Risk Stratification for Cardiac Valve Replacement. National Cardiac Surgery Database

Transcription

1 Risk Stratification for Cardiac Valve Replacement. National Cardiac Surgery Database W. R. Eric Jamieson, MD, Fred H. Edwards, MD, Marc Schwartz, BS, Joseph W. Bero, MS, Richard E. Clark, MD, and Frederick L. Grover, MD for the Database Committee of The Society of Thoracic Surgeons Division of Cardiovascular Surgery, University of British Columbia, Vancouver, Canada; Division of Cardiothoracic Surgery, University of Florida, Health Sciences Center, Jacksonville, Florida; Summit Medical Systems, Minneapolis, Minnesota; Cardiovascular and Pulmonary Research Center, Allegheny-Singer Research Institute, Pittsburgh, Pennsylvania; and University of Colorado Health Sciences Center, Denver, Colorado Background. The Society of Thoracic Surgeons National Database Committee is committed to risk stratification and assessment as integral elements in the practice of cardiac operations. The National Cardiac Surgery Database was created to analyze data from subscribing institutions across the country. We analyzed the database for valve replacement procedures with and without coronary artery bypass grafting to determine trends in risk stratification. Methods. The database contains complete records of 86,580 patients who had valve replacement procedures at the participating institutions between 1986 and 1995, inclusive. The 1995 harvest of data was conducted in late 1996 and available for evaluation in These records were used to conduct an in-depth analysis of risk factors associated with valve replacement and to provide prediction of operative death by using regression analysis. Regression models were made for six subgroups. Results. Adverse patient risk factors, including diabetes, hypertension and reoperation, but not ventricular function, increased over time. There were trends with regard to increasing age of the various population subsets. The types of prostheses used remained similar over time, with more mechanical prostheses than bioprostheses used for both aortic and mitral valve replacement. There was a trend toward increased use of bioprostheses in aortic replacements and decreased use in mitral replacements between 1991 and 1995 than between 1986 and The mortality rate was determined by patient subset for primary operation and reoperation and by urgency status. The modeling showed that the predicted and observed mortality correlated for all age groups and within patient subsets. Conclusions. Risk modeling is a valuable tool for predicting the probability of operative death in any individual patient. This large, multiinstitutional database is capable of determining modern operative risk and should provide standards for acceptable care. The study illustrates the importance of risk stratification for early death both for the patient and the surgeon. (Ann Thorac Surg 1999;67:943 51) 1999 by The Society of Thoracic Surgeons The operative mortality rates for cardiac valve replacement procedures have been reported from 1% to 15% depending upon valve position, multiple replacement, and concomitant coronary artery bypass. The quality assurance of valve replacement operations can be determined only by assessing the relationship between preoperative risk factors and operative mortality. Several reports have documented variables, either preoperative or intraoperative, that are independently related to increased risk of operative mortality [1 23]. The Veterans Administration study of preoperative risk assessment for cardiac operations proposed that the ratio of observed to expected mortality is a superior method to assess quality of care than unadjusted mortality [24]. That study found, by multivariate logistic regression analysis, that priority of reoperation, age, peripheral vascular disease, mitral Presented in part at the Thirty-third Annual Meeting of The Society of Thoracic Surgeons, San Diego, CA, Feb 3 5, Address reprint requests to Dr Jamieson, University of British Columbia, # West 10th Ave, Vancouver, Canada V5Z 4E3. valve replacement, and cardiomegaly were independent predictors of death for patients who had cardiac valve operations with or without coronary artery bypass grafting. We used the Society of Thoracic Surgeons (STS) National Cardiac Surgery Database to conduct a detailed analysis of risk factors associated with valve replacement operations to predict operative mortality. In the evaluation we paid particular attention to the influence of valve position, urgency, concomitant coronary artery bypass, and reoperation. The previous studies that used the STS database primarily documented risk stratification related to myocardial revascularization [25 29]. Material and Methods All patients registered in the database who had valve replacement operations between 1986 and 1995, incorporating the 1996 harvest, were considered for inclusion in the study. Patient records that had obviously conflicting information or insufficient information were excluded 1999 by The Society of Thoracic Surgeons /99/$20.00 Published by Elsevier Science Inc PII S (99)

2 944 JAMIESON ET AL Ann Thorac Surg RISK STRATIFICATION FOR VALVE REPLACEMENT 1999;67: from the study. The resultant population contained clinical information on 86,580 patients. The definition of each risk factor was consistent with the published guidelines in the STS National Cardiac Surgery Database manual for data managers. The data incorporated a classified coding system used to ensure that the patients, surgeons, and reporting institutions were not identified. The risk stratification modeling evaluated the influence of 51 preoperative variables on operative mortality by univariate and multivariate analysis for the overall population and for each subset. Subset models were developed for isolated aortic valve replacement, isolated mitral valve replacement, multiple valve replacement, aortic valve replacement with coronary artery bypass, mitral valve replacement with coronary artery bypass, multiple valve replacement with coronary artery bypass, and valve replacement with other thoracic procedures, including aortic aneurysm resection. The analysis addresses only the six primary models but the number of patients in all seven models is provided for the purpose of general comparison of incidences. Development of the Models The modeling process for risk stratification followed a standard equation, and the probability of operative mortality for each patient was calculated. The most direct approach to validation involves comparing the predicted mortality rate with the observed mortality rate for the test set, a standard training set and test set approach in which patients were randomly assigned into one of two groups of approximately equal size. The training set was used to test the model against observed results. Statistical Analysis Software (version 6.09 for Windows, SAS Institute, Cary, NC) was used for all analyses. Each preoperative variable contained in the STS database form was considered for inclusion in the models. A stepwise multivariate analysis was done to determine the patient characteristics independently associated with operative mortality. After determining the significant risk factors, a standard logistic regression analysis was done using the training set population to develop a risk equation of the form P 1/ 1 exp x where P is the probability of postoperative death, x B 0 X 0 B 1 X 1 B k X k, each B value is a constant associated with a specific risk factor, and the X values denote the status of the risk factor for a given patient. Validation of the Models The test set population was used to test the validity of the model. Risk factors for each patient in the test set were entered into the risk equation, and the probability of operative mortality for that patient was calculated. The most direct approach to validation involves comparing the predicted mortality with the observed mortality for the test set population across the risk spectrum. As shown in Figures 1 through 4, there was excellent agreement for each of the models. Fig 1. Observed operative mortality to predicted operative mortality by risk group for isolated aortic valve replacement, 1986 to The analysis sequence used in the validation process for each model incorporated first, stepwise logistic regression analysis to identify the independent risk factors; second, entry of these risk factors into the training set population to get the regression coefficient for each risk factor for that population; and third, the training set regression coefficients used to calculate the predicted mortality rate for the test set population. The test set population for each model encompassed the total population for that model. The predicted mortality rates were compared for deviation from the observed mortality rates across the entire risk spectrum. Detailed statistical descriptions of the modeling process and the validation techniques are beyond the scope of the present work. They will be presented in a separate publication. The independent risk factors in the final models are presented as odds ratios, including 95% confidence intervals. The odds ratio predicts the magnitude of influence on mortality if the risk factor was present compared with if it was absent. Results Patient Characteristics The patient populations grouped by subset model and the respective early mortality rates are given in Table 1. The total database encompassed 86,580 patients and the overall mortality was 7.5%. There were 26,317 patients with isolated aortic valve replacement (AVR) and 13,936 patients with isolated mitral valve replacement (MVR), with respective operative mortality rates of 4.3% and 6.4%. There were 3,840 patients with multiple valve replacement and the early or operative mortality rate was 9.6%. The addition of coronary artery bypass (CAB) to the valvular procedures at least doubled the early mortality rates. Valve repair or reconstruction was done either alone or

3 Ann Thorac Surg JAMIESON ET AL 1999;67: RISK STRATIFICATION FOR VALVE REPLACEMENT 945 Fig 2. Observed operative mortality to predicted operative mortality by risk group for aortic valve replacement plus coronary artery bypass, 1986 to Fig 4. Observed operative mortality to predicted operative mortality by risk group for mitral valve replacement plus coronary artery bypass, 1986 to in combination with valve replacement in another position. Aortic valve repair was done in 597 patients, with a mortality rate of 5.9%, whereas mitral valve repair was done in 4,167 patients, with a mortality rate of 3.0%. The combination of aortic valve replacement and any valve repair was done in 938 patients and the mortality rate was 7.4%. Mitral valve replacement plus any valve repair, done in 1,266 patients, had a mortality rate of 12.5%. Tricuspid valve repair was done as an isolated procedure in 144 patients, with a mortality rate of 13.9%. There were 1,723 patients with concomitant aortic valve replacement and aortic aneurysm, with a mortality rate of 9.7%. The percentage of patients with each of 51 patientrelated variables in the overall patient population was determined. Of the total population, 45.1% were older than 70 years, and 41.6% were between 50 and 70 years old. The ejection fraction was less than 30% in only 4.6% Fig 3. Observed operative mortality to predicted operative mortality by risk group for isolated mitral valve replacement, 1986 to of the population, and 19.8% were in New York Heart Association (NYHA) class IV. Previous valvular operation was done in 12.2%, and previous coronary artery bypass in 6.8%. Of the total population, 3.7% were classified as emergent status and 1.2% as emergent-salvage status. Myocardial infarction had occurred within 21 days in 4.4% of the population. Renal failure was present in 6.0% of the patients. The characteristics of patients in each of the six subset Table 1. Patient Population and Overall Operative Mortality Rates Operative Category Number Operative Mortality (%) AVR (isolated) 26, MVR (isolated) 13, Multiple valve replacement 3, AVR CAB 22, MVR CAB 8, Multiple valve replacement CAB 1, AVR any valve repair MVR any valve repair 1, Aortic valve repair Mitral valve repair 4, Tricuspid valve repair AVR aortic aneurysm repair 1, AVR other a MVR other a Overall 86, a Other includes left ventricular aneurysm, ventricular septal defect, atrial septal defect, congenital, cardiac trauma, cardiac transplant, permanent pacemaker, automatic implanted cardioverter defibrillator, aortic aneurysm, or carotid endarterectomy. Isolated aortic valve replacement and aortic aneurysm are not included in aortic valve replacement other. AVR aortic valve replacement; CAB coronary artery bypass; MVR mitral valve replacement.

4 946 JAMIESON ET AL Ann Thorac Surg RISK STRATIFICATION FOR VALVE REPLACEMENT 1999;67: Fig 5. Mean age by year and valve procedure. (AVR aortic valve replacement; CAB coronary artery bypass; MVR mitral valve replacement.) models were similar in incidence to those in the overall population. The mean age of the patients in the major models (AVR, MVR, AVR plus CAB and MVR plus CAB) by year from 1986 to 1995 are given in Figure 5. For each of the major models the age distribution by decade is shown in Figure 6. The AVR patients were older than MVR patients, and when combined with coronary artery bypass were considerably older. The age differential evaluation showed that, for the interval 1991 to 1995, the patients were older for isolated AVR, isolated MVR, CAB plus AVR, and CAB plus MVR than for the interval 1986 to 1990 ( p ). There was no age differential for multiple valve replacement and CAB plus multiple valve replacement for the time intervals. The types of prostheses used for isolated AVR and isolated MVR were similar for the two time periods, 1986 to 1990 and 1991 to In the latter time period, for aortic replacements, 59.2% were mechanical prostheses and 37.6% were bioprostheses. For mitral replacements, 73.1% were mechanical and 23.1% were bioprostheses. Homografts and autografts used for aortic valve replacement with and without concomitant valve procedures are incorporated in the bioprostheses group. There were 576 homograft AVR, with a mortality rate of 8.3% (n 48), and 71 autograft AVR, with a mortality rate of 5.6% (n 4). Table 2 compares the predominant models by time interval for the eight important patient characteristics. For the four major models (isolated AVR, isolated MVR, AVR plus CAB, and MVR plus CAB) there was no appreciable difference in mean age, gender distribution, ejection fraction, myocardial infarction within 21 days of operation and cardiogenic shock during both time intervals. The incidence of reoperation was not different for isolated AVR and isolated MVR, but it did increase when CAB was combined with AVR or MVR at the initial valve replacement operation. The incidence of diabetes mellitus and hypertension was higher in the later time interval for the four major models. Operative Mortality The mortality rate associated with each patient-related variable was determined for the overall population. The mortality rates were also determined for the six models, namely, AVR, CAB plus AVR, MVR, CAB plus MVR, multiple replacement, and CAB plus multiple replacement. The preoperative risk factors were initially evaluated for the six models by univariate analysis. Multiple stepwise logistic regression analysis was done for the procedure models to determine the significant preoperative risk factors. The predominant preoperative risk factors by procedures are stratified in Tables 3 through 5 by assessment of the odds ratios. The preoperative risk factors associated with highest operative mortality rates were salvage status, renal failure (dialysis dependent and non-dialysis dependent), emergent status, multiple reoperations, and NYHA class IV. The operative mortality rates for age groups, by decade, is shown in Table 6 for each subset model. The mortality rate increases significantly for patients over 70 years of age who had isolated aortic valve replacement and for patients over 60 years of age who had isolated mitral valve replacement or multiple valve replacement. The addition of coronary artery bypass increased the mortality rate significantly for all age groups and for all subset models. The operative mortality rate of each subset model by incidence and status is shown in Table 7. The observed and predicted mortality rates are shown by risk groups of the population size for each model in Figures 1 to 4. Comment The STS National Cardiac Surgery Database contains information from many institutions, both nonacademic and academic centers, on cardiac valve replacement operations in 86,580 patients during a 10-year period. The database provides a benchmark for quality assurance, because institutions can enter their patient population data and obtain direct risk-adjusted comparisons of their results with the national standard. Fig 6. Age distribution by valve procedure 1986 to (AVR aortic valve replacement; CAB coronary artery bypass; MVR mitral valve replacement.)

5 Ann Thorac Surg JAMIESON ET AL 1999;67: RISK STRATIFICATION FOR VALVE REPLACEMENT 947 Table 2. Comparison of Patient Characteristics by Time Intervals and Procedures Procedure Isolated aortic valve replacement (n) 3,441 22,876 Mean age (y) MI 21 days 0.8% 1.1% Mean EF 51.1% (n 1,921) 51.8% (n 15,654) Cardiogenic shock 1.1% 1.0% Female gender 39.6% 40.5% Reoperation 10.8% 13.2% Diabetes 8.6% 13.0% Hypertension 33.2% 43.3% Isolated mitral valve replacement (n) 2,217 11,719 Mean age (y) MI 21 days 1.3% 1.8% Mean EF 53.8% (n 1,294) 54.4% (n 8,110) Cardiogenic shock 1.8% 2.7% Female gender 64.2% 63.5% Reoperation 28.4% 29.0% Diabetes 8.8% 10.7% Hypertension 23.3% 32.9% Aortic valve replacement plus 2,594 20,119 coronary artery bypass (n) Mean age (y) MI 21 days 3.7% 6.9% Mean EF 48.5% (n 1,769) 49.9% (n 14,917) Cardiogenic shock 2.2% 1.3% Female gender 29.2% 31.6% Reoperation 9.7% 11.6% Diabetes 16.1% 23.4% Hypertension 41.5% 57.5% Mitral valve replacement plus 3,441 22,876 coronary artery bypass (n) Mean age (y) MI 21 days 12.4% 19.8% Mean EF 45.2% (n 906) 47.8% (n 5,704) Cardiogenic shock 7.6% 9.0% Female gender 48.0% 48.7% Reoperation 15.1% 16.7% Diabetes 19.4% 24.8% Hypertension 36.5% 52.8% EF ejection fraction; MI myocardial infarction. The patient population that had valve replacement operations has evolved during the past decade. The age differential evaluation identified that, in the past 5 years, patients who had AVR and MVR with or without myocardial revascularization were older than those who had the same procedures in previous years. These patient populations have an increased incidence of diabetes mellitus and hypertension. There has been an increase in reoperations for isolated AVR and for AVR and MVR combined with CAB. There has been no change in ventricular function in all subgroups over the decade. We evaluated 51 preoperative risk factors to determine risk models. Most of the population was over 50 years of age; 42% were between 50 and 70 years and 45% were more than 70 years of age. Of the total population, 12% had a previous valve operation and 7% had previous coronary artery bypass. Renal failure was present in 6% of the population. Of the total patient group, the urgency status was classified as emergent in 4% and emergentsalvage in 1.2%. Of the total population, 20% were classified as NYHA class IV, and 4.6% had an ejection fraction of less than 30%. The predominant preoperative risk factors associated with operative death were determined by variate analysis. Univariate and multivariate analysis were used to determine the most dominant independent predictors of early death. The risk models developed in this study defined the net effect of any given set of preoperative risk factors [26]. The model used statistical algorithms to calculate the probability of operative death. The presence or absence of risk factors for any given patient was statistically manipulated to provide a risk-adjusted estimate of mortality based on the total experience of all patients in the database. Logistic regression was used to

6 948 JAMIESON ET AL Ann Thorac Surg RISK STRATIFICATION FOR VALVE REPLACEMENT 1999;67: Table 3. Independent Risk Factors for Operative Mortality (Odds Ratios) for Isolated Aortic Valve Replacement and Aortic Valve Replacement plus Coronary Artery Bypass Aortic Valve Replacement Aortic Valve Replacement plus Coronary Artery Bypass Risk Factor Odds Ratio CI Risk Factor Odds Ratio CI Salvage status Salvage status DDRF DDRF ES Reoperation Mult reop NDRF NDRF ES Resuscitation Preop IABP First reoperation Female gender CS CS NYHA IV NYHA IV Inotropic agent used TVD CVA CVA MI Diabetes Female gender Obesity US COPD Diabetes LMD CHF PVD Arr Diuretics Age MI EF Arr Age EF Arr arrhythmia; CHF congestive heart failure; CI 95% confidence interval; COPD chronic obstructive pulmonary disease; CS cardiogenic shock; CVA cerebrovascular accident; CVD cerebrovascular disease; DDRF dialysis-dependent renal failure; EF ejection fraction; ES emergency status; HI hemodynamic instability; LMD left main disease; MI myocardial infarction; Mult reop multiple reoperation; NDRF non-dialysis-dependent renal failure; NYHA IV New York Heart Association class IV; PH pulmonary hypertension; Preop IABP preoperative intraaortic balloon pump; PVD peripheral vascular disease; TVD triple vessel disease; US urgent status. develop the statistical algorithms. The influence of changing patient characteristics in the patient population over time affects the conditional probability matrix to reflect current patient characteristics. The impact of preoperative risk factors in valve replacement operations has been reported in detail since 1985 [1 24]. There has been considerable variation in risk factors that have been identified as independent predictors of operative death. The differences in significance of risk factors among studies have resulted in confusion for individual surgeons and surgical groups. The predictors identified in the present study of the STS database will be Table 4. Independent Risk Factors for Operative Mortality (Odds Ratios) for Isolated Mitral Valve Replacement and Mitral Valve Replacement plus Coronary Artery Bypass Mitral Valve Replacement Mitral Valve Replacement plus Coronary Artery Bypass Risk Factor Odds Ratio CI Risk Factor Odds Ratio CI Salvage status Salvage status DDRF ES ES Renal failure NDRF CS Mult reop MI NYHA IV Diabetes US NYHA IV HI Female gender First reoperation CVD MI Reoperation Inotropic agent used LMD Diabetes US Age Age EF Abbreviations as in Table 3.

7 Ann Thorac Surg JAMIESON ET AL 1999;67: RISK STRATIFICATION FOR VALVE REPLACEMENT 949 Table 5. Independent Risk Factors for Operative Mortality (Odds Ratios) for Multiple Valve Replacement and Multiple Valve Replacement plus Coronary Artery Bypass Multiple Valve Replacement Multiple Valve Replacement plus Coronary Artery Bypass Risk Factor Odds Ratio CI Risk Factor Odds Ratio CI DDRF Salvage status Salvage status DDRF CS Preop IABP NDRF NDRF Mult reop NYHA IV Preop IABP Reop MI Diabetes US MI First reoperation PH NYHA IV Age Diabetes Age Abbreviations as in Table 3. compared with those identified by individual academic centers. The multivariate logistic regression analysis identified 30 independent preoperative risk factors among the six valvular models, isolated or in combination with CAB. The odds ratios of these risk factors ranged from one to over seven. One third of the preoperative risk factors had an odds ratio greater than 1.5. The remaining two thirds of the independent risk factors have odds ratios between 1 and 1.5. The upper level risk factors include salvage and emergent status, dialysis-dependent and non-dialysisdependent renal failure, reoperations (particularly multiple reoperations), and NYHA class IV. In the valvular models with concomitant CAB, high-risk factors also include cardiogenic shock, myocardial infarction, and diabetes mellitus. The lower level risk factors incorporate various combinations of the remaining predictive factors for each of the six models. The lower level risk factors include advanced age, female gender, urgent status, reduced ejection fraction, cerebrovascular accident, congestive heart failure, arrhythmias, and use of inotropic agents. Several investigators have determined independent risk factors for AVR in the general population and in the elderly, and also with concomitant CAB [1 11]. Most of those studies identified emergent procedures, advanced NYHA class, and renal failure as risk factors. Many Table 6. Operative Mortality Rate by Age Group and Valve Position and Population Size (1986 to 1995) Procedure Age (y) Overall AVR Mortality rate 3.1% 2.0% 3.0% 2.9% 3.2% 5.3% 8.5% 14.5% 4.3% Population 554 1,382 2,401 3,686 7,001 8,468 2, ,317 AVR CAB Mortality rate 21.4% 11.7% 6.1% 4.7% 5.1% 8.6% 12.5% 18.8% 8.0% Population ,709 6,120 10,617 3, ,713 MVR Mortality rate 2.3% 3.5% 3.4% 4.1% 6.1% 9.8% 13.4% 25.0% 6.4% Population ,791 2,815 4,062 3, ,936 MVR CAB Mortality rate 0.0% 13.2% 5.2% 8.6% 12.3% 18.4% 25.1% 42.9% 15.3% Population ,050 2,889 3, ,788 Multiple Mortality rate 5.7% 3.8% 5.1% 5.5% 9.9% 15.7% 23.7% 0.0% 9.6% Population ,840 Multiple CAB Mortality rate 0.0% 30.8% 5.4% 10.3% 16.5% 21.4% 26.6% 100% 18.8% Population ,424 AVR aortic valve replacement; CAB coronary artery bypass; Multiple multiple valve replacement; MVR mitral valve replacement.

8 950 JAMIESON ET AL Ann Thorac Surg RISK STRATIFICATION FOR VALVE REPLACEMENT 1999;67: Table 7. Operative Mortality Rate by Incidence and Status and Population Size (1986 to 1995) Procedure First/Elec First/Urge First/Emer First/Salv Reop/Elec Reop/Urge Reop/Emer Reop/Salv Overall AVR Mortality rate 3.3% 5.1% 15.9% 30.6% 5.9% 10.7% 22.6% 44.0% 4.3% Population 19,831 2, , ,317 AVR CAB Mortality rate 6.0% 8.9% 18.6% 51.7% 14.2% 14.3% 23.5% 57.1% 8.0% Population 15,963 3, , ,713 MVR Mortality rate 3.8% 7.5% 22.0% 43.6% 6.2% 13.3% 21.7% 35.1% 6.4% Population 8,364 1, , ,936 MVR CAB Mortality rate 10.4% 17.3% 31.9% 45.4% 14.2% 16.6% 25.6% 37.1% 15.3% Population 4,902 1, , ,788 Multiple Mortality rate 7.0% 11.0% 17.9% 33.3% 11.0% 22.5% 20.0% 62.5% 9.6% Population 2, ,840 Multiple CAB Mortality rate 13.9% 25.0% 22.5% 41.2% 22.5% 25.5% 60.0% 100% 18.8% Population ,424 AVR aortic valve replacement; CAB coronary artery bypass; Elec elective status; Emer emergency status; First first cardiac operation; Mult multiple valve replacement; MVR mitral valve replacement; Reop reoperation; Salv salvage status; Urge urgent status. reports found high-risk factors that, in the STS evaluation, were low-risk namely, advanced age, concomitant coronary artery disease, low ejection fraction, and peripheral vascular disease. In a study of predictors of mortality after AVR, Christakis and colleagues [19] reported endocarditis, previous operation, and coronary artery disease as predictors, whereas Scott and colleagues [1] found that those variables were not predictive. Endocarditis was not included as a potential risk factor in the present analysis. It is generally agreed that the risk is higher for patients with endocarditis compared with those who have an elective procedure for chronic valve disease. Coronary artery bypass also was not used in the regression model, because valve operations and valve operations plus CAB were evaluated independently. Previous operation and the requirement for CAB are definite independent predictors of early death in the present analysis, in accordance with the report by Christakis and colleagues [19]. The study by Scott and colleagues [1] is from an earlier era, 1967 to 1981, when coronary angiography was not done routinely for patients with chronic valve disease. During that time, higher mortality rates were not always explained by the risk profile of patients. These factors together with patient selection and the time period are important with regard to the final composition of any regression model. The most important predictors for MVR were the same as for AVR, namely, salvage and emergent status, renal failure and NYHA class IV. The predictors associated with MVR and CAB also included cardiogenic shock, myocardial infarction, and diabetes mellitus. The predictors reported in the literature include NYHA class IV, advanced age, urgent status, endocarditis, low ejection fraction, coronary artery disease, and cardiogenic shock [12 16]. The independent predictors of early death after multiple valve replacement and multiple valve replacement plus CAB were similar to those of isolated valve replacements. The predominant risk factors for multiple valve replacement were dialysis-dependent renal failure, salvage status, cardiogenic shock, multiple reoperations, and non-dialysis-dependent renal failure. For multiple valve replacement plus CAB the predominant predictors were salvage status, preoperative intraaortic balloon pump, dialysis-dependent and non-dialysis-dependent renal failure, reoperation, and NYHA class IV. The predictors identified in two reported series were previous operation, tricuspid valve procedure, tricuspid regurgitation, CAB, and mitral valve lesion [17, 18]. Perioperative variables have also been documented as independent predictors of operative mortality [8, 22]. The present risk stratification did not include perioperative variables in the predictive modeling. He and colleagues [9] identified cardiopulmonary bypass and ischemic time as predictors with AVR, which also included advanced age, concomitant mitral valve disease, and aortic insufficiency. Concomitant CAB doubled operative mortality rates in all valvular models (AVR, MVR, and multiple valve replacement). Concomitant AVR and aortic aneurysm resection doubled operative mortality rates over the rate for isolated AVR. In 1985, Scott and colleagues [1] found that aortic aneurysm resection was not a predictor of death after AVR. An important finding of the present study is that the mortality rate for mitral valve reconstruction is half that of isolated mitral valve replacement; however, a detailed evaluation of this observation is beyond the scope of this report. We also found that reoperative mortality rates are considerably higher for elective and urgent operations but not

9 Ann Thorac Surg JAMIESON ET AL 1999;67: RISK STRATIFICATION FOR VALVE REPLACEMENT 951 for emergent and emergent-salvage operations in all valvular models. The same observations regarding reoperations have been reported by Husebye and associates [21], Biglioli and associates [22], and Christenson and colleagues [23]. References 1. Scott WC, Miller DC, Haverich A, et al. Determinants of operative mortality for patients undergoing aortic valve replacement: discriminant analysis of 1479 operations. J Thorac Cardiovasc Surg 1985;89: Craver JM, Weintraub WS, Jones EL, Guyton RA, Hatcher CR Jr. Predictors of mortality, complications, and length of stay in aortic valve replacement for aortic stenosis. Circulation 1988;78(3 Pt 2)I Lytle BW, Cosgrove DM, Taylor PC, et al. Primary isolated aortic valve replacement: early and late results. J Thorac Cardiovasc Surg 1989;97: Galloway AC, Colvin SB, Grossi EA, et al. Ten-year experience with aortic valve replacement in 482 patients 70 years of age or older: operative risk and long-term results. Ann Thorac Surg 1990;49: Lund O, Pilegaard H, Nielsen TT, Knudsen MA, Magnussen K. Thirty-day mortality after valve replacement for aortic stenosis over the last 22 years: a multivariate risk stratification. Eur Heart J 1991;12: Bergus BO, Feng WC, Bert AA, Singh AK. Aortic valve replacement (AVR): influence of age on operative morbidity and mortality. Eur J Cardiothorac Surg 1992;6: Aranki SF, Rizzo RJ, Couper GS, et al. Aortic valve replacement in the elderly. Effect of gender and coronary artery disease on operative mortality. Circulation 1993;88 (5 pt 2):II Morris JJ, Schaff HV, Mullany CJ, et al. Determinants of survival and recovery of left ventricular function after aortic valve replacement. Ann Thorac Surg 1993;56: He GW, Acuff TE, Ryan WH, et al. Aortic valve replacement: determinants of operative mortality. Ann Thorac Surg 1994; 57: Chocron S, Etievent JP, Viel JF, et al. Aortic valve replacement in the elderly: a comparative assay of potential risk factor modification. J Heart Valve Dis 1995;4: Magovern JA, Pennock JL, Campbell DB, et al. Aortic valve replacement and combined aortic valve replacement and coronary artery bypass grafting: predicting high risk groups. J Am Coll Cardiol 1987;9: Christakis GT, Kormos RL, Weisel RD, et al. Morbidity and mortality in mitral valve surgery. Circulation 1985;72(Suppl II):II Scott WC, Miller DC, Haverich A, et al. Operative risk of mitral valve replacement: discriminant analysis of 1329 procedures. Circulation 1985;72(3 Pt 2):II Magovern JA, Pennock JL, Campbell DB, Pierce WS, Waldhausen JA. Risks of mitral valve replacement and mitral valve replacement with coronary artery bypass. Ann Thorac Surg 1985;39: Garcia Andrade I, Cartier R, Panisi P, Ennabli K, Grondin CM. Factors influencing early and late survival in patients with combined mitral valve replacement and myocardial revascularization and in those with isolated replacement. Ann Thorac Surg 1987;44: Szecsi J, Herijgers P, Sergeant P, Daene W, Scheys I, Flameng W. Mitral valve surgery combined with coronary bypass grafting: multivariate analysis of factors predicting early and late results. J Heart Valve Dis 1994;3: Teoh KH, Christakis GT, Weisel RD, et al. The determinants of mortality and morbidity after multiple-valve operations. Ann Thorac Surg 1987;43: Galloway AC, Grossi EA, Baumann FG, et al. Multiple valve operation for advanced valvular heart disease: results and risk factors in 513 patients. J Am Coll Cardiol 1992;19: Christakis GT, Weisel RD, David TE, Salerno TA, Ivanov J. Predictors of operative survival after valve replacement. Circulation 1988;78(3 pt 2):I Kirklin JK, Naftel DC, Blackstone EH, Kirklin JW, Brown RC. Risk factors for mortality after primary combined valvular and coronary artery surgery. Circulation 1989;79(6 pt 2): II Husebye DG, Pluth JR, Piehler JM, et al. Reoperation on prosthetic heart valves: an analysis of risk factors in 552 patients. J Thorac Cardiovasc Surg 1983;86: Biglioli P, Di Matteo S, Parolari A, Antona C, Arena V, Sala A. Reoperative cardiac valve surgery: a multivariable analysis of risk factors. Cardiovasc Surg 1994;2: Christenson JT, Velebit V, Maurice J, Simonet F, Schmuziger M. Valve reoperations identification of risk factors and comparison with first-time operations. Thorac Cardiovasc Surg 1994;42: Grover FL, Hammermeister KE, Burchfiel C. Initial report of the Veterans Administration Preoperative Risk Assessment Study for cardiac surgery. Ann Thorac Surg 1990;50: Edwards FH, Clark RE, Schwartz M. Practical considerations in the management of large multiinstitutional databases. Ann Thorac Surg 1994;58: Edwards FH, Clark RE, Schwartz M. Coronary artery bypass grafting: the Society of Thoracic Surgeons National Database experience. Ann Thorac Surg 1994;57: Edwards FH, Clark RE, Schwartz M. Impact of internal mammary conduits on operative mortality in coronary revascularization. Ann Thorac Surg 1994;57: Edwards FH, Carey JS, Grover FL, Bero J, Hartz RS. Impact of gender on coronary bypass operative mortality. Ann Thorac Surg 1998;66: Clarke RE. The STS Cardiac Surgery National Database: an update. Ann Thorac Surg 1995;59: Appendix. Members of the 1997 National Database Committee 1997 National Database Committee Members Frederick L. Grover (Chair), Anthony J. Acinapura, W. R. Eric Jamieson, John E. Mayer, Jr, Walter E. Pae, Jr, Joseph S. Carey, Stanley W. Dziuban, Jr, Richard H. Felns, Anthony L. Moulton, William C. Nugent, U. Scott Page, Kit V. Arom, Mercedes K. C. Dullum, Fred H. Edwards, T. Bruce Ferguson, Jr, David A. Fullerton, David H. Harpole, Jr, Constantine Mavroudis, Valerie W. Rusch, Richard P. Anderson (Ex-Officio). International Associate Members Bruce E. Keogh, Michael Polonius, Paul T. Sergeant, Hisayoshi Suma