Outcome Prediction After Coronary Surgery and Redo Surgery for Bleeding (From the KROK Registry)

Outcome Prediction After Coronary Surgery and Redo Surgery for Bleeding (From the KROK Registry)

Journal of Cardiothoracic and Vascular Anesthesia 33 (2019) 29302937 Contents lists available at ScienceDirect Journal of Cardiothoracic and Vascul...

1013KB Sizes 0 Downloads 30 Views

Journal of Cardiothoracic and Vascular Anesthesia 33 (2019) 29302937

Contents lists available at ScienceDirect

Journal of Cardiothoracic and Vascular Anesthesia journal homepage: www.jcvaonline.com

Original Article

Outcome Prediction After Coronary Surgery and Redo Surgery for Bleeding (From the KROK Registry) Piotr Knapik, MD, PhD*,1, Daniel Ciesla, Ingy, Wojciech Saucha, MD*, Ma»gorzata Knapik, MD*, Micha» O. Zembala, MD, PhDz, Piotr Przyby»owski, MD, PhDz,x, Bogus»aw Kapelak, MD, PhD{, Mariusz Kusmierczyk, MD, PhDk, Marek Jasi nski, MD, PhD#, Zdzis»aw Tobota, MD**, Bohdan J. Maruszewski, MD, PhD**, Marian Zembala, MD, PhDz, on behalf of KROK Investigators2 *

Department of Anaesthesiology, Intensive Therapy and Emergency Medicine, Silesian Centre for Heart Diseases, Medical University of Silesia, Zabrze, Poland y Department of Science and New Technologies, Silesian Centre for Heart Diseases, Zabrze, Poland z Department of Cardiac, Vascular and Endovascular Surgery and Transplantology, Silesian Centre for Heart Diseases, Medical University of Silesia, Zabrze, Poland x First Chair of General Surgery, Jagiellonian University, Medical College, Cracow, Poland { Department of Cardiovascular Surgery and Transplantology, John Paul II University Teaching Hospital, Jagiellonian University, Medical College, Cracow, Poland k Department of Cardiac Surgery and Transplantology, Institute of Cardiology, Warsaw, Poland # Department of Cardiac Surgery, University Teaching Hospital, Wroc»aw, Poland ** Department of Paediatric Cardiothoracic Surgery, The Children’s Memorial Health Institute, Warsaw, Poland

Objectives: To assess the reliability of EuroSCORE II in an entire population after isolated coronary artery surgery and separately among patients who underwent redo surgery due to bleeding, and to create a model predicting hospital death among patients who underwent redo surgery owing to bleeding. Design: Retrospective study based on data from the Polish National Registry of Cardiac Surgical Procedures. Setting: Multi-institutional study. Participants: The study comprised 41,353 patients who underwent isolated coronary artery surgery in Poland between January 2012 and December 2014. Interventions: None. Measurements and Main Results: EuroSCORE II reliability was estimated using the area under the receiver operating characteristics curve (AUC), the observed-to-expected surgical mortality ratio (O/E), and the Hosmer-Lemeshow test. Parameters of the function correcting the original EuroSCORE II were determined using the least squares method. The original score was adjusted using a created formula. Among the 41,353 patients, 1,406 (3.4%) underwent reexploration. Even though EuroSCORE II was reliable in predicting hospital mortality in the entire population (AUC 0.76, O/E ratio 1.08), it greatly underestimated mortality for patients who required reexploration (AUC 0.74, O/E ratio 4.33). In this DOI of original article: http://dx.doi.org/10.1053/j.jvca.2019.05.048. 1 Address reprint requests to Piotr Knapik, MD, PhD, Department of Anaesthesiology, Intensive Therapy and Emergency Medicine, Silesian Centre for Heart Diseases in Zabrze, Curie-Sk»odowskiej 9 Str, 41-800 Zabrze, Poland. E-mail address: [email protected] (P. Knapik). 2 Lech Anisimowicz, Andrzej Biederman, Dariusz Borkowski, Miros»aw Brykczynski, Pawe» Bugajski, Pawe» Cholewinski, Romuald Cichon, Marek Cisowski, Marek Deja, Antoni Dziatkowiak, Leszek A. Gryszko, Tadeusz Gburek, Ireneusz Haponiuk, Piotr Hendzel, Tomasz Hirnle, Stanis»aw Jab»onka, Krzysztof Jarmoszewicz, Marek Jasi nski, Ryszard Jaszewski, Marek Jemielity, Ryszard Kalawski, Bogus»aw Kapelak, Jacek Kaperczak, Maciej A.Karolczak, Micha» Krejca, Wojciech Kustrzycki, Mariusz Kusmierczyk, Pawe» Kwinecki, Bohdan Maruszewski, Maurycy Missima, Jacek J.Moll, Wojciech Ogorzeja, Jacek Pajak, ˛ Wojciech Pawliszak, Edward Pietrzyk, Grzegorz Religa, Jan Rogowski, Jacek Roz_ anski, Jerzy Sadowski, Girish Sharma, Janusz Skalski, Jacek Skiba, Janusz Sta˛z_ ka, Piotr St˛epi nski, Kazimierz Suwalski, Piotr Suwalski, Zdzis»aw Tobota, ºukasz Tu»ecki, Kazimierz Widenka, Micha» Wojtalik, Stanis»aw Wos, Marian Zembala and _ Piotr Zelazny. https://doi.org/10.1053/j.jvca.2019.04.028 1053-0770/Ó 2019 Elsevier Inc. All rights reserved.

P. Knapik et al. / Journal of Cardiothoracic and Vascular Anesthesia 33 (2019) 29302937

2931

subpopulation, the worst performance of the EuroSCORE II was noted among patients with the lowest predicted mortality (0.50%-0.82%) Accurate calibration was obtained by adding a coefficient and creating a nomogram. Conclusions: EuroSCORE II was reliable in a Polish population undergoing isolated coronary surgery. After redo surgery for bleeding, the observed mortality was much higher than in the overall coronary population, but the rate was made more accurate by adding a coefficient to the initially calculated EuroSCORE II. Ó 2019 Elsevier Inc. All rights reserved. Key Words: coronary artery surgery; reoperation; hospital mortality; prediction; euroSCORE II

THE IMPORTANCE OF surgical reexploration (SR) owing to postoperative bleeding or tamponade in the postoperative period after cardiac surgery is largely underestimated in the daily work of cardiac surgical departments. According to the Society of Thoracic Surgeons database, this seemingly innocent postoperative complication increased mortality 4.5-fold in more than 500,000 patients after coronary artery bypass grafting surgery (CABG).1 A decade later, a study from a large Swedish cardiac surgical registry reported that SR increased perioperative mortality from 2.8% to 12.0% in a contemporary cardiac surgical cohort.2 A similar trend (increase in hospital mortality from 1.8% to 13.2%) also was observed in patients after isolated CABG in a study recently published by the authors of the present study, based on the same data set used for the present study.3 Apart from excessive mortality, the authors of the present study also found that the need for SR was associated with the development of major postoperative complications. SR is, therefore, a procedure with a profound, direct effect on patient survival. A mortality risk assessment associated with the patient’s profile has been used widely in cardiac surgery for many years. The most popular tool in Europe is the European System for Cardiac Operative Risk Evaluation (EuroSCORE), first introduced in 1999.4 A new version of this model, EuroSCORE II, was presented for the first time at the 25th Congress of the European Society of Cardiothoracic Surgery in 20115 and described in a publication the following year.6 Since then, EuroSCORE II has been used widely in Europe. The modification of the original EuroSCORE was very much expected and awaited. In 2014, only 2 years after the launch of EuroSCORE II, it already was possible to conduct a meta-analysis of 22 articles that assessed its credibility. Good overall performance of this score was confirmed; however, it became clear that validation in various populations needed to be the focus of additional studies.7 Thus, the authors of the present study decided to assess the reliability of EuroSCORE II in the overall CABG population in Poland. Undoubtedly, this is a reliable tool that can estimate mortality risk precisely in every single patient qualified for cardiac surgery. However, what happens to the risk prediction when it is necessary to return to the operating room to perform a minor procedure that dramatically increases the surgical risk? Is it possible to identify patients who would be the most likely to die after the cascade of adverse events caused by surgical bleeding? Today, these questions remain unanswered. EuroSCORE II is now commonly used in thousands of cardiac surgical departments, and numerous studies have already confirmed its reliability. Extending this popular tool (with

necessary modifications) to patients after SR seems to be an attractive and logical concept, even if EuroSCORE was not originally designed for this purpose. This might be easier and more practical than creating separate scoring systems for the prediction of hospital mortality in patients undergoing SR. The primary aim of the present study was to assess the reliability of EuroSCORE II in the whole CABG population and in CABG patients requiring SR. If the EuroSCORE II model proved to be ineffective in predicting mortality after SR, the secondary aim was to create a EuroSCORE II-based model for predicting mortality among patients requiring SR.

Material and Methods This analysis was based on retrospective registry data. The only source of information was the Polish National Registry of Cardiac Surgical Operations (KROK Registry). This registry is obligatory, and all cardiac surgical centers in Poland have been entering data into the registry since 2006. All patients who underwent CABG between January 2012 and December 2014 and whose data were collected in the KROK Registry were analyzed. The personal data of patients were protected in accordance with Polish law. The sophisticated personal data protection system provided by the National Centre for Healthcare Information Systems was used during data processing. The KROK Registry makes it possible to assess each patient in terms of the most important preoperative variables. The mortality risk using EuroSCORE II was calculated for each patient since January 2012. Information on individual risk factors included all major comorbidities with the inclusion of variables necessary to calculate EuroSCORE II. Patients with missing information on key, irreplaceable variables (namely age, sex, weight, height, and surgical status) were excluded. Categorical binary variables with missing data defaulted to the most common value present in the majority of cases. All patients requiring SR in the postoperative period were identified. The primary aim of the present study was to assess the reliability of EuroSCORE II in predicting in-hospital mortality in the whole CABG population and in patients requiring SR. The secondary aim of the study was to create a reliable model for predicting mortality among patients requiring SR. For the purpose of the primary aim, EuroSCORE II discrimination and accuracy were assessed. Receiver operating characteristics curve analysis was used to estimate the performance of EuroSCORE II in predicting hospital

2932

P. Knapik et al. / Journal of Cardiothoracic and Vascular Anesthesia 33 (2019) 29302937

mortality. The area under the curve (AUC) with a 95% confidence interval (CI) was calculated. Calibration of the model was assessed using the Hosmer-Lemeshow goodness-of-fit test. The observed-to-expected surgical mortality ratio (O/E) was calculated for the whole population and separately for each consecutive quartile of EuroSCORE II. All calculations were performed for the overall CABG cohort and separately in patients who underwent SR after CABG. For the purpose of the secondary aim, the original EuroSCORE II was adjusted to the population of patients who underwent SR. Parameters of the function correcting the original EuroSCORE II were determined using the least squares method. A formula adjusting the original score was created and a corresponding nomogram was drawn. The least squares method is a standard approach in regression analysis to approximate the solution of overdetermined systems. To verify the results of the calculations, the study group was divided into a derivation cohort (75% of the population undergoing SR) and validation cohort (25% of the population undergoing SR). The derivation cohort was divided into 20 subgroups according to the EuroSCORE II percentile, and then the observed mortality in each compartment was calculated. The scatter plot with the data created this way indicated the possibility of matching with the help of a power or logarithmic function. Using the least squares method, the parameters of binding functions were selected to minimize the sum of the residuals. A function that obtained the best possible level of data fit for the derivation and validation sample was selected. The goodness of the match was assessed using the Hosmer-Lemeshow test, with both samples (derivation and validation samples) divided into 10 categories. Based on the obtained results, a nomogram was created. Continuous variables are presented as mean and standard deviations, and categorical variables are presented as percentages. To test differences between groups, the chi-square,

Mann-Whitney U, and Student t tests were used, as appropriate. For each analysis, a 2-tailed p value  0.05 was considered to be significant. Analyses were performed and graphs created with Dell Statistica (data analysis software system), version 13 (Dell, Aliso Viejo, CA). Results A total of 41,358 patients who underwent CABG in all cardiac surgical centers in Poland from 2012-2014 were subject to statistical analysis. In this population, 1,406 patients (3.4%) underwent SR in the early postoperative period. Five patients were excluded owing to missing data on key variables. In the end, 41,353 patients were included in the analysis. Reliability of EuroSCORE II The histogram of expected mortality based on EuroSCORE II is presented in Fig 1. The majority of patients (72.8%) had a EuroSCORE <2%. The mean EuroSCORE II was greater in patients who underwent SR compared with those who did not (3.0% § 5.2% v 2.0% § 3.1%; p < 0.001). Overall, the in-hospital mortality rate was 2.2% (910 observed deaths), and the mortality rate predicted by EuroSCORE II was 2.0% § 3.24% (from 0.5% to 85.1%), with a mean of 845 expected deaths (p = 0.123). EuroSCORE II presented good discrimination in identifying patients who were at risk for in-hospital death in the whole population (AUC 0.763, 95% CI 0.746-0.780; p < 0.001); however, the results did not match well for predicting in-hospital mortality (HosmerLemeshow p = 0.006). The O/E was 1.08 for all patients after CABG. Generally, EuroSCORE II was found to be reliable and slightly underestimated mortality in the whole Polish CABG population (Fig 2).

Fig 1. Distribution of EuroSCORE II in the analyzed population. SR, surgical reexploration.

P. Knapik et al. / Journal of Cardiothoracic and Vascular Anesthesia 33 (2019) 29302937

2933

Fig 2. Receiver operating characteristics curve analysis and the performance of EuroSCORE II in predicting hospital mortality in the whole population (left) and in the population requiring surgical reexploration (right). SR, surgical reexploration.

Fig 3. Mean hospital mortality and observed-to-expected surgical mortality ratio in all patients (left) and in the population requiring surgical reexploration (right). EuroSCORE ranges are specified below the bars. The dotted line indicates the optimal observed-to-expected surgical mortality ratio. O/E, observed-to-expected surgical mortality ratio; SR, surgical reexploration.

In-hospital mortality for patients who required SR was 13.2% (185 observed deaths), and the mortality predicted by EuroSCORE II was 3.0% § 5.2%, with a mean of only 43 expected deaths (p < 0.001). The AUC still showed good discrimination for a patient being at risk for in-hospital death in this subpopulation (AUC 0.746, 95% CI 0.706-0.785; p < 0.001); however, the model was not calibrated for predicting in-hospital mortality (Hosmer-Lemeshow p < 0.001). The O/E was 4.3 for patients who required SR in the postoperative period. In conclusion, among patients who required SR in the postoperative period, the EuroSCORE II model showed satisfactory discrimination but greatly underestimated mortality (see Fig 2).

The results for each quartile of EuroSCORE II for all patients and for patients who underwent SR are shown in Fig 3. In patients who required SR, the worst calibration of EuroSCORE II was noted in the lowest quartile; however, calibration of the model across the remaining quartiles also was very poor (see Fig 3). Prediction of Hospital Death With Adjusted EuroSCORE II Among patients requiring SR, EuroSCORE II showed good discrimination, but the calibration was extremely poor. To

2934

P. Knapik et al. / Journal of Cardiothoracic and Vascular Anesthesia 33 (2019) 29302937

Table 1 Patients in the Derivation and Validation Cohorts Variable Age >65 y Female sex CCS class IV NYHA class III or IV Recent MI <90 d Pulmonary hypertension LVEF <30% Previous PCA/stent Persistent or chronic AF Left main stem lesion Triple vessel disease Cigarette smoking Hypercholesterolemia Diabetes mellitus Arterial hypertension BMI >35 kg/m2 Renal failure COPD Past TIA, RIND, stroke Past treatment of CAD PVD Poor mobility Cardiogenic shock Use of IABP IV nitrates or heparin Previous cardiac surgery Nonelective surgery On-pump surgery

Derivation Cohort (n = 1,062) 572 222 200 153 360 8 46 282 59 297 658 193 651 342 939 42 176 91 52 16 136 83 58 27 192 17 477 687

Validation Cohort (n = 344)

53.9% 20.9% 18.8% 14.4% 33.9% 0.8% 4.3% 26.6% 5.6% 28.0% 62.0% 18.2% 61.3% 32.2% 88.4% 4.0% 16.6% 8.6% 4.9% 1.5% 12.8% 7.8% 5.5% 2.5% 18.1% 1.6% 44.9% 64.7%

177 75 54 54 127 0 10 100 21 90 205 51 210 112 305 10 45 32 14 2 50 33 16 5 70 3 167 231

51.5% 21.8% 15.7% 15.7% 36.9% 0.0% 2.9% 29.1% 6.1% 26.2% 59.6% 14.8% 61.1% 32.6% 88.7% 2.9% 13.1% 9.3% 4.1% 0.6% 14.5% 9.6% 4.7% 1.5% 20.4% 0.9% 48.6% 67.2%

p 0.474 0.780 0.218 0.617 0.338 0.229 0.310 0.400 0.804 0.561 0.472 0.179 0.984 0.955 0.979 0.465 0.144 0.758 0.629 0.293 0.465 0.353 0.656 0.333 0.390 0.465 0.266 0.442

Abbreviations: AF, atrial fibrillation; BMI, body mass index; CAD, carotid artery; CCS, Canadian Coronary Score; COPD, chronic obstructive pulmonary disease; IABP, intra-aortic balloon pump; IV, intravenous; LVEF, left ventricular ejection fraction; MI, myocardial infarction; NYHA, New York Heart Association; PCA, percutaneous coronary angioplasty; PVD, peripheral vascular disease; RIND, reversible ischemic neurologic deficit; SR, surgical reexploration; TIA, transient ischemic attack.

adjust for this, the parameters of a function correcting the original EuroSCORE II were determined using the least squares method. The adjusted EuroSCORE was created on the basis of the following formula: pffiffiffiffiffiffiffiffiffiffi mod ES II ¼ 12:95  ES II  6:11 The model was created with the use of a derivation cohort (75% of the population undergoing SR) and validation cohort (25% of the population undergoing SR). Both cohorts were randomly created (Table 1). In the derivation cohort, the mortality rate was 13.7% (145 observed deaths), and the mortality rate predicted by the adjusted EuroSCORE II was 13.5% § 11.7%, with a mean of 143 expected deaths (p = 0.878). The AUC showed a good discrimination in the SR population (AUC 0.754, 95% CI 0.7110.796; p < 0.001). According to the Hosmer-Lemeshow test, the model was well-calibrated for the prediction of hospital mortality (p = 0.731) and O/E was 1.01. In the validation cohort, the mortality rate was 11.6% (40 observed deaths), and the mortality rate predicted by the adjusted EuroSCORE II was 13.1% § 10.5%, with a mean of 45 expected deaths (p = 0.580). The AUC showed good discrimination in the SR population (AUC 0.715, 95% CI 0.6170.813; p < 0.001). According to the Hosmer-Lemeshow test,

the model was well-calibrated for the prediction of hospital mortality (p = 0.308) and O/E was 0.89. A nomogram of the calculations of predicted mortality, created on the basis of the aforementioned formula, is shown in Fig 4 (the initial EuroSCORE II is shown on the horizontal axis, and the calculated hospital mortality in patients requiring SR is shown on the vertical axis). The formula allows for a simple, manual conversion of the initially predicted mortality to the mortality predicted when the patient required SR in the postoperative period. Discussion EuroSCORE II is used commonly in Europe and no significant change is expected in its use in the coming years. Therefore, the authors decided to begin their research from the general assessment of the reliability of EuroSCORE II in the conditions in Poland. EuroSCORE II was found to be generally reliable but not ideally calibrated; it slightly underestimated mortality in the whole Polish CABG population. The predicted mortality rate was underestimated to different degrees within subsets of the overall population. The authors decided to use the approach presented in the work published by Borracci et al.,8 who divided the research population into quartiles

P. Knapik et al. / Journal of Cardiothoracic and Vascular Anesthesia 33 (2019) 29302937

2935

Fig 4. Nomogram for calculations of predicted mortality. Initial predicted in-hospital mortality is shown on the horizontal axis, and predicted in-hospital mortality after surgical reexploration is shown on the vertical axis. SR, surgical reexploration.

and analyzed O/E separately for each of them.8 Using this method, the authors of the present study found that at low EuroSCORE ranges (0.5%-1.24%), the risk was overestimated, whereas at higher ranges, the risk was underestimated (see Fig 2). This was not owing to the percentage of imputed missing data because the percentage was similar across all quartiles. The authors tried to compare these results to those available in the medical literature and found contradictory information. In a meta-analysis including 22 studies, the reliability of EuroSCORE II in a pooled number of 145,592 various surgical procedures was assessed.7 For isolated CABG patients, EuroSCORE II generally overestimated mortality, with an O/E of 0.804; however, underestimated results were observed in the high-risk CABG group.7 In contrast, differing results can be found in the study by Borracci et al.8 For isolated CABG procedures, the AUC was 0.794, and mortality was profoundly underestimated (with an O/E ratio as high as 1.50). The authors pointed out that medical registries tend to have incomplete data (eg, creatinine clearance, left ventricular function, or systolic pulmonary artery pressure), which is needed to correctly estimate the EuroSCORE II.8 This may be quite relevant because the authors of the present study also had to impute missing data for these variables in some patients. The divergence of results in the medical literature should not preclude careful and critical analysis of the available observational studies.9 Hogervorst et al.10 recently confirmed that EuroSCORE II underestimated mortality in the Netherlands, Spain,11 and Canada,12 but in the United Kingdom,

performance of the model was satisfactory.13 EuroSCORE II was unreliable in patients 70 years old in the United Kingdom,14 but in Canada and Turkey, EuroSCORE II had good discrimination capacity in high-risk octogenarians.15,16 Howell et al.17 stated that EuroSCORE II did not improve risk prediction in high-risk patients undergoing adult cardiac surgery compared with original additive and logistic EuroSCOREs.17 Some important problems (eg, frailty) also may not be properly addressed in the prediction of outcomes.18,19 Some authors wonder whether EuroSCORE II could be replaced by another, more accurate model. Many research articles have compared EuroSCORE II with the American Society of Thoracic Surgeons score; however, a recent meta-analysis revealed that both scoring systems perform similarly.20 The risk of isolated CABG has been increasing steadily; therefore, predictive models should be updated constantly. In a study on trends in isolated CABG over the last decade, Kindo et al. pointed out that risk profiles for this procedure are increasing.21 However, it is difficult to adjust models to constantly changing, diverse populations. A solution may soon be available. In a very recent study, a machine learning model proved to be more accurate in predicting mortality after elective cardiac surgery than EuroSCORE II.22 Machine learning is related to computational statistics and focuses on making predictions using computers, resulting in a progressive improvement in performance. Once such solutions using machine learning and artificial intelligence permanently enter medical practices, the aforementioned problems no longer will be important. Each center would be able to work on its own, constantly updated predictive model. In the

2936

P. Knapik et al. / Journal of Cardiothoracic and Vascular Anesthesia 33 (2019) 29302937

meantime, EuroSCORE II is the standard tool, and the primary aim of the present study was to assess the reliability of EuroSCORE II in patients after isolated CABG whose postoperative course was complicated with SR. An analysis of the medical literature indicates that the number of patients requiring SR after CABG seems to be relatively constant over time.1,2325 Moreover, recently published studies tend to report a higher risk of reexploration-related mortality than those published in the 1990s.26 There has not been an improvement in the mortality risk along with progress being made in cardiac surgery, and there could be multiple factors involved in increases in mortality with SR, such as postoperative complications, blood transfusions, resternotomy, periods of hemodynamic instability, or many other factors.26 SR is a complication that profoundly influences all known methods of outcome prediction. Surviving patients after SR had a significantly lower EuroSCORE II compared with nonsurvivors.27 Patients who died had a similar preoperative EuroSCORE II, regardless of whether they underwent SR.28 Patients without SR had an observed mortality rate that was half of the predicted rate, whereas patients who underwent redo surgery had an observed mortality rate double than expected.29 None of the available scoring systems has been adjusted so far to work properly under these varying conditions. It is instructive to comment on why the O/E in the present study was so greatly underestimated. What is it about SR that increases the mortality rate so much? The reasons likely are multifactorial; however, it must be kept in mind that SR is a procedure usually performed during the early postoperative period and might be associated with hemodynamic instability, periods of organ hypoperfusion, low blood pressure, excessive administration of fluids and blood products, etc. As a result, some patients can develop serious neurologic, pulmonary, renal, and abdominal complications, and these in turn are associated with a sharp deterioration of prognosis.3 In the present study, the authors aimed to use EuroSCORE II for a purpose for which it was not originally intended. The idea is not new. Even as early as 2005, Toumpoulis et al. tried to use the original EuroSCORE to predict 3-month mortality, prolonged length of stay, and specific postoperative complications and achieved very good results. Interestingly, in their study, EuroSCORE was unable to predict reexploration for bleeding.30 More recently, Barili et al.31 tried to use EuroSCORE II to predict long-term outcomes, but the results were rather disappointing.31 The use of EuroSCORE II to predict mortality in a subset of patients who developed a specific complication is, however, a completely different approach. Hospital mortality for patients who required SR in the present study’s population was 13.1%, and the predicted mortality was only 3.0% § 5.2%. O/E was greatly underestimated; however, the AUC still showed good discrimination in this specific group. Underestimated mortality rates were observed across all quartiles, but the least reliable result was among patients with low expected mortality, originally ranging from 0.5% to 0.85% (O/E 7.81). A secondary aim of the present study was to create a model for predicting hospital death specifically for patients requiring SR in the postoperative period. The original EuroSCORE was

adjusted using sophisticated statistical methods and the creation of a complicated formula. The authors present here a nomogram drawing that could be used by clinicians in everyday practice. When a patient suddenly has to return to the operating room, the original EuroSCORE may be adjusted easily to predict hospital mortality after SR more accurately. This study has one important limitation. The analysis relied entirely on data present in the medical registry, requiring a retrospective design of the study. Analyzing large amounts of data from sources such as this has associated problems, such as selection bias and missing data. Evidence-based medicine, however, must be viewed with an open mind and the shortcomings appreciated.9 In conclusion, EuroSCORE II was reliable in a Polish population undergoing isolated CABG. With SR, predicted mortality was much greater and could be made more accurate by adding a coefficient to the calculated EuroSCORE II. Acknowledgments The authors wish to thank Mrs. Jolanta Ciesla for editorial help in preparing the manuscript and to Mr Patryk Korecki for statistical support. Conflict of Interest The authors have no conflicts of interest to disclose. References 1 Mehta RH, Sheng S, O’Brien SM, et al. Reoperation for bleeding in patients undergoing coronary artery bypass surgery: Incidence, risk factors, time trends, and outcomes. Circ Cardiovasc Qual Outcomes 2009;2:583–90. 2 Ruel M, Chan V, Boodhwani M, et al. How detrimental is reexploration for bleeding after cardiac surgery? J Thorac Cardiovasc Surg 2017;154:927–35. 3 Knapik P, Knapik M, Zembala MO, et al. In-hospital and mid-term outcomes in patients reoperated on due to bleeding following coronary artery surgery (from the KROK Registry) [E-pub ahead of print]. Interact Cardiovasc Thorac Surg doi:10.1093/icvts/ivz089, Accessed 9 April 2019. 4 Nashef SA, Roques F, Michel P, et al. European system for cardiac operative risk evaluation (EuroSCORE). Eur J Cardiothorac Surg 1999;16:9–13. 5 Nashef SA, Sharples L, Nilsson J, et al. Euroscore II. Interact Cardiovasc Thorac Surg 2011;13:S73. 6 Nashef SA, Roques F, Sharples LD, et al. EuroSCORE II. Eur J Cardiothorac Surg 2012;41:734–45. 7 Guida P, Mastro F, Scrascia G, et al. Performance of the European System for Cardiac Operative Risk Evaluation II: A meta-analysis of 22 studies involving 145,592 cardiac surgery procedures. J Thorac Cardiovasc Surg 2014;148:3049–57. 8 Borracci RA, Rubio M, Celano L, et al. Prospective validation of EuroSCORE II in patients undergoing cardiac surgery in Argentinean centres. Interact Cardiovasc Thorac Surg 2014;18:539–43. 9 Muckart DJ, Malbrain MN. The future of evidence-based medicine: Is the frog still boiling? Anaesthesiol Intensive Ther 2017;49:329–35. 10 Hogervorst EK, Rosseel PMJ, van de Watering LMG, et al. Prospective validation of the EuroSCORE II risk model in a single Dutch cardiac surgery centre. Neth Heart J 2018;26:540–51. 11 Garcia-Valentin A, Mestres CA, Bernabeu E, et al. Validation and quality measurements for EuroSCORE and EuroSCORE II in the Spanish cardiac surgical population: A prospective, multicentre study. Eur J Cardiothorac Surg 2016;49:399–405.

P. Knapik et al. / Journal of Cardiothoracic and Vascular Anesthesia 33 (2019) 29302937 12 Kieser TM, Rose MS, Head SJ. Comparison of logistic EuroSCORE and EuroSCORE II in predicting operative mortality of 1125 total arterial operations. Eur J Cardiothorac Surg 2016;50:509–18. 13 Chalmers J, Pullan M, Fabri B, et al. Validation of EuroSCORE II in a modern cohort of patients undergoing cardiac surgery. Eur J Cardiothorac Surg 2013;43:688–94. 14 Poullis M, Pullan M, Chalmers J, et al. The validity of the original EuroSCORE and EuroSCORE II in patients over the age of seventy. Interact Cardiovasc Thorac Surg 2015;20:172–7. 15 Kalender M, Adademir T, Tasar M, et al. Validation of EuroSCORE II risk model for coronary artery bypass surgery in high-risk patients. Kardiochir Torakochirurgia Pol 2014;11:252–6. 16 Luc JGY, Graham MM, Norris CM, et al. Predicting operative mortality in octogenarians for isolated coronary artery bypass grafting surgery: A retrospective study. BMC Cardiovasc Disord 2017;17:275. 17 Howell NJ, Head SJ, Freemantle N, et al. The new EuroSCORE II does not improve prediction of mortality in high-risk patients undergoing cardiac surgery: A collaborative analysis of two European centres. Eur J Cardiothorac Surg 2013;44:1006–11. 18 Bagnall NM, Faiz O, Darzi A, et al. What is the utility of preoperative frailty assessment for risk stratification in cardiac surgery? Interact Cardiovasc Thorac Surg 2013;17:398–402. 19 Reichart D, Rosato S, Nammas W, et al. Clinical frailty scale and outcome after coronary artery bypass grafting. Eur J Cardiothorac Surg 2018;54:1102–9. 20 Sullivan PG, Wallach JD, Ioannidis JP. Meta-analysis comparing established risk prediction models (EuroSCORE II, STS Score, and ACEF Score) for perioperative mortality during cardiac surgery. Am J Cardiol 2016;118:1574–82.

2937

21 Kindo M, Hoang Minh T, Perrier S, et al. Trends in isolated coronary artery bypass grafting over the last decade. Interact Cardiovasc Thorac Surg 2017;24:71–6. 22 Allyn J, Allou N, Augustin P, et al. A comparison of a machine learning model with EuroSCORE II in predicting mortality after elective cardiac surgery: A decision curve analysis. PLoS One 2017;12:e0169772. 23 Munoz JJ, Birkmeyer NJ, Dacey LJ. Northern New England Cardiovascular Disease Study Group. Trends in rates of reexploration for hemorrhage after coronary artery bypass surgery. Ann Thorac Surg 1999;68:1321–5. 24 Sellman M, Intonti MA, Ivert T. Reoperations for bleeding after coronary artery bypass procedures during 25 years. Eur J Cardiothorac Surg 1997;11:521–7. 25 Kinnunen EM, De Feo M, Reichart D, et al. Incidence and prognostic impact of bleeding and transfusion after coronary surgery in low-risk patients. Transfusion 2017;57:178–86. 26 Biancari F, Airaksinen KE, Lip GY. Benefits and risks of using clopidogrel before coronary artery bypass surgery: Systematic review and meta-analysis of randomized trials and observational studies. J Thorac Cardiovasc Surg 2012;143:665–75. 27 Kristensen KL, Rauer LJ, Mortensen PE, et al. Reoperation for bleeding in cardiac surgery. Interact Cardiovasc Thorac Surg 2012;14:709–13. 28 Frojd V, Jeppsson A. Reexploration for bleeding and its association with mortality after cardiac surgery. Ann Thorac Surg 2016;102:109–17. 29 Ranucci M, Bozzetti G, Ditta A, et al. Surgical reexploration after cardiac operations: Why a worse outcome? Ann Thorac Surg 2008;86:1557–62. 30 Toumpoulis IK, Anagnostopoulos CE, Swistel DG, et al. Does EuroSCORE predict length of stay and specific postoperative complications after cardiac surgery? Eur J Cardiothorac Surg 2005;27:128–33. 31 Barili F, Pacini D, D’Ovidio M, et al. The impact of EuroSCORE II risk factors on prediction of long-term mortality. Ann Thorac Surg 2016;102:1296–303.