Long-Term Results of Heart Operations Performed by Surgeons-in-Training
Background— We investigated the association between trainees performing supervised operations and late outcomes of patients undergoing cardiac surgery.
Methods and Results— Data were prospectively collected on patients who underwent coronary artery bypass graft surgery, aortic valve replacement, or a combination of these between 1998 and 2005 at the Maritime Heart Center, Halifax, Canada. In-hospital mortality and a composite outcome of in-hospital mortality, stroke, bleeding, intra-aortic balloon pump insertion, renal failure, and sternal infection was compared between teaching (n=1054) and nonteaching cases (n=5877). Late survival and cardiovascular hospital readmissions were also examined. To adjust for baseline risk disparities, we used logistic regression for dichotomous in-hospital outcomes and Cox proportional hazards regression for survival data. Resident cases were significantly more likely to have high-risk features such as depressed ventricular function, redo operation, and urgent or emergent procedure. Resident as primary operator was not independently associated with in-hospital mortality (OR, 1.09; 95% CI, 0.75 to 1.58; P=0.66) or with the composite outcome (OR, 1.01; 95%, CI 0.82 to 1.26; P=0.90). The Kaplan-Meier event-free survival of the 2 groups was equivalent at 1, 3, and 5 years (log-rank P=0.06). By Cox regression, resident cases were not associated with late death or cardiovascular rehospitalization (hazard ratio, 1.05; 95% CI, 0.94 to 1.17; P=0.42).
Conclusions— Cases performed by senior-level cardiac surgery residents were more likely to have greater acuity and complexity than staff surgeon-performed cases. However, clinical outcomes were similar in the short- and long-term. Allowing residents to perform cardiac surgery is not associated with adverse patient outcomes.
Teaching young surgeons how to perform heart operations can be a balancing act. The literature is sparse, but it seems that in our profession, this apprenticeship has taken place over entire decades within tight quality control. In recent years, a number of factors conspire in potentially diluting the overall experience available to trainees. Patients coming to cardiac surgery have a higher risk profile and many programs throughout the world have to deal with legislation implementing a reduction in working hours.1,2 At the same time, institutions report that training surgeons in this field remains possible and safe, whether they perform coronary artery bypass graft surgery (CABG), aortic valve replacement (AVR),3,4 or operations that are technically more challenging such as mitral valve surgery5 or off-pump revascularization.6,7
One component that is missing in previously published reports is an analysis of long-term outcomes. If residents’ operating does not seem to affect hospital morbidity and mortality, does this also apply to the long-term success of the operation? Despite the best supervision, which may largely be responsible for good early results, it is conceivable that a few less-than-perfect stitches in a CABG surgery can reduce graft patency over the long term, which may in turn result in suboptimal clinical results. We therefore sought to examine the outcomes of resident-performed cases compared with staff cases in terms of event-free long-term survival.
Patients and Methods
Consecutive cases performed by residents and staff surgeons at the Maritime Heart Center between January 1, 1998, and December 31, 2005, were reviewed retrospectively. All patient data were collected prospectively in the institutional database, and audits reveal that less than 5% of data points are missing.5 For the purpose of this study, we excluded a minority of patients who were out of province referrals and did not have subsequent links to the provincial administration databases. The training system and how the surgical teaching takes place at this institution have been described previously.3,5 Briefly, in the Canadian system, cardiac training is a 6-year, full-fledged stand alone program to which medical students can match directly after medical school without prior general surgical training. In our study, resident operators were senior-level cardiac surgery trainees in postgraduate years 3 to 6. Because of the presence of a high trainers-to-trainees ratio, a senior-level resident has the privilege of selecting the cases with which he or she is to be involved. A resident case is defined as a procedure performed from “skin to skin” by the trainee with the staff surgeon acting as the first assistant or supervising while another person directly assists the resident. This means the resident performed all technical aspects of the operation (cannulation for bypass, vascular anastomoses, valve implantation, and so on) as well as, and importantly, being involved in the constant intraoperative dialogue that takes place with the anesthetist and the perfusionist to determine the conduct of the operation (myocardial protection, weaning from bypass, and so on). If a senior-level resident was scrubbed in a case but did not perform it as primary operator for whatever reason and, instead, assisted a staff surgeon, this case was labeled a staff surgeon case.
Resident cases were identified using case logs compiled by the residents themselves. In this period, there were 6 residents; for the purposes of this study, they were treated as a single surgeon. During the study period, the resident trainees were involved in a variety of operations. However, to minimize potential heterogeneity among cases and to allow for more meaningful comparisons, we included in the study only those adult patients undergoing isolated CABG, isolated AVR, or combined CABG and AVR either as a primary or redo procedure. The incidence of an in-hospital composite outcome, including death, stroke, reoperation for bleeding, intra-aortic balloon pump inserted during the operation or postoperatively, new postoperative renal failure, and deep sternal wound infection was compared between cases that were performed by residents and those performed by staff surgeons. The rates of the individual components of the composite were also examined among resident- and staff-performed cases. Separate logistic regression models of both in-hospital mortality and the composite outcome as the dependent variables were fit in which all variables with a univariate association significant at a probability value <0.10 or that were felt to be clinically relevant were included in the model. A variable selection algorithm was not used. Effect estimates were described using ORs and their associated 95% CIs. Each model’s discriminatory capacity and calibration were assessed by computing the area under the receiver operating characteristic curve (C statistic) and by using the Hosmer-Lemeshow goodness-of-fit test, respectively.8,9
The association between resident trainee versus staff as operator and longitudinal clinical outcomes postcardiac surgery was also examined and constituted the main focus of this study. Patients that were discharged alive after CABG and/or AVR were linked to government administrative hospitalization and mortality databases. The primary clinical end point of interest in this study was the freedom from all-cause mortality and a first hospital readmission for cardiovascular ischemic events (unstable angina, myocardial infarction, transient ischemic attack, or stroke), prosthetic valve-related complications (recurrent stenosis/insufficiency or endocarditis), dysrhythmias (atrial or ventricular fibrillation or flutter, ventricular tachycardia, or cardiac arrest), congestive heart failure, and cardiac procedures (valve repair/replacement, percutaneous or surgical revascularization). Survival times were analyzed using the Kaplan-Meier technique and multivariate analysis was performed using Cox proportional hazards regression to adjust for baselines disparities in risk.10 Variables that were significantly associated with the primary outcome on univariate analysis at a probability value <0.10 or were clinically relevant were included in the model. Effect estimates were described using hazard ratios and their associated 95% CIs. The proportionality assumption of the Cox regression model was assessed graphically and with the use of Schoenfeld residuals.11,12 Risk-adjusted event-free survival curves were generated using the corrected group prognosis method.13 With respect to the longitudinal outcomes analysis, the study had a statistical power of 82% (beta error=0.18) to detect a 30% increase in the relative risk of the composite outcome of late death or cardiovascular rehospitalization for those patients who had a resident versus staff surgeon as the primary operator.
Statistical analysis was performed using Statistical Analysis System software version 8.2 (SAS Institute, Cary, NC). Continuous variables were compared using unpaired t tests or nonparametric equivalents where appropriate. Proportions were compared using χ2 tests or Fisher exact test where appropriate and Mantel-Haenszel χ2 statistics with one degree of freedom were used to test for trends. All probability values were 2-sided. The study was in compliance with our local institutional research ethics board and received full approval.
Statement of Responsibility
The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.
Between January 1998 and December 2005, 9510 patients were operated on at our institution, of which 1648 were resident cases (17.3%). There were 7 staff surgeons for the entire time period and 6 residents over the 7-year period. The breakdown of training cases was (percent of total cases staff and resident combined): CABG only: 947 (14.1%); AVR±CABG: 248 (21.9%); mitral valve repair/replacement±CABG: 142 (27.8%); aortic root/arch/descending or thoracoabdominal pathology: 91 (31.1%); transplant: 17 (25.4%); and other: 203 (25.0%).
For this analysis, we included only CABG and/or AVR of patients residing in the province of Nova Scotia (to allow for linkage to administrative data). Therefore, the study population included 6931 patients, of which residents performed 1054 and staff surgeons performed 5877 operations. Patient demographics at baseline were similar between the 2 groups with respect to many prognostically important variables such as age, diabetes, and extracardiac vasculopathy. However, residents operated on patients with a significantly higher proportion of left ventricular dysfunction, clinical congestive heart failure, preoperative atrial fibrillation, preoperative intra-aortic balloon pump, previous cardiac surgery, and anemia (all P<0.05; Table 1). In addition, resident cases were significantly more likely to be urgent or emergent operations and combined procedures (Table 2). The durations of cardiopulmonary bypass and aortic clamping were on average 10 minutes longer for resident-performed isolated CABG cases, but operative durations were not significantly different between the 2 groups for isolated AVR and CABG/AVR cases (Table 2). The observed unadjusted in-hospital mortality was higher among resident cases; however, this difference was not statistically significant (3.98% versus 2.91%, P=0.06). In addition, the incidence of the composite end point did not significantly differ between the 2 groups (Table 2).
In multivariate analysis using logistic regression modeling, resident as operator was not independently associated with in-hospital mortality or the composite outcome of death, stroke, reoperation for bleeding, intra- or postoperative insertion of an intra-aortic balloon pump, renal failure, or deep sternal wound infection (Table 3). Similarly, resident was not associated with in-hospital mortality or the composite outcome when strata of procedure type was examined separately. All models had good discriminatory capacity and were a good fit to the data (Table 3).
Late outcomes were also examined. Of 6931 patients undergoing CABG and/or AVR at our institution, 6718 patients were discharged alive (96.9%). Among these, 4 patients did not have a valid health card number and we were unable to link them to the healthcare utilization databases. The remaining 6714 patients were successfully linked to the administrative hospitalization and mortality data sets and were included in the longitudinal analyses (5703 staff surgeon-performed cases and 1011 resident-performed cases). Median follow-up to event or censor was 2.7 years (interquartile range 1.0 to 4.7 years, maximum 8.2 years). There were 2522 late deaths or rehospitalizations for cardiovascular events or procedures during the study period (37.6%). The Kaplan-Meier event-free survival of staff- versus resident-performed cases was equivalent at 1, 3, and 5 years: 81.3% versus 79.1%, 68.2% versus 66.7%, and 58.6% versus 55.8%, respectively (log rank P=0.06). In a fully risk-adjusted time-to-first event multivariate Cox proportional hazards model, resident as operator was not associated with long-term death or hospital readmission for cardiovascular reasons (adjusted hazard ratio, 1.05; 95% CI, 0.94 to 1.17; P=0.42). The variables that emerged as independent predictors of late death and rehospitalization are shown in Table 4. Crude and adjusted survival curves are depicted in the Figure. Assessment of the Cox model using tests of proportionality and Schoenfeld residuals revealed that the effects of the predictor variables were the same at all values of time (constant hazard ratios over time) suggesting that the proportionality assumption of the model was not violated.
A separate Cox model was fit for patients undergoing AVR (with or without concomitant CABG). A similar association between resident case and outcomes was observed (hazard ratio, 1.10; 95% CI, 0.86 to 1.40; P=0.46).
Previously published reports have shown that allowing residents to perform cardiac surgical operations under staff supervision does not adversely affect in-hospital patient outcomes.2–5,7 Our data support these findings using a much larger series of patients. Furthermore, we have found that these results are maintained over a median follow-up of 3 years with a maximum follow-up of 8 years. These results suggest that resident training does not adversely affect morbidity and mortality in the intermediate term after cardiac surgery.
The absolute differences in the pre- and intraoperative characteristics of the 2 groups studied here are small. Overall, resident cases were sicker and more complex patients. This was expected because the residents choose which cases they want to scrub for and tend to pick those that are more complex. This, we believe, is beneficial in terms of residency training and exposure to difficult decision-making. It appears that the clustering of risk factors in the 2 groups has such an effect that short- and medium-term outcomes remain comparable. For example, prolonged ventilation in the resident group in unadjusted analysis may be associated with a higher rate of preoperative intra-aortic balloon pump use. A more detailed analysis of observed versus expected outcomes was done previously3 and is not duplicated here.
The main focus was what happens to patients after discharge from our service. Of the many independent risk factors influencing medium-term outcome in this report, all seem to have a small hazard ratio, suggesting small individual contributions. The usual predictors of poorer long-term outcome are confirmed in this analysis (age, comorbidities and acuity of procedure; Table 4).
Training and quality assurance go hand in hand and by examining the few existing reports, it can be seen how crude and adjusted results, as well as the tools used to evaluate them, are evolving. In reference to the Bristol report on off-pump training, Blackstone describes the sophisticated analyses (and their caveats) to which these cases are amenable.7,14 Lim and Tsui showed how adopting a team approach to training ensures a continued good level of exposure despite an absolute reduction in hours worked.2 In an interesting but not unexpected finding in their analysis, the main factors influencing whether an operation became a teaching case were the consultant in charge and the complexity of the procedure. Although few in number, these publications show collectively that the subtle chemistry of training can be broken down into smaller components and examined in analytic fashion. Of utmost importance, patient safety is preserved in all these positive reports. Intuitively, this must be the case in the overwhelming majority of programs. Extrapolating from the level of an individual institution to the wider world is conceptually difficult but not impossible. Haan and colleagues examined the STS database and drew a comparison between residency and nonresidency programs. Without teasing out further which cases were done by residents, training programs had longer crossclamp and perfusion times but short-term patient outcomes did not appear to be affected.4
What cannot be described in simple, retrospective reports such as ours is how the team, containing trainees in this instance, adapts to change for an optimal outcome. In the conduct of an operation there is a permanent interaction between patient and system factors or in a temporal sequence between pre- and intraoperative factors.15 In a landmark study from Great Ormond St Hospital in London, de Leval and colleagues used human factors researchers to observe team behavior and adjustment in performing complex pediatric cardiac surgery.16 Recovery from major as well as serial minor adverse events leads to avoidance of death and/or near-miss. This sophisticated type of analysis may be applicable in the future to other fields of cardiac surgery, including training.
Several limitations should be noted. The single-center nature of the study limits its generalization. However, it is reasonable to believe that similar results could be observed in other jurisdictions and programs. Despite risk adjustment using multivariable analysis that controls for differences in patient baseline characteristics, intra- and postoperative variables, the study’s observational design, and lack of random treatment allocation cannot completely exclude residual confounding from unrecognized or incompletely measured variables. The potential for selection bias is all the more important because residents are free to choose which cases they perform. Furthermore, there may be some intangible factor at play that attracts residents to certain cases and confers a protective effect, which is falsely attributed to resident as primary operator. Such a covariate may obviate the possible existence of a negative, deleterious association between resident as primary operator and clinical outcome after cardiac surgery. This would seem unlikely given that by most important well-established risk factors the adverse outcomes the resident cases were higher risk, yet the outcomes were not significantly different. Therefore, the influence of a residual confounding variable, if present, is likely to be minimal. Another limitation is the use of clinical outcomes instead of more sensitive measures of technical success such as angiographic patency. However, the objective of this study was to examine “hard” clinical end points such as death and ischemic events requiring rehospitalization, the opportunity to acquire communitywide outcomes being a strength of this work. Despite these limitations, our results strongly support the view that cardiac operations performed by surgeons in training are not associated with adverse outcomes after hospital discharge.
Sources of Funding
D.K. was supported by a Canadian Institutes of Health Research Team Grant in Cardiovascular Outcomes Research to the Canadian Cardiovascular Outcomes Research Team.
Presented at the American Heart Association Scientific Sessions, November 4–7, 2007, Orlando, Fla.
Lim E, Tsui S, on behalf of the registrars and consultant cardiac surgeons of Papworth Hospital 2003–2005. Impact of the European Working Time Directive on exposure to operative cardiac surgical training. Eur J Cardiothorac Surg. 2006; 30: 574–577.
Cook NR. Use and misuse of the receiver operating characteristic curve in risk prediction. Circulation. 2007; 115: 928–935.
Lemeshow S, Hosmer DW. A review of goodness of fit statistics for use in the development of logistic regression models. Am J Epidemiol. 1982; 115: 92–106.
Cox DR. Regression analysis and life tables. J R Stat Soc. 1972; 34: 187–220.
Barlow WE, Prentice RL. Residuals for relative risk regression. Biometrika. 1988; 75: 65–74.
Ghali WA, Quan H, Brant R, van Melle G, Norris CM, Faris PD, Galbraith PD, Knudtson ML, for the APPROACH (Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease) Investigators. Comparison of 2 methods for calculating adjusted survival curves from proportional hazards models. JAMA. 2001; 286: 1494–1497.
Stoica SC, Sharples LD, Ahmed I, Roques F, Large SR, Nashef SA. Preoperative risk prediction and intraoperative events in cardiac surgery. Eur J Cardiothorac Surg. 2002; 21: 41–46.