rationale recently has prompted more attention to, and examination of, perioperative handoffs.9–11

The vulnerability of the surgical patient has engendered 2 divergent strategies by surgeons and anesthesia providers. Surgeons have traditionally relied on the “ship’s captain” model of care, where 1 surgeon is responsible for a patient’s surgery. By contrast, anesthesiologists have embraced a systems-based, care team model, now the dominant model, by which several anesthesia providers participate during a patient’s surgery.12,13 This model is endorsed by the American Society of Anesthesiologists (ASA)13; however, the safety of care transitions in this anesthesia care model has received limited attention.

This investigation leverages quality improvement data from the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) and administrative data documenting the number of anesthesia providers involved in an operation to investigate the safety of care transitions in a team-based model of anesthesia care. Our primary aim was to estimate the extent to which handoffs by attending anesthesiologists would be associated with postoperative complications after we adjusted for surgical and patient factors. We additionally investigated transitions in care by in-room anesthesia providers.

Back to Top | Article Outline


Human Subjects Protections

For quality improvement purposes, Mayo Clinic, Rochester, MN, has maintained a database of surgical patients as part of institutional participation in the ACS-NSQIP.14 This study was approved, and the requirement for informed written consent was waived by the Institutional Review Board of Mayo Clinic.

Back to Top | Article Outline

Methodologic Assumptions and Approaches

Given a dearth of published data on the topic, we made a key epidemiologic assumption about care transitions as a risk factor based on previous clinical and statistical experience. We assumed that care transitions are more risky in some institutions than others and for some kinds of surgeries and patients compared with others. Thus, any potential association between care transitions and outcomes may be subject to effect modification, or interaction depending on practice setting and type of surgery. Effect modification would complicate a modeling approach for causal inference, and efforts to statistically adjust for diverse surgical types and/or institutions would likely be inadequate, thus increasing the chances of a spurious finding.15–17 In practical terms, any potential association between care transitions and complications in an observational study would be meaningful only in a specific context, because a global estimate of the average effect of handoffs across diverse settings would be too generalized to apply to any specific setting. Care transitions are a complex process that often are performed differently according to institutions and surgical type, with effects potentially varying by method of assessment, such as whether the assessment is interventional or observational.18–20

Back to Top | Article Outline

Practice Setting

The setting for this study is a single academic tertiary care center where the anesthesia practice environment has used a stable anesthesia care team model: anesthesia providers, including attending anesthesiologists and nurse anesthetists, historically have trained at the same institution, have long-standing working relationships, and practice standardized anesthetics in a care team model. At Mayo Clinic, there are no locum tenens anesthesiologists or locum tenens Certified Registered Nurse Anesthetists (CRNAs). Staff CRNAs and anesthesiologists typically work within one core area with minimal crossover between specialties. Of note, there was no standardized handoff protocol in place during the study period.

Back to Top | Article Outline

Selection of Patients

We used patient exclusion to create a homogeneous analytic stratum, or single-level stratification where conclusions would be more likely to be accurate and where any estimate of effect would have a target population for future intervention. The statistical implication of a homogenous sample is a discounted number of variables needed for risk adjustment and discounted discrimination ability of a model.21 We queried patients from the Mayo Clinic NSQIP data registry between April 26, 2006, and January 28, 2010. All patients underwent elective colorectal procedures for which there were greater than 10 instances of a given Current Procedural Terminology codes 44140, 44143, 44145, 44150, 44155, 44160, 44204, 44205, 44207, 44210, 44211, 44212, 45110, 45113, 45395, and 45397.22 All patients were of ASA physical status (ASAPS) classification less than V, and no cases were performed by student nurses under the direction of an instructor. All patients had received a general anesthetic under the care of resident physicians and/or CRNAs and attending anesthesiologists providing medical direction. No systematic changes in surgical practice or perioperative care occurred during the study period that would affect a potential association between care transitions and complications. All surgeons were board certified.

Back to Top | Article Outline


We identified and defined potential confounders of possible associations between care transitions and outcomes. Operative duration was initially defined with 1-hour increments. Estimated blood loss was defined a categorical variable using the classification system devised by Gawande et al.23 To account for case complexity, we additionally included work relative value units (RVUs) and classification of surgical extent modeled on the work of Cohen et al. and the American College of Surgeon Colorectal Surgical Risk Calculator.24–26 We also used data describing the attending surgeon experience, defined as having performed >40 colectomies in the sample or not, as well as a continuous variable for postgraduate year of the surgical assistant.

Additional potential model covariates were selected a priori based on a review of the literature, presumed explanatory ability, and clinical applicability.21,26 The surgical homogeneity of the sample permitted inclusion of fewer variables than may have been required for a more heterogenous population.27 The following ACS-NSQIP preoperative variables were coded as normal versus not: functional status before surgery, dyspnea, and congestive heart failure. ASAPS was included as a categorical variable. Preoperative health states and postoperative clinical data were classified using ACS-NSQIP definitions.28

Back to Top | Article Outline

Anesthesia Care Providers and Transitions

The a priori primary exposure of interest was the number of attending anesthesiologists caring for a patient between entry and exit from the operating room. The a priori secondary exposure of interest was number of in-room providers. Data describing the anesthesia care providers were available from administrative data. Each time a new anesthesia team member (attending anesthesiologist, resident, or CRNA) assumed care of a patient, it was documented electronically in the anesthesia information management system. In this study, the presence of any additional attending anesthesiologist, but not in-room provider, for a case was equivalent to a handoff in care. The following alternative exposures were selected for investigation before analyses: total number of care providers (number of anesthesiologists plus number of in-room providers) and total number of in-room care providers (consisting of anesthesiology residents and CRNAs).

In this study, CRNAs provide anesthesia care under medical direction of an attending anesthesiologist. In conjunction with the attending anesthesiologist, a CRNA or resident would participate in preoperative assessment and intubation. This person would remain with the patient during maintenance of anesthesia and notify the attending anesthesiologist of progress during the case. The attending anesthesiologist would participate in preoperative assessment and be present for any procedures, induction, and emergence from anesthesia and any critical event occurring during the surgical course. The variable “high-provider number” was defined a priori as having 3 or more in-room providers and 2 or more attending anesthesiologists. With this definition, a high-provider team would include at least 1 handoff among attending anesthesiologists and at least 2 handoffs among in-room providers.

Anesthesia care transitions in the study followed different patterns for attending anesthesiologists and in-room providers. Attending anesthesiologists arrive to work in operating rooms in the morning, assess their patients preoperatively, and participate with in-room providers during induction, procedures, and emergence from anesthesia. Attending anesthesiologists typically cover 2–3 operating rooms at a time. Because surgical work is accomplished during the day, an operating room may close, and the anesthesiologist covering the closing room will be available to take over another from another anesthesiologist. Through the use of a hierarchy set in advance, each anesthesiologist is assigned a position in this relief system. Decisions to transition care between attending anesthesiologists are primarily based on workflow in the operating suites and not time-of-day transitions. In rare instances when a patient is unstable during a case, handoffs would be specifically avoided. In other instances, attending anesthesiologists may briefly yield and subsequently reassume care during a case, such as to participate in a meeting, but this would be an extremely rare event on the whole. Thus, for attending anesthesiologists, the number of handoffs is equal or nearly equal to the number of anesthesiologists minus 1. A “night call” anesthesiologist will arrive around 6:00 PM but will take over day cases if, and only if, the number and acuity of emergency cases is limited.

For in-room providers, care transitions primarily are attributable to brief breaks in the morning, for lunch, and in the afternoon. Because the timing of breaks varies, an in-room provider may receive a morning break anytime between 8:30 AM and 11:30 AM, depending on workflow. Providers do not always receive a morning or afternoon break, but they generally do. Other than breaks, care transitions among in-room providers may rarely be as the results personal, idiosyncratic needs such as having a doctor’s appointment. Similarly, break patterns for dinner or relief of shift from a case in the evening do not follow any set pattern. In-room providers may work past 5:00 PM if a case is ongoing. The care transitions between in-room providers are not asymmetric (1-way), so the number of providers may underestimate the number of care transitions. With the data-capture method used here, potential misclassification of provider number is unlikely but, if present, would result in underestimation of the number of providers; the result would be misclassification of exposure more likely to bias findings to the null, or to underestimate any association between provider number and outcome.

Back to Top | Article Outline


The primary end point was major complication and/or death within 30 days of index surgery as recorded in the Mayo Clinic, Rochester, MN, ACS-NSQIP database. Data linking anesthesia therapies to 30-day postoperative events are limited, and surgical quality metrics such as National Quality Forum–endorsed performance measures 0534, 1543, 1540, 0706 rely on a black-box model, wherein quality is defined by occurrence of postoperative complications including death for up to 30 days.29 For this reason, and consistent with previous investigations,30 the following ACS-NSQIP-defined events were defined as major complications: acute renal failure, bleeding that required a transfusion of 4 units or more of red blood cells within 72 hours after surgery, cardiac arrest requiring cardiopulmonary resuscitation, coma of 24 hours or longer, myocardial infarction, unplanned intubation, ventilator use for 48 hours or more, pneumonia, stroke, wound disruption, deep or organ-space surgical-site infection, superficial surgical-site infection, sepsis, septic shock, and systemic inflammatory response syndrome. Deep venous thrombosis and pulmonary embolism were excluded from this suite because the link between these outcomes and anesthesia handoffs was deemed tenuous.31 These outcomes were, however, included in sensitivity analyses.

Back to Top | Article Outline


Descriptive Statistics

Differences in the distributions of clinical characteristics were tested with analysis of variance for continuous variables and χ2 or Fisher exact tests, as appropriate, for categorical variables. The association between surgical extent and complication was tested with the χ2 test as well.

Back to Top | Article Outline

Primary Hypothesis Testing

To test for adjusted associations between anesthesia providers and outcome, we evaluated a series of unadjusted and adjusted multivariable logistic regression models with addition of candidate covariates with alternative variable forms to inspect for effect size attenuation rather than whether the covariate was independently associated with major complication. We present saturated, nonparsimonious models. We also inspected unadjusted and adjusted associations between operative duration, estimated blood loss, and anesthesia providers to characterize potential confounding.

Multicollinearity of all covariates used in the same models was assessed by tolerance values, all of which exceeded 0.5, well above the prespecified cutpoint of 0.1 that would suggest incompatibility or distortion in models. Models were assessed by the Hosmer–Lemeshow goodness-of-fit test.32 All analyses were performed in SAS version 9.3 (Cary, NC).

Back to Top | Article Outline

Sensitivity Analyses and Evaluation of Robustness

We performed multiple sensitivity analyses to investigate findings including nonparsimonious models, alternative variable forms, serial exclusions, and serial inspection of associations of covariates with complications. The first type of sensitivity analyses focused on additional variables and variable forms describing surgical intensity. Nonparsimonious logistic regression models were created by forcing and/or exchanging additional variables into the models as well as by altering the variable structures for blood loss and operative duration, in an attempt to diminish associations between anesthesia providers and major complication with alternative adjustment. These alternatives included adjustment for surgical extent, surgeon experience, postgraduate year of the surgical assistant, and dyspnea as well as new variable structures for operative duration to a linear variable, an ordinal categorical variable with additional levels beyond blood loss of 500, 750, and 1000 mL, unordered categorical variable structure for blood loss as well as multiple, analogous alterations of the variable structures for operative duration. Unordered covariates were made ordered to test for trends. Final, saturated, nonparsimonious models included >70 events and 6 or 7 degrees of freedom.

We further tested associations between provider transitions and outcomes by using conditional logistic regression with m:n matching (PROC PHREG) on 30-minute duration of surgery. Because the number of providers is an integer and may exceed 2, and because operative duration is the strongest correlate of provider number, this method was chosen as analogous but superior to propensity score matching to investigate whether residual confounding by operative duration was responsible for a spurious association.

Additional sensitivity analyses used serial exclusions to examine whether other factors were driving statistically significant associations between anesthesia care transitions and complications. These exclusions included the following: cases lasting longer than 4 hours, cases that were contaminated or dirty cases, and cases that ended after 5:00 PM. In these analyses, as a positive control, we investigated whether greater ASAPS and more contaminated wound classification remained associated with postoperative complication in a statistically significant manner.33,34 Second, we examined the directions and statistical significances of estimated blood loss and operative duration with complication to assess for potential “overadjustment,” a potentially disruptive effect of adjustment wherein adjustment for 2 related variables results in 1 variable having a paradoxical association with the outcome. In this case, we examined our findings for spurious, statistically significant protective effects of high blood loss, and long duration of operation.

Back to Top | Article Outline

Post Hoc, Exploratory Analyses of Anesthesia Therapy, Provider Number, and Outcome

We conducted exploratory analyses related to transitions of care, postoperative complications, and intraoperative patient experience. These analyses aimed to assess associations between multiple variables and both the outcomes and the number of attending anesthesiologists. In these crude analyses, if a variable (such as phenylephrine dosing) was found to be associated with both the exposure (number of attending anesthesiologists) and the outcome, then it may deserve scrutiny as a potential mechanistic variable. To avoid inappropriate conclusions from limited data in these exploratory analyses, we selected the nonparametric Spearman correlation as a simple measure of association with few statistical assumptions. We evaluated associations with and without adjustment as specified. These post hoc, exploratory analyses require 24 hypothesis tests (12 variables tested against exposure and outcome), so we considered a Bonferroni-corrected P value of 0.0021 to be statistically significant. For these post hoc, exploratory analyses, we queried the anesthesia information management system to obtain the following variables: vital signs components of the surgical Apgar score using 10-minute nonoverlapping intervals,35 compliance with National Quality Forum–endorsed performance measure for normothermia within the operating room,36 total crystalloid and colloids administered intraoperatively, total neostigmine dose, total IV opioids administered assessed as oral morphine equivalents, total bolus dose of phenylephrine, total bolus dose of ephedrine, and total number of medication administrations in the operating room. Repeated dosing as a source of error when calculating oral morphine equivalents from IV opioids and neuromuscular blockade was not addressed because of the limitations of the data available.37,38 Given the post hoc nature of this analysis, we specified that a potential mechanistic variable should demonstrate association with both complications and with handoffs among attending anesthesiologists. These P values were not adjusted for multiple comparisons.

Back to Top | Article Outline


In all, the study included 927 patients undergoing elective colectomy in the institutional ACS-NSQIP sample to compose the analytic stratum (Table 1). Of these patients, 71 experienced a major complication, including 7 who died. In all, 105 complications occurred. The total complications in this study were dominated by infection-related events, including superficial surgical-site infection,26 pneumonia,12 sepsis14 and septic shock,10 deep surgical-site infection,10 and wound disruption.5 Complications not clearly related to infection such as reintubation,8 myocardial infarction,4 postoperative transfusion,4 acute renal failure,4 stroke,3 prolonged ventilation,3 coma,1 and cardiopulmonary resuscitation1 were less common. Only 1 patient experienced an intraoperative event—all other complications and deaths occurred postoperatively. The following variables were associated with postoperative complication in unadjusted analyses: wound classification (P < 0.01), ASAPS (P < 0.01), dyspnea (P < 0.01), and congestive heart failure (P < 0.03.) In this highly homogenous sample, a number of risk factors were not associated with postoperative complication. These variables included age, patient’s functional status, work RVU, surgical extent, surgeon experience, or postgraduate year of the surgical trainee.

Table 1

Table 1

Table 2

Table 2

We tested the distributions of covariates by the number of attending anesthesiologists (Table 2). Wound classification was significantly associated with the number of attending anesthesiologists (P < 0.01) as were work RVUs (P < 0.01) and surgical extent classified as partial colectomy versus other (P < 0.01). Both estimated blood loss (P < 0.01) and operative duration (P < 0.01) also were significantly associated with attending number. Composition of the anesthesia team also was associated with the number of attending anesthesiologists wherein greater numbers of attending anesthesiologists was associated with greater numbers of in-room providers (P < 0.01) and the presence of a high-provider anesthesia team (P < 0.01).

Back to Top | Article Outline

Assessment of Potential Confounding and Primary Hypothesis Testing

All measures of anesthesia provider numbers were significantly associated with postoperative complications, and greater estimated blood loss was associated (odds ratio [OR] = 1.48, 95% confidence interval [CI] 1.03–2.14, P = 0.0364) with postoperative complication (Table 3). Operative duration was not significantly associated with postoperative complications, however.

Table 3

Table 3

Table 4

Table 4

Adjusted associations between anesthesia care team and postoperative complications are presented in Table 4. All models were additionally adjusted for age, ASAPS, work RVU, wound classification, estimated blood loss, and operative duration. All descriptions of greater provider number, whether total number of anesthesia providers, number of attending anesthesiologists, or use of a high-provider team, demonstrated statistically significant associations with postoperative complication after adjustment. Estimated blood loss and surgical duration were not associated with postoperative complication after adjustment.

Back to Top | Article Outline

Sensitivity Analyses for All Providers

In a separate adjusted model (not shown) including both in-room providers and attending anesthesiologists, associations varied little from those shown with associations for in-room providers (OR = 1.31, 95% CI 0.95–1.82, P = 0.1006) and attending anesthesiologist (OR = 1.39, 95% CI 1.05–1.85, P = 0.0237). When attending anesthesiologist number and in-room provider number were collapsed into a single variable, additional anesthesia providers were associated with significantly increased odds of postoperative complication (OR = 1.58, 95% CI 1.20–2.08, P = 0.0012).

Back to Top | Article Outline

Sensitivity Analyses for Surgical Factors

We conducted additional sensitivity analyses for surgical variables of surgeon experience, surgical duration, and wound classification. In additional and/or alternative analyses, adjustment for experience of the surgeon and first assistant or surgical extent yielded associations that were not materially different from those presented. In additional sensitivity analyses using exclusion, we restricted the analyses only to the shorter cases, excluding cases longer than 4 hours (an exclusion of 25% of the sample), and the adjusted associations persisted (for attending anesthesiologists OR = 1.48, P = 0.0290). After excluding contaminated or dirty cases (i.e., exclusion based on wound classification), 860 patients remained, and attending handoffs remained significantly associated with complications even after adjustment (OR = 1.40, P = 0.0329). In addition, results were unchanged when deep venous thrombosis was included among complications. Finally, we applied m:n matching by 30-minute category of operative duration with conditional logistic regression to investigate whether adjustment for operative duration was adequate. Adjusted associations were not materially different with this method (attending anesthesiologist OR = 1.44, P = 0.0100; in-room provider OR = 1.44, P = 0.0260; high-provider team OR = 2.05, P = 0.0140).

Back to Top | Article Outline

Sensitivity Analyses for After-Hours Care

When analyses were restricted cases to ending before 5:00 PM (restriction based on time of day), the sample included 46 events and 745 patients, and attending anesthesiologist handoffs were significantly associated with complications before adjustment (OR = 1.88, P = 0.0005) and after adjustment (OR = 1.85, P = 0.0020) for age, ASA classification, wound classification, work RVU, operative duration, and estimated blood loss.

Back to Top | Article Outline

Post Hoc, Exploratory Analyses of Anesthesia Therapy, Provider Number, and Outcome

We tested crude and adjusted Spearman associations between multiple variables and both complications and number of attending anesthesiologists. These included the following 12 variables: compliance with intraoperative normothermia reporting, surgical Apgar score, the vital signs components of the surgical Apgar score, crystalloid volume infused, colloid volume infused, red cell transfusion volume, dose of neostigmine reversal, dose of ephedrine bolus, dose of phenylephrine bolus, use of phenylephrine infusion, oral morphine equivalents, and total number of medication administrations. Only intraoperative transfusion of red cells was significantly associated with both complications (correlation 0.10396, P = 0.0016) and number of attending anesthesiologists (correlation 0.13980, P < 0.0001). When Spearman correlations were adjusted for blood loss, operative duration, and wound classification, the associations between intraoperative red cell transfusion were altered (with complications, correlation 0.06331, P = 0.0549; with number of attending anesthesiologists, correlation 0.07960, P = 0.0157). None of these tests met the prespecified threshold of P < 0.0021 to be considered statistically significant.

Back to Top | Article Outline


We used an observational study design to test whether the number of attending anesthesiologists was associated with the odds of postoperative complications. We found consistently positive, independent associations between postoperative complications and attending anesthesiologists as well as the number of in-room providers (residents and nurse anesthetists) and care by high-provider-number teams. These associations were only minimally affected after extensive attempted adjustments for operative duration, blood loss, surgical extent, and other potential markers of surgical intensity and patient factors. Further sensitivity analyses accounting for after-hours care, very long cases, and conditional logistic regression with matching demonstrated the same significant associations. Two validation end points, the ASAPS and wound classification, which are both known risk factors for complications, additionally confirmed the reliability of the models.33 These findings provide empirical evidence that intraoperative care transitions may contribute to surgical outcomes used in quality measurement.

Features of the study design inform conclusions regarding the strength and generalizability of the findings. This observational study design does not specifically test a hypothesis about the comparative effectiveness of single versus multiple provider care models. These findings are intentionally restricted to elective colectomy patients at a single institution; thus, the study was not designed to be generalizable to all practice setting or case mixes. A “generalizable” magnitude of association between handoffs and complications is not likely to be consistent across anesthesia practices globally. However, the specific findings in this setting do challenge the conventional wisdom of daily practice that handoffs, as executed in real time, are care-neutral events. Thus, these findings may be relevant to any surgical practice where anesthesia care transitions occur.

These present findings are internally robust, and this restricted analytic approach may serve as an example for future observational (not interventional) studies aiming to investigate local practice effects. Although residual confounding can never be excluded entirely, the associations between care transitions and complications in this study do not appear to be the result of residual confounding from unmeasured heterogeneity among patients or unmeasured surgical factors. The selection of a highly homogeneous surgical sample, or analytic stratum, minimizes the potential effect of unmeasured predictors of complication. This is demonstrated by the limited predictive ability of otherwise predictive variables in this study. For example, there was no association between postoperative complication and age, patient’s functional status, work RVU, or surgical extent, among other variables, so these variables were not classic confounders in this sample (Table 1). If the adjusted models failed to include important patient or surgical factors, then one would expect the regression models to demonstrate poor predictive ability, because the model would not have accounted for other important, causative factors. On the contrary, the model presented here demonstrated good predictive ability with c-statistics of 0.691 (number of attendings), 0.688 (number of in-room providers), and 0.691 (high-provider team). These results compare reasonably well with the American College of Surgeons Colorectal Risk Calculator, a widely used tool using >20 variables with a reported c-statistic ranging from 0.681 to 0.728 for predicting mortality.26 Model calibration in the present study was reasonable. The comparable performance of the model created in the current study is likely a result of successful selection of a highly homogenous sample27,39 as well as delineation of the most explanatory patient factors associated with complications.21 If surgical factors were independently associated with complications in this sample, but not adequately included, then we would expect the models in the current study to have poor predictive ability. On the contrary, the models developed in this study had good predictive ability, despite the “discounting” of the c-statistic expected with a homogeneous sample. Moreover, statistically significant associations between handoffs and complications were virtually unchanged before and after the multiple adjustment strategies and alternative exclusions of longer cases, contaminated or dirty cases, and cases ending later in the day. Our final models were nonparsimonious, including additional variables, to demonstrate that no additional adjustment affected associations or levels of significance. Although residual confounding from unmeasured factors is always possible in observational studies, we found no evidence that residual confounding explained the significant associations presented.

The approaches using serial adjustment methods additionally benefitted from low misclassification rates due to the high fidelity reporting of covariate and outcome data in the ACS-NSQIP. The reported rate of major complications in the present study of elective colectomies is lower than has been reported for other series of colectomies, but these series also included emergent colectomies.30,40 There are no large, readily available multi-institution or national data sets currently that possess the necessary variables and outcome fidelity to test the hypothesis under investigation.41 Moreover, results from such studies, if possible, would be of limited relevance because of their heterogeneous samples and care settings.

Surgery is common worldwide, with >200 million surgical procedures occurring annually.8 In the United States alone, of the 34.8 million patients discharged from United States hospitals each year, 46% had at least 1 surgical procedure.42 Expenditures for surgical procedures approach $400 billion annually, and each major surgical complication is estimated to cost an average of $11,500.43,44 Surgical quality is increasingly defined by the occurrence of complications or death postoperatively.29 Despite this, there is only one published study investigating associations between anesthesia care coordination and complications, which used different methods but yielded similar strength, direction and magnitude of association.45 A review of the literature found 2 studies linking patient handoffs and any postoperative quality variance. One analysis of 60 surgical malpractice claims suggested that communication breakdowns may contribute to surgical complications.46 A more recent study of 360 handoffs among surgical patients suggested that poorer-quality handoffs were associated with longer stays in the postanesthesia care unit.7 Other studies of handoffs including nonsurgical patients suggest this aspect of care may be a critical target for process improvement.47,48

The associations presented are not assumed to be identical in other practice settings and may, in fact, be an underestimate of magnitudes of association in some other settings. Results in the present study are derived from a single institution where the anesthesia practice model is mature, handoffs are part of the culture of care and not reflective of patient decompensation in the operating room, and the anesthesia personnel have been historically sheltered from health care personnel disruption. If we assume that stable, established care teams transitioning care for elective surgery patients are more likely to be safe than transitions among unfamiliar personnel caring for emergency surgery patients, then the associations between handoffs and complications in this study could be an underestimate of an association in other settings.

Although the magnitude of association between handoffs and complications is not likely to be consistent across anesthesia practices globally, these specific findings do challenge the conventional wisdom of daily practice that handoffs, as executed in real time, are care-neutral events. Thus, these findings may be relevant to any surgical practice where anesthesia care transitions occur. This observational study design does not specifically test a hypothesis about the comparative effectiveness of single versus multiple provider care models.

Although associations between anesthesia provider number and postoperative complications were independent, consistent, and robust, the mechanisms for this effect are unclear. This association may be related to knowledge loss at handoff or patient exposure to provider-specific variation. No previous studies have investigated knowledge loss with operative handoffs or provider-specific quality variations in anesthetic care within the same case.49 One potential source of knowledge loss contributing to postoperative complication involves perioperative antibiotics, which are among the endorsed performance measures for surgical patients.29 Although initiation of antibiotic prophylaxis is subject to minimal practice variation,50,51 timely redosing of antibiotics could be subject to variation when multiple providers share care for a patient, thus increasing risk of postoperative infection. This is unlikely to be the mechanism the present study given that associations persisted after exclusion of cases lasting >4 hours when redosing would occur.

Handoffs may expose patients to provider-specific variations in other therapies, even during brief and temporary care transitions in the operating room, particularly with IV fluids and opioid therapy. For both of these therapies, clinical trials have linked therapy variation to postoperative outcomes for patients who undergo colectomy.52,53 Alternative explanations include increased risk of infection attributable to unsterile medication administration in which IV catheters are used or anesthesia providers acting as sources of fomite.54,55 These explanations, however, are unlikely to account for the associations in full. In-room providers rather than supervising anesthesiologists would typically administer medications intraoperatively, and the association between care transitions and complications was present for both groups. This study lacks data describing fomite introduction, but data linking fomites to postoperative infection are not determinative, and postoperative infection after bowel surgery is likely a common end point for diverse anatomic and physiologic disturbances.56

We undertook exploratory analyses of patient’s intraoperative experience, provider numbers, and complications. These analyses validated the association between care transitions and complications but demonstrated no compelling correlations linking provider numbers, complications, and a number of variables describing patient’s operative experience. These variables included hemodynamic stability, fluid therapies, opioid therapies, muscle relaxation, normothermia, or the number of medication administration episodes as a proxy for possible iatrogenic bacteremia. These secondary analyses were exploratory and are not definitive.

Future work in anesthesia care handoffs may use prospective, single-center observational studies to investigate which kinds of surgeries are most susceptible to anesthesia care transitions and evaluate potential mechanisms of care lapses. Future work on anesthesia practice variation attributable to provider transitions may reveal how anesthesia care transitions can affect surgical outcomes and what components would make up an anesthesia handoff checklist.

This investigation demonstrated significant, independent associations between more anesthesia providers and more postoperative complications after elective colorectal surgery, although a mechanism is not demonstrated. Given that transitions of care in anesthesia are commonplace, this work merits independent validation and future investigations that evaluate the essential components of care transitions, how anesthesia care transitions can become care neutral or potentially improve patient care, and assess the impact of intraoperative handoff checklists on patient outcome. Future observation studies may focus on local features of care transitions and case-mix specific associations.

Back to Top | Article Outline


Name: Joseph A. Hyder, MD, PhD.

Contribution: This author conceived the study idea, organized the study, helped analyze the data, and helped prepare the manuscript.

Attestation: Joseph A. Hyder has reviewed and approved the final manuscript. Joseph A. Hyder attests to the integrity of the original data and analysis reported in this manuscript. Joseph A. Hyder is the archival author.

Name: J. Kyle Bohman, MD.

Contribution: This author helped design the study, interpret data, and prepare the manuscript.

Attestation: J. Kyle Bohman has reviewed and approved the final manuscript. J. Kyle Bohman attests to the integrity of the original data and analysis reported in this manuscript.

Name: Daryl J. Kor, MD.

Contribution: This author helped design the study, interpret data, and prepare the manuscript.

Attestation: Daryl J. Kor has reviewed and approved the final manuscript.

Name: Arun Subramanian, MBBS.

Contribution: This author helped design the study, interpret data, and prepare the manuscript.

Attestation: Arun Subramanian has reviewed and approved the final manuscript.

Name: Edward A. Bittner, MD, PhD.

Contribution: This author helped design the study, interpret data, and prepare the manuscript.

Attestation: Edward A. Bittner has reviewed and approved the final manuscript.

Name: Bradly J. Narr, MD.

Contribution: This author helped design the study, interpret data, and prepare the manuscript.

Attestation: Bradly J. Narr has reviewed and approved the final manuscript.

Name: Robert R. Cima, MD, MA.

Contribution: This author helped design the study, interpret data, and prepare the manuscript.

Attestation: Robert R. Cima has reviewed and approved the final manuscript.

Name: Victor M. Montori, MD, MSc.

Contribution: This author helped design the study, interpret data, and prepare the manuscript.

Attestation: Victor M. Montori has reviewed and approved the final manuscript.

This manuscript was handled by: Sorin J. Brull, MD, FCARCSI.

Back to Top | Article Outline


The authors wish to acknowledge Roxanne Hyke, RN, Sharon Nehring, RN, Gregory Wilson, RRT, and Melissa Passe, RRT for their support with data management at Mayo Clinic, Rochester, MN.

Back to Top | Article Outline


1. Sentinel Event Data: Root Causes by Event Type 2004–2012. 2012 Washington, DC The Joint Commission
2. Crossing the Quality Chasm: A New Health System for the 21st Century. 2001 Washington, DC The National Academies Press
3. National Quality Forum (NQF), Preferred Practices and Performance Measures for Measuring and Reporting Care Coordination: A Consensus Report. 2010 Washington, DC NQF
4. Desai SV, Feldman L, Brown L, Dezube R, Yeh HC, Punjabi N, Afshar K, Grunwald MR, Harrington C, Naik R, Cofrancesco J Jr. Effect of the 2011 vs 2003 duty hour regulation-compliant models on sleep duration, trainee education, and continuity of patient care among internal medicine house staff: a randomized trial. JAMA Intern Med. 2013;173:649–55
5. Riesenberg LA, Leitzsch J, Little BW. Systematic review of handoff mnemonics literature. Am J Med Qual. 2009;24:196–204
6. Riesenberg LA, Leitzsch J, Massucci JL, Jaeger J, Rosenfeld JC, Patow C, Padmore JS, Karpovich KP. Residents’ and attending physicians’ handoffs: a systematic review of the literature. Acad Med. 2009;84:1775–87
7. Bittner EA, George E, Eikermann M, Schmidt U. Evaluation of the association between quality of handover and length of stay in the post anaesthesia care unit: a pilot study. Anaesthesia. 2012;67:548–9
8. Weiser TG, Regenbogen SE, Thompson KD, Haynes AB, Lipsitz SR, Berry WR, Gawande AA. An estimation of the global volume of surgery: a modelling strategy based on available data. Lancet. 2008;372:139–44
9. Jayaswal S, Berry L, Leopold R, Hart SR, Scuderi-Porter H, Digiovanni N, Phillips A. Evaluating safety of handoffs between anesthesia care providers. Ochsner J. 2011;11:99–101
10. Tan JA, Helsten D. Intraoperative handoffs. Int Anesthesiol Clin. 2013;51:31–42
11. Boat AC, Spaeth JP. Handoff checklists improve the reliability of patient handoffs in the operating room and postanesthesia care unit. Paediatr Anaesth. 2013;23:647–54
    12. Meeusen V, van Zundert A, Hoekman J, Kumar C, Rawal N, Knape H. Composition of the anaesthesia team: a European survey. Eur J Anaesthesiol. 2010;27:773–9
    13. ASA House of Delegates. Statement on the Anesthesia Care Team. 2009 Washington, DC American Society of Anesthesiologists
    14. Fink AS, Campbell DA Jr, Mentzer RM Jr, Henderson WG, Daley J, Bannister J, Hur K, Khuri SF. The National Surgical Quality Improvement Program in non-veterans administration hospitals: initial demonstration of feasibility. Ann Surg. 2002;236:344–53
    15. Bosco JL, Silliman RA, Thwin SS, Geiger AM, Buist DS, Prout MN, Yood MU, Haque R, Wei F, Lash TL. A most stubborn bias: no adjustment method fully resolves confounding by indication in observational studies. J Clin Epidemiol. 2010;63:64–74
    16. Stürmer T, Joshi M, Glynn RJ, Avorn J, Rothman KJ, Schneeweiss S. A review of the application of propensity score methods yielded increasing use, advantages in specific settings, but not substantially different estimates compared with conventional multivariable methods. J Clin Epidemiol. 2006;59:437–47
      17. Grosse-Sundrup M, Henneman JP, Sandberg WS, Bateman BT, Uribe JV, Nguyen NT, Ehrenfeld JM, Martinez EA, Kurth T, Eikermann M. Intermediate acting non-depolarizing neuromuscular blocking agents and risk of postoperative respiratory complications: prospective propensity score matched cohort study. BMJ. 2012;345:e6329
        18. Haynes AB, Weiser TG, Berry WR, Lipsitz SR, Breizat AH, Dellinger EP, Herbosa T, Joseph S, Kibatala PL, Lapitan MC, Merry AF, Moorthy K, Reznick RK, Taylor B, Gawande AASafe Surgery Saves Lives Study Group. . A surgical safety checklist to reduce morbidity and mortality in a global population. N Engl J Med. 2009;360:491–9
        19. Urbach DR, Govindarajan A, Saskin R, Wilton AS, Baxter NN. Introduction of surgical safety checklists in Ontario, Canada. N Engl J Med. 2014;370:1029–38
        20. Gawande AA. When checklists work and when they don’t. The Incidental Economist. 2014
          21. Dimick JB, Osborne NH, Hall BL, Ko CY, Birkmeyer JD. Risk adjustment for comparing hospital quality with surgery: how many variables are needed? J Am Coll Surg. 2010;210:503–8
          22. American Medical Association. CPT – Current Procedural Terminology.
          23. Gawande AA, Kwaan MR, Regenbogen SE, Lipsitz SA, Zinner MJ. An Apgar score for surgery. J Am Coll Surg. 2007;204:201–8
          24. Cohen ME, Dimick JB, Bilimoria KY, Ko CY, Richards K, Hall BL. Risk adjustment in the American College of Surgeons National Surgical Quality Improvement Program: a comparison of logistic versus hierarchical modeling. J Am Coll Surg. 2009;209:687–93
          25. Dummit LA The Basics: Relative Value Units. 2009 Washington, DC National Health Policy Forum
            26. Cohen ME, Bilimoria KY, Ko CY, Hall BL. Development of an American College of Surgeons National Surgery Quality Improvement Program: morbidity and mortality risk calculator for colorectal surgery. J Am Coll Surg. 2009;208:1009–16
            27. Kwok AC, Lipsitz SR, Bader AM, Gawande AA. Are targeted preoperative risk prediction tools more powerful? A test of models for emergency colon surgery in the very elderly. J Am Coll Surg. 2011;213:220–5
            28. Khuri SF, Daley J, Henderson W, Hur K, Gibbs JO, Barbour G, Demakis J, Irvin G 3rd, Stremple JF, Grover F, McDonald G, Passaro E Jr, Fabri PJ, Spencer J, Hammermeister K, Aust JB. Risk adjustment of the postoperative mortality rate for the comparative assessment of the quality of surgical care: results of the National Veterans Affairs Surgical Risk Study. J Am Coll Surg. 1997;185:315–27
            29. NQF Endorses Surgical Measures. National Quality Forum. 2012 Available at: Accessed February 19, 2015
            30. Khuri SF, Daley J, Henderson W, Barbour G, Lowry P, Irvin G, Gibbs J, Grover F, Hammermeister K, Stremple JF. The National Veterans Administration Surgical Risk Study: risk adjustment for the comparative assessment of the quality of surgical care. J Am Coll Surg. 1995;180:519–31
            31. Regenbogen SE, Ehrenfeld JM, Lipsitz SR, Greenberg CC, Hutter MM, Gawande AA. Utility of the surgical apgar score: validation in 4119 patients. Arch Surg. 2009;144:30–6
            32. Hosmer DW, Lemeshow S Applied Logistic Regression. 20002nd ed. New York Wiley
            33. Prasad V, Jena AB. Prespecified falsification end points: can they validate true observational associations? JAMA. 2013;309:241–2
            34. Campbell DA Jr, Henderson WG, Englesbe MJ, Hall BL, O’Reilly M, Bratzler D, Dellinger EP, Neumayer L, Bass BL, Hutter MM, Schwartz J, Ko C, Itani K, Steinberg SM, Siperstein A, Sawyer RG, Turner DJ, Khuri SF. Surgical site infection prevention: the importance of operative duration and blood transfusion–results of the first American College of Surgeons-National Surgical Quality Improvement Program Best Practices Initiative. J Am Coll Surg. 2008;207:810–20
            35. Hyder JA, Kor DJ, Cima RR, Subramanian A. How to improve the performance of intraoperative risk models: an example with vital signs using the surgical apgar score. Anesth Analg. 2013;117:1338–46
            36. National Quality Forum. Surgery Patients with Perioperative Temperature Management. 2008 Available at: Accessed February 29, 2015
            37. Opioids -- Equianalgesic Dosages. Available at:
            38. Butterworth JF MD, Wasnick JD Morgan and Mikhail’s Clinical Anesthesiology. 20135th ed. Stamford, CT Appleton and Lange
            39. Merkow RP, Hall BL, Cohen ME, Dimick JB, Wang E, Chow WB, Ko CY, Bilimoria KY. Relevance of the c-statistic when evaluating risk-adjustment models in surgery. J Am Coll Surg. 2012;214:822–30
            40. Regenbogen SE, Bordeianou L, Hutter MM, Gawande AA. The intraoperative Surgical Apgar Score predicts postdischarge complications after colon and rectal resection. Surgery. 2010;148:559–66
            41. Gordon HS, Johnson ML, Wray NP, Petersen NJ, Henderson WG, Khuri SF, Geraci JM. Mortality after noncardiac surgery: prediction from administrative versus clinical data. Med Care. 2005;43:159–67
            42. DeFrances CJ, Lucas CA, Buie VC, Golosinskiy A. 2006 National Hospital Discharge Survey. Natl Health Stat Report. 2008;5:1–20
            43. Eappen S, Lane BH, Rosenberg B, Lipsitz SA, Sadoff D, Matheson D, Berry WR, Lester M, Gawande AA. Relationship between occurrence of surgical complications and hospital finances. JAMA. 2013;309:1599–606
            44. Dimick JB, Chen SL, Taheri PA, Henderson WG, Khuri SF, Campbell DA Jr. Hospital costs associated with surgical complications: a report from the private-sector National Surgical Quality Improvement Program. J Am Coll Surg. 2004;199:531–7
            45. Saager L, Hesler BD, You J, Turan A, Mascha EJ, Sessler DI, Kurz A. Intraoperative transitions of anesthesia care and postoperative adverse outcomes. Anesthesiology. 2014;121:695–706
            46. Greenberg CC, Regenbogen SE, Studdert DM, Lipsitz SR, Rogers SO, Zinner MJ, Gawande AA. Patterns of communication breakdowns resulting in injury to surgical patients. J Am Coll Surg. 2007;204:533–40
            47. Horn J, Bell MD, Moss E. Handover of responsibility for the anaesthetised patient—opinion and practice. Anaesthesia. 2004;59:658–63
            48. Pezzolesi C, Schifano F, Pickles J, Randell W, Hussain Z, Muir H, Dhillon S. Clinical handover incident reporting in one UK general hospital. Int J Qual Health Care. 2010;22:396–401
            49. Slogoff S, Keats AS. Does perioperative myocardial ischemia lead to postoperative myocardial infarction? Anesthesiology. 1985;62:107–14
            50. Bratzler DW, Ma A, Nsa W. Surgical care improvement project adherence and postoperative infections. JAMA. 2010;304:1670
            51. Stulberg JJ, Delaney CP, Neuhauser DV, Aron DC, Fu P, Koroukian SM. Adherence to surgical care improvement project measures and the association with postoperative infections. JAMA. 2010;303:2479–85
            52. Brandstrup B, Tønnesen H, Beier-Holgersen R, Hjortsø E, Ørding H, Lindorff-Larsen K, Rasmussen MS, Lanng C, Wallin L, Iversen LH, Gramkow CS, Okholm M, Blemmer T, Svendsen PE, Rottensten HH, Thage B, Riis J, Jeppesen IS, Teilum D, Christensen AM, Graungaard B, Pott FDanish Study Group on Perioperative Fluid Therapy. . Effects of intravenous fluid restriction on postoperative complications: comparison of two perioperative fluid regimens: a randomized assessor-blinded multicenter trial. Ann Surg. 2003;238:641–8
            53. Marret E, Remy C, Bonnet FPostoperative Pain Forum Group. . Meta-analysis of epidural analgesia versus parenteral opioid analgesia after colorectal surgery. Br J Surg. 2007;94:665–73
            54. Krueger CA, Murray CK, Mende K, Guymon CH, Gerlinger TL. The bacterial contamination of surgical scrubs. Am J Orthop (Belle Mead NJ). 2012;41:E69–73
              55. Ritter MA. Operating room environment. Clin Orthop Relat Res. 1999:103–9
              56. Ibrahimi OA, Sharon V, Eisen DB. Surgical-site infections and routes of bacterial transfer: which ones are most plausible? Dermatol Surg. 2011;37:1709–20
              © 2016 International Anesthesia Research Society