Risk Model of Thoracic Aortic Surgery in 4707 Cases From a Nationwide Single-Race Population Through a Web-Based Data Entry System
The First Report of 30-Day and 30-Day Operative Outcome Risk Models for Thoracic Aortic Surgery
Background— The objective of this study was to collect integrated data from nationwide hospitals using a web-based national database system to build up our own risk model for the outcome from thoracic aortic surgery.
Methods and Results— The Japan Adult Cardiovascular Surgery Database was used; this involved approximately 180 hospitals throughout Japan through a web-based data entry system. Variables and definitions are almost identical to the STS National Database. After data cleanup, 4707 records were analyzed from 97 hospitals (between January 1, 2000, and December 31, 2005). Mean age was 66.5 years. Preoperatively, the incidence of chronic lung disease was 11%, renal failure was 9%, and rupture or malperfusion was 10%. The incidence of the location along the aorta requiring replacement surgery (including overlapping areas) was: aortic root, 10%; ascending aorta, 47%; aortic arch, 44%; distal arch, 21%; descending aorta, 27%; and thoracoabdominal aorta, 8%. Raw 30-day and 30-day operative mortality rates were 6.7% and 8.6%, respectively. Postoperative incidence of permanent stroke was 6.1%, and renal failure requiring dialysis was 6.7%. OR for 30-day operative mortality was as follows: emergency or salvage, 3.7; creatinine >3.0 mg/dL, 3.0; and unexpected coronary artery bypass graft, 2.6. As a performance metric of the risk model, C-index of 30-day and 30-day operative mortality was 0.79 and 0.78, respectively.
Conclusion— This is the first report of risk stratification on thoracic aortic surgery using a nationwide surgical database. Although condition of these patients undergoing thoracic aortic surgery was much more serious than other procedures, the result of this series was excellent.
Thoracic aortic surgery is one of the most challenging in the entire field of surgery. Because several aspects can combine to make aortic surgery difficult (eg, location of the lesion, patient’s age, degree of atherosclerosis, and so on), it would be helpful for aortic surgeons to know the nature of parameters affecting the outcome of aortic surgery and the reasons why. Although many hospitals and surgeons have tried to clarify these parameters in their own units, to date, there has been no report of any nationwide study.1–3 A risk stratification of surgical procedures would help to improve quality control. In the cardiovascular surgery field, 2 major sophisticated databases are available: the STS National Database in the United States and the EuroSCORE system in Europe. Nishida et al reported recently that the EuroSCORE system worked well in Japanese patients undergoing aortic surgery in a single hospital.4 However, this does not mean that we should not endeavor to establish our own database on aortic surgery. To this end, in this article, we report for the first time on the risk stratification for thoracic aortic surgery based on data from a single nation. We used data from the Japan Adult Cardiovascular Surgery Database (JACVSD) to build up this stratification. Construction of this national database started in 2000 and a web-based data collection system had been introduced in 2002. The number of participating hospitals has gradually increased and currently approximately 180 hospitals are enrolled. After data cleaning, the risk model was developed using 4707 cases from 97 hospitals throughout Japan. In the future, we hope that this risk model will contribute to improved quality control of aortic surgery not only in Japan, but also throughout the world.
The JACVSD started in 2000 to estimate surgical outcomes after cardiovascular procedures in many centers throughout Japan. The database currently captures clinical information from approximately 180 hospitals (32.7% of all Japanese units performing cardiac surgery in 2005). The data collection form has a total of 255 variables (definitions are available online at www.jacvsd.umin.jp), and these are almost identical to those in the STS National Database (definitions are available online at http://wts.org). JACVSD constructed software for a web-based data collection system, and through this system, the data manager of each participating hospital was responsible for forwarding their data electronically to the central office. Although participation in the JACVSD is voluntary, data completeness is a high priority. The accuracy of submitted data was maintained by a data audit; this was achieved by random, monthly visits by administrative office members to a participating hospital when data were checked against clinical records. The validity of JACVSD data has been confirmed further by an independent comparison of the volume of cardiac surgery at a particular hospital entered in the JACVSD versus that reported to the Japanese Association for Thoracic Surgery (JATS) Registry.5 Data were excluded from 14 centers that had entered less than 90% of the data entered into the JATS Registry. Exclusion of data from these 14 centers did not affect the outcome in terms of establishing preoperative risk.
We examined all cardiovascular surgery procedures relating to thoracic aortic surgery between January 1, 2000, and December 31, 2005. Initially, those JACVSD records that had been obtained without patients’ informed consent were excluded from this analysis. Records with missing age (or out of range), sex, or 30-day status (see “End Points” for explanation) were also excluded. With the exception of body surface area, and preoperative creatinine value, all missing or out-of-range values were imputed using the variable-specific median value. After this data cleaning, the population for this risk model analyses resulted in 4707 aortic procedures from 97 participating sites throughout Japan.
The primary outcome measure of JACVSD was 30-day mortality, defined as death within 30 days of an operation regardless of the patient’s geographic location. It included death within 30 days of an operation even if the patient has been discharged from the hospital. The secondary outcome measure was 30-day operative mortality, which was exactly the same as the 30-day operative mortality but as expressed in the STS National Database. This meant that any patient who died within the index hospitalization, regardless of the length of hospital stay, and including any patient who died after being discharged from hospital up to 30 days from the date of the operation. Major morbidity was as defined as any of the following 5 postoperative in-hospital complications: stroke, reoperation for whatever reason, need for mechanical ventilation for more than 24 hours after surgery, renal failure, or deep sternal wound infection.6 In this analysis, composite operative mortality or major morbidity was used as the third end point.
The statistical model was multiple logistic regression; the variables entered in the model were selected using bivariate tests, χ2 tests for categorical covariates, and unpaired t tests or Wilcoxon rank sum tests for continuous covariates. All variables significant at the P<0.2 level were entered into the model provided they were present in at least 2% of the sample. A multivariate stepwise logistic regression analysis was then performed for each of the 3 outcomes. Stability of the model was checked every time the variable was eliminated. In the case of continuous variables, in which the relationship with outcome was not linear such as preoperative creatinine, we determined cutoff points. When all statistically nonsignificant variables had been eliminated from the model, “goodness-of-fit” testing was used to assess how well the model had discriminated and the area under the receiver operating characteristic curve was used to assess how well the model could discriminate between patients who lived from those who had died. Model calibration (the degree to which observed outcomes were similar to the predicted outcomes from the model across patients) was examined by comparing observed with predicted average within each of 10 equal-sized subgroups arranged in increasing order of patient risk. To evaluate the model calibration, the Hosmer-Lemeshow test for the lack of “goodness of fit” was applied.
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.
Risk Profile for Study Population
The JACVSD aortic patient population (total JACVSD aortic patient records=4707), had an average age of 66.5 years (SD 12.5); 67.5% were males. In this population, 25.5% required either an emergency (patient was taken to theater immediately after the diagnosis) or salvage (patient was transferred to the operating room under resuscitation) procedure; an additional 7.5% of this population required an urgent procedure (definition of emergency and urgent are same as the STS). Other preoperative comorbidities were chronic lung disease, 11.0% (of patients), renal failure, 9.2%; acute myocardial infarction; 3.2%; reoperation, 7.9%; and rupture or malperfusion, 9.7%. The replacement surgery involved aortic root in 10.1%; ascending aorta, 46.6%; aortic arch, 44.4%; distal aorta, 20.6%; descending aorta, 26.9%; and thoracoabdominal, 8.1%. Some patients had surgery involving more than one area. An abbreviated risk profile for the study population is shown in Table 1.
Outcomes of aortic surgery in the JACVSD study population were 30-day mortality (6.71%), 30-day operative mortality (8.58%), and composite 30-day mortality or major morbidity (30.13%). These rates and those for the specific major morbidities are presented in Table 2.
Three different risk models were developed and the final logistic model with the ORs and 95% CIs for the ORs are presented in Table 3. Among the 3 models, there were 9 overlapping variables: status (elective, urgent, emergency, or salvage); renal failure or abnormal preoperative creatinine levels; reoperation; patient’s age; lung disease (none, mild, moderate, or severe); acute myocardial infarction; neurological impairment, rupture, or malperfusion; unexpected coronary artery bypass graft (CABG); or CABG surgery.
To evaluate the models’ performance, both a C-index (measure of model discrimination), which was the area under the receiver operating characteristic curve, and the Hosmer-Lemeshow test (measure of model calibration across risk groups to evaluate “goodness of fit”) were evaluated. Figures 1 through 3⇓⇓ demonstrate the calibration of the models or how well the rates for the predicted event matched those of the observed event among patient risk subgroups. The details of model performance metrics are shown in Table 4.
Thoracic aortic surgery is one of the most challenging in the entire field of surgery. This is particularly the case for emergency surgical cases, which might be at greater risk of experiencing postoperative brain damage, or be in a worse preoperative condition due to systemic atherosclerosis than other categories of cardiac patients undergoing, for example, CABG or valve surgery. The basic concept of aortic surgery is simple, cut and sew. However, surgery involving the thoracic aorta is not so simple because in the operating room, the brain is susceptible to ischemia during aortic arch reconstruction, and severe atherosclerosis can lead to embolization of multiple organs, whereas any rupture of the thoracic aorta will require a quick procedure in response before it becomes too late. Acute dissection produces chemical substances, which can cause a violent reaction in vital organs. As a result, achieving good results from surgery of the thoracic aorta is quite challenging.
In this study, 4707 procedures were examined from 97 surgical units throughout Japan. The mean age was 66.5 (SD 12.8) years; two thirds of patients were male. Preoperatively, renal failure (serum creatinine >2.0 mg/dL) was seen in 9.2% of the patients and chronic lung disease in 11%, which was higher than in the isolated CABG population, which we had previously studied.7 Reoperation cases were seen in 7.9%, which was also much higher than in this CABG population. The incidence of acute myocardial infarction was 3.2% among the population as a whole, which increased to 10.1% in the 30-day mortality group; prior rupture or malperfusion had occurred in 9.7% of cases. One third of our cases were for nonelective procedures, which was similar to that found in other countries.
In terms of the section along the aorta requiring replacement surgery, many cases included multiple areas, for example, the ascending aorta plus aortic arch or aortic arch plus distal aortic arch; almost half of the cases involved ascending aorta or aortic arch replacement. In many hospitals in Japan, surgeons do not hesitate to operate on the aortic arch area. Although the 97 hospitals participating in this study were not always well-known specialist aortic centers, our results confirmed that surgeons in Japan were not hesitating to operate on the aortic arch area if needed.
Postoperatively, permanent stroke was reported in 6.14% of patients and 6.69% patients needed hemodialysis. Although these numbers were slightly higher than reports from a few leading hospitals worldwide,8,9 we feel they are acceptable. Prolonged ventilation was seen in 18.6% of patients, which was much higher than expected. This number was overestimated, however, because the definition of prolongation was more than 24 hours, which was the same as in the previously mentioned CABG series. Because of the high number of cases that involved acute dissection, on reflection, we feel this definition should have been extended to 74 hours.
The 30-day mortality rate of the whole population was 6.71%, which was better than we expected. The JATS publishes an Annual Report of all Registry data, and the most recent version had reported that the number of thoracic aortic procedures was 8907 with an overall hospital mortality rate of 8.96%.10 The 30-day operative mortality associated with our overall procedure was 8.58%; this was similar to that of the JATS Registry report. Because of the huge variation between thoracic aorta procedures, depending on differences in the vulnerability of surgery at the various sites as well as the methods used for bypass machines and for protecting the integrity of the brain’s function, it is difficult to discuss the overall outcome. Among our 4707 cases, the 30-day operative mortality rate of aortic arch replacement was 9.69% and that of the distal arch was 9.40%. Mortality rates from the world’s leading hospitals for patients undergoing aortic arch replacement have been between 4% and 7%.9,11–13 On the other hand, results from standard hospitals were not as good as from these leading hospitals. For example, the UK Cardiac Surgical Register in 2000 reported that the mortality rate for the aortic arch procedure was 28%.14 The situation is the same for thoracoabdominal aortic surgery. Although a few leading hospitals are reporting excellent data such as an operative mortality rate of 5% to 14%, a multicenter discharge database in the United States covering 20% of that country’s hospitals, including outcomes from many nonleading hospitals, reported 22.3% mortality from thoracoabdominal aorta procedure.15 Although these data were old, and results are very likely to be much improved by now, nevertheless, these high mortality rates may be indicative of the real world of thoracic aortic surgery. In comparison with this information, the Japanese results for thoracic aortic procedure presented here were better than the US data. Our data were not just from the few top leading hospitals, but rather they represent 97 hospitals that cover more than 20% of hospitals performing thoracic aortic surgery in Japan. The reason for this better result in Japan is not easy to clarify and is multifactorial. Patients with thoracic aortic disease are more prevalent in Asia than in Western countries, and Japanese surgeons may be better experienced at handling such patients than their Western counterparts. Furthermore, the socioeconomic situation in Japan, including health insurance systems, is much different from the Western world. Japanese public health insurance covers the majority of medical expenses, and only a small payment is necessary from the individual patient after surgery. One benefit of this generous insurance system is that, in Japan, patients can stay in the hospital for several weeks without incurring excessive extra charges. Patients can also choose their surgeon and hospital regardless of their insurance policy or level of coverage. Surgeons are at liberty to perform surgical procedures and medical interventions of their own choice with less pressure and interference from the hospital’s fund managers or the insurance companies than would be the case in many other countries. In Japan, a surgeon’s salary is at a fixed, flat rate; therefore, it could be argued that they think more highly of retaining their reputation rather than their remuneration with the end result that they would make every effort to save a patient’s life even if the costs were high in monetary terms. However, this generous situation may soon change because of recent increasing economic pressure from the government. Postoperative care in an intensive care unit or coronary care unit is mainly maintained by doctors who are either trainee surgeons or intensive care specialists, and not by the nursing staff, a system that may also be responsible for the better outcomes than seen elsewhere.
To evaluate the outcome of surgical procedure in this study, we mainly used 30-day mortality and 30-day operative mortality rates. It has been said that 30-day mortality or 30-day operative mortality rates represent a biased interval.16 Some reports compare the outcome after a few days, whereas others recommend comparing after 60 days or longer.17 The purpose of the database was to establish a means of quality control, and to accomplish this purpose, we needed a parameter, or parameters, that would allow us to compare our latest outcomes with old ones or with other parties’ outcomes. When we started this project, the 2 major cardiac surgery databases were the STS National Database and the EuroSCORE, both of whom were using 30-day mortality and 30-day operative mortality as the main outcome parameters. Although we recognized the limitations of these parameters, nevertheless, we learned a lot from these 2 major databases and decided to use the same parameters.
From our risk model, important variables affecting the 30-day operative mortality rates were status of emergency or salvage procedure (OR, 3.67; 95% CI, 2.80 to 4.81), preoperative high creatinine levels >3.0 mg/dL (OR, 2.97; 95% CI, 1.91 to 4.61), unexpected CABG (OR, 2.58; 95% CI,1.25 to 5.34), left main coronary artery disease (OR, 2.38; 95% CI, 1.24 to 4.57). Other factors, like rupture or malperfusion, chronic lung disease (moderate to severe), bad left ventricular function (ejection fraction <30%), and diabetes with treatment, were also significant risk factors for the 30-day or 30-day operative mortality in our series of patients. These parameters were almost the same as those reported in the previous studies from several major centers.8,11 Through this risk model, we have constructed our own calculator for the expected mortality from thoracic aortic surgery, similar to the EuroSCORE or STS National Database Calculator. We feel certain that the employment of this calculator will contribute to improvement in the quality control of daily practice in the field of thoracic aortic surgery.
The C-indexes (the area under the receiver operating characteristic curve) of 30-day mortality and 30-day operative mortality in this study were 0.79 and 0.78, respectively. Several hospitals have tried to use the EuroSCORE for risk stratification of their results from thoracic aortic procedures, and the C-index from the original EuroSCORE varied between 0.58 and 0.68.3,13 In comparison with these data, our risk model was shown to be a reliable model at this point with little fluctuation. Strictly speaking, we should have divided our data into 2 data sets, the analyzing data set and the validation data set, for the validation of our risk model. We fully acknowledge the importance of a validation data set to achieve a more accurate risk model. Unfortunately, the volume of our data (4707 cases) was not as large as that in the STS National Database, which contains over 500|000 pieces of data. When the EuroSCORE was developed in 1999, they used all 19|030 cases for the risk model construction and did not divide them into 2 data sets.18 The validation of their risk model was done 3 years later by using the data of STS National Database.19 Just as the Europeans did with their validation methodology, we also wanted to construct our new risk model by first using all of our data under initially limited conditions with a relatively small number of samples. When the volume of our data reaches an adequate size, hopefully in the near future, we fully intend to perform a validation by dividing our data into 2 data sets.
Several limitations exist in this study. Although the data that we analyzed came from 97 hospitals from all over Japan, it cannot be assumed that the cases retrieved were representative of all Japanese cases. Furthermore, these 97 hospitals were relatively high-volume centers in comparison to many in Japanese hospitals, which might have affected our result even if the data were not particularly large. Third, we did not divide our data into analyzing and validation data sets because of the relatively small volume of our data. As covered in the preceding “Discussion,” it is our intention to perform a validation of our risk model by dividing into the 2 data sets as soon as the volume of our data becomes large enough.
We have reported the first risk stratification study on thoracic aortic surgery that uses a nationwide cardiovascular surgery database. By analyzing 4707 procedures from 97 hospitals throughout Japan, 30-day and 30-day operative mortality rates were 6.71% and 8.58%, respectively. The results were quite satisfactory for a nationwide outcome of thoracic aortic surgery, and this system will contribute to improving the quality control of surgical practice in thoracic aortic procedures.
We thank all the data managers and hospitals, which participated in the Japan Adult Cardiovascular Surgery Database project, for their great efforts in entering the data.
Sources of Funding
S.T. received a research grant from the Management of JACVSD.
Presented at the American Heart Association Scientific Sessions, November 4–7, 2007, Orlando, Fla.
Nashef SA, Roques F, Michel P, Gauducheau E, Lemeshow S, Salamon R. European system for cardiac operative risk evaluation (EuroSCORE). Eur J Cardiothorac Surg. 1999; 16: 9–13.
Kawachi Y, Nakashima A, Toshima Y, Arinaga K, Kawano H. Risk stratification analysis of operative mortality in heart and thoracic aorta surgery: comparison between Parsonnet and EuroSCORE additive model. Eur J Cardiothorac Surg. 2001; 20: 961–966.
Barmettler H, Immer FF, Berdat PA, Eckstein FS, Kipfer B, Carrel TP. Risk-stratification in thoracic aortic surgery: should the EuroSCORE be modified? Eur J Cardiothorac Surg. 2004; 25: 691–694.
Nishida T, Masuda M, Tomita Y, Tokunaga S, Tanoue Y, Shiose A, Morita S, Tominaga R. The logistic EuroSCORE predicts the hospital mortality of the thoracic aortic surgery in consecutive 327 Japanese patients better than the additive EuroSCORE. Eur J Cardiothorac Surg. 2006; 30: 578–582;discussion 582–583.
Shroyer AL, Coombs LP, Peterson ED, Eiken MC, DeLong ER, Chen A, Ferguson TB Jr, Grover FL, Edwards FH. The Society of Thoracic Surgeons: 30-day operative mortality and morbidity risk models. Ann Thorac Surg. 2003; 75: 1856–1864;discussion 1864–1865.
Motomura N, Miyata H, Tsukihara H, Okada M, Takamoto S, JCVSD organization. First report on 30-day and operative mortality in risk model of isolated CABG in Japanese. Ann Thorac Surg. 2008; in press.
Roques F, Nashef SA, Michel P, Gauducheau E, de Vincentiis C, Baudet E, Cortina J, David M, Faichney A, Gabrielle F, Gams E, Harjula A, Jones MT, Pintor PP, Salamon R, Thulin L. Risk factors and outcome in European cardiac surgery: analysis of the EuroSCORE multinational database of 19030 patients. Eur J Cardiothorac Surg. 1999; 15: 816–822;discussion 822–823.
Nashef SA, Roques F, Hammill BG, Peterson ED, Michel P, Grover FL, Wyse RK, Ferguson TB. Validation of European System for Cardiac Operative Risk Evaluation (EuroSCORE) in North American cardiac surgery. Eur J Cardiothorac Surg. 2002; 22: 101–105.