Ann Thorac Surg 2009;88:S43-S62. doi:10.1016/j.athoracsur.2009.05.055
© 2009 The Society of Thoracic Surgeons
Report of STS Quality Measurement Task Force
The Society of Thoracic Surgeons 2008 Cardiac Surgery Risk Models: Part 3—Valve Plus Coronary Artery Bypass Grafting Surgery
David M. Shahian, MDa,*,
Sean M. O'Brien, PhDb,
Giovanni Filardo, PhD, MPHc,
Victor A. Ferraris, MDd,
Constance K. Haan, MDe,
Jeffrey B. Rich, MDf,
Sharon-Lise T. Normand, PhDg,
Elizabeth R. DeLong, PhDb,
Cynthia M. Shewan, PhDh,
Rachel S. Dokholyan, MPHb,
Eric D. Peterson, MD, MPHb,
Fred H. Edwards, MDe,
Richard P. Anderson, MDi,
a Massachusetts General Hospital, Boston, Massachusetts
b Duke Clinical Research Institute, Durham, North Carolina
c Institute for Health Care Research and Improvement, Baylor Health Care System, Dallas, Texas
d University of Kentucky Chandler Medical Center, Division of Cardiovascular and Thoracic Surgery, Lexington, Kentucky
e University of Florida, Division of Cardiothoracic Surgery, Jacksonville, Florida
f Sentara Cardiovascular Research Institute, Norfolk, Virginia
g Department of Health Care Policy, Harvard Medical School, and Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts
h The Society of Thoracic Surgeons, Chicago, Illinois
i Seattle, Washington
Abbreviations and Acronyms AVR = aortic valve replacement; CABG = coronary artery bypass graft surgery; MI = myocardial infarction; MVR = mitral valve replacement; MVRepair = mitral valve repair; NCD = National Adult Cardiac Surgery Database; QMTF = Quality Measurement Task Force; STS = The Society of Thoracic Surgeons
* Address correspondence to Dr Shahian, Massachusetts General Hospital, 55 Fruit St, Boston, MA 02114 (Email: dshahian{at}partners.org).
| Drs Shahian, O'Brien, Filardo, Ferraris, Haan, Rich, Normand, DeLong, Shewan, Peterson, Edwards, Anderson, and Ms Dokholyan, have no conflicts of interest to declare regarding this work.
|
 |
Abstract
|
|---|
Background: Since 1999, The Society of Thoracic Surgeons (STS) has published two risk models that can be used to adjust the results of valve surgery combined with coronary artery bypass graft surgery (CABG). The most recent was developed from data for patients who had surgery between 1994 and 1997 using operative mortality as the only endpoint. Furthermore, this model did not specifically consider mitral valve repair plus CABG, an increasingly common procedure. Consistent with STS policy of periodically updating and improving its risk models, new models for valve surgery combined with CABG have been developed. These models specifically address both perioperative morbidity and mitral valve repair, and they are based on contemporary data.
Methods: The final study population consisted of 101,661 procedures, including aortic valve replacement (AVR) plus CABG, mitral valve replacement (MVR) plus CABG, or mitral valve repair (MVRepair) plus CABG between January 1, 2002, and December 31, 2006. Model outcomes included operative mortality, stroke, deep sternal wound infection, reoperation, prolonged ventilation, renal failure, composite major morbidity or mortality, prolonged postoperative length of stay, and short postoperative length of stay. Candidate variables were screened for frequency of missing data, and imputation techniques were used where appropriate. Stepwise variable selection was employed, supplemented by advice from an expert panel of cardiac surgeons and biostatisticians. Several variables were forced into models to insure face validity (eg, atrial fibrillation for the permanent stroke model, sex for all models). Based on preliminary analyses of the data, a single model was employed for valve plus CABG, with indicator variables for the specific type of procedure. Interaction terms were included to allow for differential impact of predictor variables depending on procedure type. After validating the model in the 40% validation sample, the development and validation samples were then combined, and the final model coefficients were estimated using the overall 100% combined sample. The final logistic regression model was estimated using generalized estimating equations to account for clustering of patients within institutions.
Results: The c-index for mortality prediction for the overall valve plus CABG population was 0.75. Morbidity model c-indices for specific complications (permanent stroke, renal failure, prolonged ventilation > 24 hours, deep sternal wound infection, reoperation for any reason, major morbidity or mortality composite, and prolonged postoperative length of stay) for the overall group of valve plus CABG procedures ranged from 0.622 to 0.724, and calibration was excellent.
Conclusions: New STS risk models have been developed for heart valve surgery combined with CABG. These are the first valve plus CABG models that also include risk prediction for individual major morbidities, composite major morbidity or mortality, and short and prolonged length of stay.
Risk models for cardiac surgery were first developed almost 2 decades ago, and most of these early models focused on isolated coronary artery bypass graft surgery (CABG) [1–4]. The results of this frequently performed surgical procedure have often been used as the sole marker to assess the quality of care delivered by cardiac surgical programs. Risk-adjusted results for CABG have been used for hospital and regional quality improvement initiatives, public reporting, pay for performance reimbursement programs, decision support, patient counseling, and clinical research. Earlier models focused primarily on mortality prediction, but subsequent models have been developed for both risk-adjusted morbidity and length of stay [5].
The other commonly performed category of cardiac surgery consists of operations on the heart valves, either alone or in combination with CABG. Relative to isolated CABG procedures, which are declining in frequency, the proportion of valve cases is steadily increasing. To better assess the overall performance of cardiac surgery programs, to discern the factors that are most significantly related to patient outcomes, and to aid in physician and patient decision-making, risk models have now also been developed for heart valve surgery [6–18].
Unlike risk models for isolated CABG, a relatively standardized procedure, valve surgery encompasses a much more diverse group of operations. There are four cardiac valves, and they may malfunction in a number of quite different ways (eg, stenosis, regurgitation, infection, and so forth). The valves may be repaired or replaced with a wide range of techniques and prosthetics. In some cases, procedures may be performed on multiple valves, or the valve procedure may be combined with CABG.
Given the heterogeneity of heart valve surgery, it is not surprising that a variety of risk-modeling techniques has been applied. At one extreme, the European System for Cardiac Operative Risk Evaluation (EuroSCORE) algorithm, developed by a European consortium, groups all cardiac operations together in a single risk model with indicator variables included to account for valve procedures [14, 18]. Although this approach is simple and easy to apply, recent studies by van Gameren and associates [19] have suggested that a dedicated valve risk model may have better discrimination and calibration than the EuroSCORE algorithm when applied to valve surgery patients. Combined models for aortic and mitral valve procedures with or without CABG have been developed by Jin and colleagues [12] and by Ambler and associates [13]. The 2001 valve models developed by The Society of Thoracic Surgeons (STS) [6] consisted of one model for all isolated valve procedures and one model for valve procedures combined with CABG, and a 2007 risk model derived from the New York Cardiac Surgery Reporting System used a similar stratification [8].
Unified valve models reflect the fact that many risk factors are common to both aortic and mitral valve surgery. They offer simplicity, and they also permit larger sample sizes for development and validation [12]. However, there are significant differences between aortic and mitral valvular disease in both pathophysiology and outcomes, and both also differ substantially from isolated CABG [11]. Some investigators advocate separate aortic and mitral valve models to have more homogeneous patient populations. Examples include models developed by STS, the New York Cardiac Surgery Reporting System, and the Northern New England Cardiovascular Disease Study Group [7, 9, 10]. Some of these models have been developed solely for isolated valve replacement, some have included CABG as a separate predictor variable in the isolated valve model, and some models have focused specifically on valve plus CABG. All these decisions involve a tradeoff—the more homogeneous the study group, the fewer patients are available for model development and validation [12].
Because of the large number of valve surgery patients available for analysis in the STS National Adult Cardiac Surgery Database (NCD), our approach has favored separate models for valve plus CABG versus isolated valve surgery. The STS Quality Measurement Task Force (QMTF) presumes that when adequate numbers of patients are available for study, relatively homogeneous operative categories result in more accurate risk prediction. Furthermore, recent studies by van Gameren and colleagues [19] suggest that the valve plus CABG group may be the most difficult to model accurately, thus meriting its own algorithm.
Several new features were added to the 2008 valve plus CABG models described in this report. First, recognizing that mitral valve repair is often different in both etiology and outcomes than replacement, the QMTF has included interactions between surgery type and several key predictor variables. Fitting a single model with several such interactions is useful. It allows for pooling information across related groups of valve procedures without making an a priori assumption that the effect of key risk factors is constant across these groups. Finally, new models have been developed for specific major complications of each valve plus CABG procedure, as well as for composite morbidity, mortality, and for both short and prolonged postoperative length of stay.
The authors of this report are members of the STS QMTF who were involved in this risk model development project.
|
Study Population and Endpoints
|
|---|
Our general approaches to variable selection and risk model development have been described in the companion articles on isolated CABG (Part 1) and isolated valve surgery (Part 2). Details specific to the valve plus CABG models are included in this report.
Study Population
The study population for this analysis consisted of single aortic or mitral valve surgical procedures combined with CABG performed on adult patients between January 1, 2002, and December 31, 2006. Only the following procedures were included: (1) isolated aortic valve replacement (AVR) plus CABG; (2) isolated mitral valve replacement (MVR) plus CABG; and (3) isolated mitral valve repair (MVRepair) plus CABG.
Because of the relatively small number of pulmonic, tricuspid, multiple valve procedures, and aortic repairs, these cases were not included in the current models. Patients undergoing isolated valve surgery without CABG were excluded from the current analysis, but these cases are the focus of a separate model described in Part 2 of this three-part series. Patients with missing sex data (n = 17) were excluded because these patients are not allowed in the analysis dataset used for creating STS database participant feedback reports. Patients on dialysis preoperatively (n = 2,443) were excluded when developing the risk model for prediction of postoperative renal failure. The final study population comprised 101,661 patient operations (66,074 AVR plus CABG; 13,663 MVR plus CABG; and 21,924 MVRepair plus CABG) from 814 STS NCD participating groups.
Characteristics of the study population are summarized in Table 1.
Training and Validation Samples
The study population was randomly divided into a 60% training (development) sample and a 40% test (validation) sample. The development sample was used to identify predictor variables and estimate model coefficients. Data from the validation sample were used to assess model fit, discrimination, and calibration. After choosing variables and assessing model fit, the development and validation samples were subsequently combined, and the final model coefficients were estimated using the combined (development plus validation) data.
Endpoints
In developing the valve plus CABG risk models, we used the same nine endpoints that were analyzed in the STS isolated CABG (Part 1) and the STS isolated valve (Part 2) models. Morbidities in all three models are recorded only in-hospital, in contrast to the operative mortality endpoint defined below (although beginning with version 2.61, sternal infection will be recorded at 30 days): (1) operative mortality: death during the same hospitalization as surgery, regardless of timing or within 30 days of surgery regardless of venue; (2) permanent stroke (CVA): a central neurologic deficit persisting longer than 72 hours; (3) renal failure: a new requirement for dialysis or an increase of the serum creatinine to more than 2.0 mg/dL and double the most recent preoperative creatinine level; (4) prolonged ventilation (> 24 hours); (5) deep sternal wound infection; (6) reoperation for any reason; (7) major morbidity or mortality, a composite defined as the occurrence of any of the above endpoints; (8) prolonged postoperative length of stay (PLOS): length of stay (LOS) more than 14 days (alive or dead); and (4) short postoperative length of stay (SLOS): LOS less than 6 days and patient alive at discharge.
Endpoint frequencies in the study population are presented in Table 2.
|
Separate Versus Combined Models
|
|---|
Given the variety of approaches used in previous models by STS and other developers, we investigated the option of developing separate models for the AVR plus CABG and MVR plus CABG populations, and we also studied how best to subdivide the mitral plus CABG population into repair versus replacement. Although we had a large study population available, many of the individual outcomes were relatively rare. We were concerned that the number of events would be too small to permit reliable estimation of the model coefficients in separate models for each valve. Thus, in theory, the development of separate custom models for each valve type could be inferior to a single combined model because the custom models would have a smaller sample size and hence larger variance.
As described in detail in Part 2 of this series (isolated valve surgery), we performed preliminary empirical analyses to compare two alternative strategies (separate versus combined AVR plus CABG and MVR/Repair plus CABG) for developing these risk models. We first developed separate models for the three subpopulations (AVR plus CABG, MVR plus CABG, and MVRepair plus CABG), then modeled all three subpopulations together in a single model. In the latter approach, we included several interaction terms to allow the effect of certain risk factors to differ across the specific valve subpopulations. These strategies were used to develop risk models for operative mortality and permanent stroke, using a 60% development sample and a separate 40% validation sample. The performance of the combined model was then assessed separately within each subpopulation and compared to the model that was developed specifically for that subpopulation. In the case of mortality, the combined model had better discrimination (larger c-index) than the corresponding custom model in each of the three subpopulations (AVR plus CABG, MVR plus CABG, MVRepair plus CABG). For stroke, the combined model had better discrimination in two of the three populations (all except AVR plus CABG). Finally, when explained variation was quantified by the generalized R2 index of Nagelkerke [20], the combined model had greater explained variation than the custom model in each subpopulation for each endpoint. These results provide empirical support for the use of a single model with several interactions, which allows pooling of information across valve groups without assuming that the effect of risk factors is constant.
|
Selection of Candidate Predictor Variables
|
|---|
The candidate variables for the STS valve plus CABG models were identical to those in the STS isolated valve models, described in Part 2 of this series. They differed from the isolated CABG model variables in the following specific areas: (1) Percutaneous coronary intervention (PCI) occurring 6 hours or less before surgery was present in only 315 patients (0.3%) in the valve plus CABG study population, and was not included as a candidate variable. (2) Infectious endocarditis was not included in the isolated CABG model but was considered for the valve plus CABG model. Although this risk factor was rarely present (0.8% active endocarditis) in the overall valve plus CABG population, it was included for consistency with the isolated valve model. Active endocarditis was present in 2.6% of patients undergoing mitral replacement plus CABG. (3) Mitral stenosis was rarely present among isolated CABG patients (0.35%). However, it was not uncommon (4.9%) among patients undergoing valve plus CABG surgery and was included as a candidate variable. It was present in 17.3% of mitral replacements and 5.0% of mitral repairs.
An indicator for valve procedure (AVR, MVR, MVRepair) was included in the combined valve plus CABG model, as previously noted.
|
Missing Data
|
|---|
Missing data are uncommon in the STS NCD, with a frequency of less than 1% missing for most variables. Model variables with more than 1% missing were ejection fraction (5.9%), New York Heart Association functional class (3.8%), tricuspid insufficiency (2.6%), aortic insufficiency (2.1%), mitral insufficiency (1.5%), and creatinine/dialysis (1.2%).
To make full use of the available data, binary risk factors were modeled as yes versus no or missing. Thus, missing values were analyzed as if the endpoint did not occur. Missing data on categorical variables were imputed to the lowest risk value, which, in most instances, was the mode. Missing data on continuous variables were imputed to the conditional median. For ejection fraction, we conditioned on congestive heart failure and sex. For body surface area, we conditioned on sex. For serum creatinine, we conditioned on renal failure.
Although multiple imputation is generally preferred on statistical grounds [21], we chose single imputation for this analysis based largely on practical considerations, including computational intensity. Furthermore, the fraction of missing data was small, and single and multiple imputation would give similar results. Finally, multiple imputation is primarily used for calculating appropriate standard error estimates, but an adjustment to the standard errors would not impact our study results or the published risk algorithms. In a separate sensitivity analysis, we compared predicted risk estimates from our final models to risk estimates that were derived from analogous models using multiple instead of single imputation. For each endpoint, the relative difference in predicted risk was less than 6% (eg, an absolute difference of 5.0% versus 5.3%) for all patients in the development and validation samples, and it was less than 2% (eg, an absolute difference of 5.0% versus 5.1%) for 99% of patients. A summary of these analyses including regression coefficients and covariance matrices is available at www.sts.org/riskmodels.
|
Final Variable Selection Procedure
|
|---|
Variables were initially selected using an automated stepwise model selection algorithm. The stepwise procedure began with a model that included all of the candidate variables except for interaction terms. Age, body surface area, and month of surgery were forced into each model. As in the isolated CABG and isolated valve models described in Parts 1 and 2 of this series, month of surgery was used only to adjust for time trends in the frequency of adverse outcomes over the 5-year study period. We adjusted for this to reduce potential confounding by time trends when estimating regression coefficients for the variables that are of primary interest (ie, patient preoperative risk factors—see example in Part 1). Surgery date was categorized into 6-month intervals and modeled as a linear trend across the ordinal categories. Surgery date is not included in the final risk prediction algorithm, and a patient's predicted risk does not depend on it. The published intercept parameter has been adjusted to incorporate the time trend, and this adjusted intercept reflects the baseline risk for a reference period of July to December 2006.
Other variables were selected in a stepwise fashion using a significance criterion of 0.05 for entry and removal. Ordinal categorical variables were initially coded such that removing an indicator variable caused a category to be combined with the lowest risk category (the reference group). In the case of myocardial infarction (MI), there were two outcomes (permanent stroke, prolonged length of stay) in which "MI 1 to 21 days" was retained but "MI less than 24 hours" was removed. For these two cases, the two MI categories were replaced by the single category "MI 21 days or less." The stepwise procedure was performed separately for each endpoint. Multiple interaction terms consisting of predictor variable and surgery type were also evaluated, and two additional interaction terms (age by reoperation and age by emergent status) were forced into the models (see Tables 3 and 5).
The results of this initial selection process were then reviewed by surgeon members of the QMTF for face validity and consistency with previous STS or other valve models: (1) preoperative atrial fibrillation was forced into the model for permanent stroke; (2) an indicator variable for dialysis was forced into any model that included creatinine (this did not apply to the renal failure model, as patients with preoperative dialysis were excluded); (3) sex was forced into all models; and (4) each variable that interacted with surgery group was also included as a main effect.
After validating the model in the 40% validation sample, the development and validation samples were then rejoined, and the final model coefficients were estimated using the overall 100% combined sample. The final logistic regression model was estimated using generalized estimating equations with empirical (sandwich) standard error estimates to account for clustering of patients within institutions [22]. An independence working correlation matrix was used to apply the generalized estimating equations. With this approach, the estimated regression coefficients were identical to those obtained using ordinary logistic regression, but the standard errors were adjusted to account for the clustered data structure.
|
Results
|
|---|
Risk Factors, Outcomes, and Predictor Variables
Table 1 presents the distribution of risk factors and endpoints in the overall 2002 to 2006 study population. Because there are three valve plus CABG categories, space limitations prevent display of the bivariate relationships for each predictor variable, endpoint, and valve plus CABG group. These are available upon request from STS.
Table 2 summarizes the overall frequency of adverse outcomes as well as the outcomes for the three major valve groups. Table 3 lists the candidate predictor variables and their coding schemes.
Assessment of Model Fit and Discrimination
The Hosmer-Lemeshow test was not employed to assess overall calibration. Large sample sizes make a significant p value almost inevitable, as all risk models are only approximations of reality [23]. Rather, we assessed calibration graphically by plotting observed versus predicted event rates within deciles of predicted risk in the development and validation samples (Fig 1). These plots were constructed for the overall sample and for subgroups based on surgery type (AVR plus CABG, MVR plus CABG, MVRepair plus CABG); age (< 60, 60 to 79, 
80 years); sex (male, female); diabetes mellitus (yes/no); status (elective, nonelective); and ejection fraction (
40, > 40). Because of space constraints, only the overall sample results in the validation sample are presented. Additional results are available at www.sts.org/riskmodels.
In general, the models were well calibrated in the validation sample. The average absolute difference between observed versus predicted event rates across the decile categories ranged from 0.1% for deep sternal wound infection to 0.96% for prolonged length of stay. There was a very slight tendency for the models to overpredict risk in the highest decile. Although perfect prediction would be ideal, a slight overprediction implies that the model will give adequate credit to surgeons who take on patients with several model risk factors.
Discrimination was assessed by determining the c-statistic, also known as the area under the receiver operating characteristic (ROC) curve. Table 4
presents the discrimination of the various models. In the validation sample, the c-index of the overall valve plus CABG operative mortality model was 0.750, and the c-indices of the morbidity models ranged from 0.617 for reoperation to 0.724 for renal failure and short length of stay.
Odds Ratios
Table 5
presents the odds ratios and 95% confidence intervals (CI) derived from these models. "Not applicable" indicates that those predictors were not included in a particular risk model.
Odds ratios that do not interact with surgery type are summarized in Part A of Table 5. Several variables interact with surgery type, and the odds ratios for these variables differ for some of the endpoints depending on the specific type of surgery, as summarized in Tables 5B, C, and D (AVR plus CABG, MVR plus CABG, MVRepair plus CABG). For example, in the model for prolonged length of stay, the odds ratio for active endocarditis is 1.81 (95% CI: 1.41 to 2.32) for AVR plus CABG; 2.08 (95% CI: 1.62 to 2.67) for MVR plus CABG; and 2.98 (95% CI: 1.86 to 4.77) for MVRepair plus CABG.
Final Model Intercept and Coefficients
The algorithms for calculating predicted risk values, including the intercepts and regression coefficients, are presented in the Appendix.
|
Limitations
|
|---|
The limitations of the STS valve plus CABG models are similar to those discussed in Part 1 of this series.
|
Conclusion
|
|---|
A new STS model has been developed for valve surgery combined with CABG. This model includes specific indicator variables for each major type of valve plus CABG procedure (AVR plus CABG, MVR plus CABG, MVRepair plus CABG). Models have been developed for operative mortality, individual morbidity endpoints, a composite morbidity or mortality endpoint, and short and prolonged postoperative length of stay. Overall model performance is excellent.
|
Appendix
|
|---|
Regression Coefficients and Variable Definitions for STS 2008 Valve Plus CABG Models
For each endpoint, the formula for calculating a patient's predicted risk of the endpoint has the form:
where x
1, x
2, ... , xn denote patient preoperative risk factors (eg, quantitative variables such as age, and comorbidities coded as 1=present, 0=absent); and β0, β1, ... , βn denote regression coefficients (numerical constants). Regression coefficients for each endpoint are presented in Appendix Table 1. The variables x
1, x
2, ... , xn are the same for each endpoint and are defined in Appendix Table 2. The regression coefficient for the time trend is not presented. Instead, the intercept has been adjusted to incorporate the time trend. This adjusted intercept reflects the baseline risk for a reference period of July–December 2006.
|
Footnotes
|
|---|
This author is deceased. Former Chair, Quality, Research and Patient Safety Council, The Society of Thoracic Surgeons, Chicago, IL. 
|
References
|
|---|
- Hannan EL, Kilburn Jr H, O'Donnell JF, Lukacik G, Shields EP. Adult open heart surgery in New York State. An analysis of risk factors and hospital mortality rates. JAMA 1990;264:2768-2774.[Abstract/Free Full Text]
- Edwards FH, Clark RE, Schwartz M. Coronary artery bypass grafting: the Society of Thoracic Surgeons National Database experience Ann Thorac Surg 1994;57:12-19.[Abstract]
- O'Connor GT, Plume SK, Olmstead EM, et al. Multivariate prediction of in-hospital mortality associated with coronary artery bypass graft surgery. Northern New England Cardiovascular Disease Study Group. Circulation 1992;85:2110-2118.[Abstract/Free Full Text]
- Shahian DM, Blackstone EH, Edwards FH, et al. Cardiac surgery risk models: a position article Ann Thorac Surg 2004;78:1868-1877.[Abstract/Free Full Text]
- Shroyer AL, Coombs LP, Peterson ED, et al. The Society of Thoracic Surgeons: 30-day operative mortality and morbidity risk models Ann Thorac Surg 2003;75:1856-1864.[Abstract/Free Full Text]
- Edwards FH, Peterson ED, Coombs LP, et al. Prediction of operative mortality after valve replacement surgery J Am Coll Cardiol 2001;37:885-892.[Abstract/Free Full Text]
- Jamieson WR, Edwards FH, Schwartz M, Bero JW, Clark RE, Grover FL. Risk stratification for cardiac valve replacement. National Cardiac Surgery Database. Database Committee of The Society of Thoracic Surgeons Ann Thorac Surg 1999;67:943-951.[Abstract/Free Full Text]
- Hannan EL, Wu C, Bennett EV, et al. Risk index for predicting in-hospital mortality for cardiac valve surgery Ann Thorac Surg 2007;83:921-929.[Abstract/Free Full Text]
- Hannan EL, Racz MJ, Jones RH, et al. Predictors of mortality for patients undergoing cardiac valve replacements in New York State Ann Thorac Surg 2000;70:1212-1218.[Abstract/Free Full Text]
- Nowicki ER, Birkmeyer NJ, Weintraub RW, et al. Multivariable prediction of in-hospital mortality associated with aortic and mitral valve surgery in Northern New England Ann Thorac Surg 2004;77:1966-1977.[Abstract/Free Full Text]
- Nowicki ER. What is the future of mortality prediction models in heart valve surgery? Ann Thorac Surg 2005;80:396-398.[Free Full Text]
- Jin R, Grunkemeier GL, Starr A. Validation and refinement of mortality risk models for heart valve surgery Ann Thorac Surg 2005;80:471-479.[Abstract/Free Full Text]
- Ambler G, Omar RZ, Royston P, Kinsman R, Keogh BE, Taylor KM. Generic, simple risk stratification model for heart valve surgery Circulation 2005;112:224-231.[Abstract/Free Full Text]
- 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.[Abstract/Free Full Text]
- Gardner SC, Grunwald GK, Rumsfeld JS, et al. Comparison of short-term mortality risk factors for valve replacement versus coronary artery bypass graft surgery Ann Thorac Surg 2004;77:549-556.[Abstract/Free Full Text]
- Grover FL, Edwards FH. Similarity between the STS and New York State databases for valvular heart disease Ann Thorac Surg 2000;70:1143-1144.[Free Full Text]
- Rankin JS, Hammill BG, Ferguson Jr TB, et al. Determinants of operative mortality in valvular heart surgery J Thorac Cardiovasc Surg 2006;131:547-557.[Abstract/Free Full Text]
- Roques F, Nashef SA, Michel P. Risk factors for early mortality after valve surgery in Europe in the 1990s: lessons from the EuroSCORE pilot program J Heart Valve Dis 2001;10:572-577.[Medline]
- van Gameren M, Kappetein AP, Steyerberg EW, et al. Do we need separate risk stratification models for hospital mortality after heart valve surgery? Ann Thorac Surg 2008;85:921-930.[Abstract/Free Full Text]
- Nagelkerke NJD. A note on a general definition of the coefficient of determination Biometrika 1991;78:691-692.[Abstract/Free Full Text]
- Little RJA, Rubin DB. Statistical analysis with missing data2nd ed.. Hoboken, NJ: Wiley-Interscience; 2002.
- Liang KY, Zeger SL. Longitudinal data-analysis using generalized linear-models Biometrika 1986;73:13-22.[Abstract/Free Full Text]
- Marcin JP, Romano PS. Size matters to a model's fit Crit Care Med 2007;35:2212-2213.[Medline]
This article has been cited by other articles:
|
|
|
|
 
S. L. Rohde, R. A. Baker, P. J. Tully, S. Graham, H. Cullen, and J. L. Knight
Preoperative and Intraoperative Factors Associated With Long-Term Survival in Octogenarian Cardiac Surgery Patients.
Ann. Thorac. Surg.,
January 1, 2010;
89(1):
105 - 111.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
