HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Add to Personal Folders Download to citation manager Author home page(s): Thomas L. Higgins Floyd D. Loop Delos M. Cosgrove, III Permission Requests Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Higgins, T. L. Articles by Cosgrove, D. M. Search for Related Content PubMed PubMed Citation Articles by Higgins, T. L. Articles by Cosgrove, D. M., III

Ann Thorac Surg 1997;64:1050-1058
© 1997 The Society of Thoracic Surgeons

---

Original Article: Cardiovascular

ICU Admission Score for Predicting Morbidity and Mortality Risk After Coronary Artery Bypass Grafting

Departments of Cardiothoracic Anesthesia, Thoracic and Cardiovascular Surgery, and Biostatistics and Epidemiology, The Cleveland Clinic Foundation, Cleveland, Ohio

Accepted for publication April 7, 1997.

	Abstract

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 
Background. This study was performed to develop an intensive care unit (ICU) admission risk score based on preoperative condition and intraoperative events. This score provides a tool with which to judge the effects of ICU quality of care on outcome.

Methods. Data were collected prospectively on 4,918 patients (study group n = 2,793 and a validation data set n = 2,125) undergoing coronary artery bypass grafting alone or combined with a valve or carotid procedure between January 1, 1993, and March 31, 1995. Data were analyzed by univariate and multiple logistic regression with the end points of hospital mortality and serious ICU morbidity (stroke, low cardiac output, myocardial infarction, prolonged ventilation, serious infection, renal failure, or death).

Results. Eight risk factors predicted hospital mortality at ICU admission, and these factors and five others predicted morbidity. A clinical score, weighted equally for morbidity and mortality, was developed. All models fit according to the Hosmer-Lemeshow goodness-of-fit test. This score applies equally well to patients undergoing isolated coronary artery bypass grafting.

Conclusions. This model is complementary to our previously reported preoperative model, allowing the process of ICU care to be measured independent of the operative care. Sequential scoring also allows updated prognoses at different points in the continuum of care.

	Introduction

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 
Outcome after coronary artery bypass grafting (CABG) is determined by the preoperative status of the patient, as well as by technical factors in the operating room and intensive care unit (ICU). Preoperative risk stratification in the cardiac surgical population consistently identifies advanced age, emergency status, left ventricular dysfunction, a previous heart operation, diabetes, female sex, small body size, and renal disease as factors that affect outcome [1–5].

However, morbidity and mortality after CABG are also influenced by surgical and anesthetic techniques [6, 7], including time on cardiopulmonary bypass [7], completeness of revascularization [8], efficacy of myocardial protection [9], hemodynamic management [10], and unforeseen events in the operating room. The patient's prognosis at arrival to the ICU may differ from the preoperative prognosis. In the present study, we evaluated the relative contribution of preoperative condition, operating room events, and physiologic measurements at ICU admission to outcome and describe a risk stratification score based on a patient's status at ICU admission.

	Material and Methods

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 
We prospectively collected data on 4,918 consecutive adult patients undergoing CABG between January 1, 1993, and March 31, 1995. Preoperative data were collected by resident physicians, and ICU admission data were collected by trained physician assistants. Data collected during the first 15 months (n = 2,793) were used to develop the model, and data collected during the next 12 months (n = 2,125) were used to validate the model. Patients undergoing repeat operation, simultaneous carotid endarterectomy, or repair or replacement of the mitral or aortic valve (or both) were included in the study, because these "combined" CABG patients represent a large (27%) and growing part of our surgical population. A separate analysis was also conducted on the subset of patients (n = 2,035; 73% of the developmental data set) undergoing isolated coronary revascularization. Six patients underwent both a primary and a repeat revascularization procedure during the study period, but only the first operation was considered for analysis. The reliability of data collection was assessed by comparing the data with chart review of a random subset. This review showed a relevant agreement of 98% to 100% of patients for each variable used in the final models.

We chose more than 100 risk factors from the literature, our clinical experience, and our own work [1, 8, 11] and examined the influence of each on outcome at discharge. Emergency cases were those characterized by unstable angina, unstable hemodynamics, or ischemic valvular dysfunction that could not be controlled medically. Left ventricular function was categorized as normal (ejection fraction [EF] 0.60], mildly impaired (EF 0.50 to 0.59), mildly to moderately impaired (EF 0.46 to 0.49), moderately impaired (EF 0.41 to 0.45), moderately to severely impaired (EF 0.36 to 0.40), or severely impaired (EF 0.35) based on the cardiologist's evaluation of the ventriculogram at cardiac catheterization or by echocardiography if a ventriculogram was not recorded. Diabetes or chronic obstructive pulmonary disease was diagnosed only if the patient had a history of the condition and was maintained on appropriate medications. Valvular pathology was evaluated by stenosis versus regurgitation and repair versus replacement. Cardiopulmonary bypass time was the total of all bypass runs if a second or subsequent period of bypass was conducted. The need for a second or greater bypass run was also considered as a separate factor in the analysis.

Morbidity was defined as the presence of one or more of the following during hospitalization: (1) cardiac complication: low cardiac output (sustained cardiac index 1.8 L • min^-1 • m^-2 despite adequate preload and inotropic support) or myocardial infarction documented by electrocardiography and enzyme criteria and that required the use of an intraaortic balloon pump (IABP) or a ventricular assist device (IABP and assist devices were resorted to when vasoactive drug support failed to achieve a cardiac index >2 L/m or mean arterial pressure >60 mm Hg without significant side effects (eg, persistent ischemia, arrhythmias); (2) prolonged ventilatory support: mechanical ventilation for 72 hours or more; (3) central nervous system complication: focal brain lesion confirmed by clinical findings, computed tomographic scan, or both, or diffuse encephalopathy with more than 24 hours of severely altered mental status, failure to awaken postoperatively, or both; (4) renal failure: urine output of <400 mL/24 hours, institution of dialysis, or both; (5) serious infection: culture-proven pneumonia (blood or sputum), mediastinitis, wound infection, sepsis syndrome, or septic shock; or (6) death, because early hospital death might preclude the diagnosis of other morbidities. Mortality included all deaths during the hospitalization for the operation, regardless of length of stay.

	Development of the Logistic Regression Model

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 
The association of each factor with morbidity and mortality was evaluated using a ² test or Fisher's exact test for categoric variables and Student's t test for continuous variables. Because of the large number of variables analyzed, only the most significant are presented in Tables 1 and 2. Continuous variables significantly related to outcome were plotted using a locally weighted smoothing scatterplot technique [12] to assess linearity on the logit scale. If the relationship to risk was linear, the factor was entered into the model as a continuous variable. If the relationship to risk was not linear, appropriate cut points were determined from the locally weighted smoothing scatterplots to divide the distribution of values into categories. Locally weighted smoothing scatterplots were also used to divide variables that were continuous in the logistic models into categories for use in a clinical risk stratification score (see below).

Two models were developed: one each for mortality and morbidity. The number of terms allowed in a logistic model was limited to 10% of the number of outcome events, with only the most significant factors included to avoid overfitting the models [13]. The goodness of fit of each final logistic model was evaluated using the Hosmer-Lemeshow ² statistic [14]. In the event of a low number of expected events, the lowest deciles were combined for statistical analysis, with appropriate reduction in the degrees of freedom. Receiver operating characteristic curves were used to generate a C-statistic (area under the curve) to measure and compare the accuracy of the models [15, 16]. C-statistic values closer to 1.0 indicate better discrimination by the model.

	Development of the Clinical Model and Risk Stratification Score

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 
A clinical model was created by first categorizing continuous predictors in the logistic regression models by using the locally weighted smoothing scatterplot procedure to identify cut points and then refitting the logistic models using all predictors as categoric variables. For example, for cardiopulmonary bypass time the cut point of 160 minutes was based on the much higher mortality above this value in the locally weighted smoothing scatterplots. Cut points were not based on being one or two standard deviations above the mean. Then a numeric score was calculated by summing the regression coefficient estimates of the logistic models for morbidity and mortality, multiplying by 2, and then rounding to the nearest integer to produce a clinical model that equally weights mortality and morbidity. This clinical model was evaluated for its goodness of fit and discrimination on the developmental data set, the subset of patients who underwent isolated coronary bypass without other simultaneous procedures, and the validation data set. Individual levels of morbidity and mortality risk were then combined into five levels of risk to facilitate the application of the scoring system when smaller numbers of patients are available, as might be seen in quarterly analysis or at institutions with smaller surgical volumes. The observed values in the validation set were then compared with the 95% confidence intervals of the predicted values generated from the developmental data set.

Most statistical analyses were performed with SAS version 6.0 (Statistical Analysis Systems, Cary, NC). The receiver operating characteristic analyses used programs developed by Metz and associates [15, 16].

	Results

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 
Univariate risk factors for morbidity and mortality are presented in Table 1 (categoric variables) and Table 2 (continuous variables). Selected variables that were not statistically significant by univariate analysis included low urine output in the operating room, history of chronic obstructive pulmonary disease requiring medication, history of renal failure requiring dialysis, chronic hypertension, race other than white, prior carotid operation, prior lung operation, and preoperative use of antiarrhythmics or calcium channel antagonists. Ventricular aneurysectomy or carotid operation in addition to CABG also did not achieve statistical significance.

Fifty-eight variables qualified for entry into the multiple logistic regression analysis. The logistic regression models for mortality and morbidity are presented in Tables 3 and 4. These models were fit on 2,440 patients having complete data on these variables. To avoid overfitting, the mortality model was restricted to the first eight significant factors, because there were only 78 deaths. Significant factors that would have been entered into the model without this restriction were central venous pressure and mean arterial pressure values at ICU admission, and a history of congestive heart failure. No such limits applied to the morbidity model, which has 13 factors, including all eight factors in the mortality model, plus central venous pressure and alveolar–arterial gradient at ICU admission, preoperative serum creatinine level, reoperation status, and body surface area.

View larger version (24K): [in this window] [in a new window]   Fig 1. . Mortality by intensive care unit (ICU) admission risk group. Risk is estimated by summing the individual elements in Table 5. Results are displayed for the developmental set (1993–94) and the prospective validation set (1994–95). All observed (validation) values fall within the 95% confidence intervals established from the developmental set.

View larger version (28K): [in this window] [in a new window]   Fig 2. . Morbidity by intensive care unit (ICU) admission risk group. See legend for Figure 1.

The accuracy and goodness of fit of both the logistic and clinical models were very similar indicating that the clinical severity score, although not a probability, can be useful in categorizing risks of mortality and morbidity in CABG patients.

	Comment

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 
Preoperatively, patients undergoing open heart operations should be expected to be concerned about the possible outcome of the operation, including morbidity and mortality. Payors and health providers need similar information. Several preoperative risk status models have been developed to predict outcome after coronary artery operations [1–5]. These models all predict mortality, and possibly morbidity, with good accuracy but they also have shortcomings [17]. The patient's physiologic response to the operation and anesthesia cannot always be predicted from preoperative information. Unexpected intraoperative events, management of anesthesia [10], efficacy of myocardial preservation [9], completeness of myocardial revascularization, and uneventful reoperation can neutralize or amplify preoperative risk. All these factors can affect the prognosis. Thus reevaluation and assessment of a patient's condition at ICU admission complements the preoperative evaluation and allows for a proactive ICU management plan and better prediction of outcome.

Previous work from our group [1] identified 13 preoperative risk factors for morbidity and mortality after CABG. Four of these factors are also significant in the ICU admission models: preoperative serum creatinine level, age, prior cardiac operation, and a history of vascular disease.

Data on serum albumin level, a significant factor in the ICU model, were not available for consideration in our preoperative risk model (Table 6). However, preoperative hematocrit, used previously as a measure of chronic disease, appears to account for some of the same information as albumin (Pearson correlation coefficient between hematocrit and serum albumin = 0.43). If serum albumin level is not considered as a variable, hematocrit would appear as a fifth factor common to the preoperative and ICU admission models.

This study also confirmed that intraoperative events such as cardiopulmonary bypass time and the use of an IABP are as important as preoperative risk factors in predicting outcome. Cardiopulmonary bypass time was analyzed as a continuous factor in the logistic model and it showed increase in mortality and morbidity with increasing time. We use 160 minutes as the cutpoint, which is 1.58 times the median time (101 minutes). The 160 minutes represents the 89.1 percentile and was chosen based on locally weighted smoothing scatterplot curve analysis demonstrating a sharp increase in poor outcome above this value.

Emergency procedures are significant preoperative predictors of poor outcome, but were not significant in the ICU admission logistic regression models. Emergency patients are more likely to have other risk factors on ICU arrival such as low cardiac index, decreased serum albumin, higher alveolar–arterial oxygen gradient, elevated central venous pressure, and tachycardia. These were included in the logistic and clinical models. It is clear that scoring based on physiologic variables is more patient specific and is preferable to the term "emergency," a definition that might vary between institutions. Several other physiologic measurements at ICU admission are predictors of outcome in this study, namely alveolar–arterial oxygen gradient, heart rate, cardiac index, central venous pressure, and arterial bicarbonate.

The use of physiologic measurements to predict outcome was pioneered by the APACHE prognostic system [21] and successfully predicts mortality, resource use, and ICU length of stay in CABG patients [22]. The present study confirms the importance of physiologic measurements at ICU admission as outcome predictors.

Some of the physiologic variables (such as body temperature and hematocrit) scored by APACHE-III, a benchmark for determining outcome in noncardiac intensive care units, may have a different relationship with outcome in CABG patients [21]. During cardiopulmonary bypass most patients are cooled systematically and their temperature at ICU admission can be less than 37°C. This does not reflect a pathologic status. Hemodilution is now a routine practice in cardiac surgery. Immediate postoperative hematocrit of 24% or greater is generally considered adequate. Experimental studies by our group [23] and our clinical experience demonstrated that hematocrit greater than 28% is necessary to maintain hemodynamic stability in patients with impaired ventricular function. In the present study, patients with hematocrit levels of 24% or greater were more likely to require transfusion in the operating room (p > 0.001). Therefore low hematocrit values (about 24%) are not necessarily a marker of increased risk in CABG patients, although it has predictive ability in other ICU patients.

This ICU risk stratification score relates to both morbidity and mortality as end points. Statistically, the contribution of patient management to outcome is more evident when the end point occurs more frequently. In this study the incidence of morbidity (10.4%) is higher than the incidence of overall mortality (3.1%). Therefore, morbidity better reflects the ICU and hospital length of stay and cost. Although many of the risk factors predict both morbidity and mortality, the logistic regression models demonstrate that some of the risk factors for mortality and morbidity were different. The clinical model included the most significant factors that encompassed both outcomes, and can be a useful tool for benchmark comparisons.

This ICU admission model is part of a sequential evaluation of the probability of morbidity or mortality. In the present study, we investigated the sequential relationship of how the prognosis of mortality changes during the perioperative period (Fig 3). In patients with preoperative risk scores of 6 or less (n = 2,118) the mortality was 2.3%; of these patients 97% had an ICU admission score of 14 or less with a mortality rate of 1.7%. The other 3% had an ICU admission score greater than 14 with a mortality of 19.4%.

View larger version (45K): [in this window] [in a new window]   Fig 3. . Sequential analysis of mortality risk in cardiac surgical patients. Use of preoperative and intensive care unit (ICU) risk stratification scores for sequential analysis of mortality risk. A preoperative score of 7 was used as the cut point between a low and high risk. An ICU admission score of 14 was used as the cut point to determine the level of risk.

This sequential analysis demonstrates that the ICU admission scoring model identifies patients with high risk of mortality based on postoperative status who were not identified by preoperative scoring alone. Thus discrimination of outcome based on ICU admission stratification is superior to stratification by preoperative status alone. This is borne out by comparison of receiver operating characteristic curves. In preoperative models for cardiac surgery, the area under the receiver operating characteristic curve (the C-statistic) for mortality prediction ranges from 0.74 to 0.83 [1, 4, 5, 17]. The comparable C-statistic for the ICU admission clinical model is 0.87. Similar results (a C-statistic of 0.85) have been reported for APACHE-III in the cardiac surgical setting [22]. However, caution should be exercised in comparing C-statistics across different populations in different studies, because C-statistics in independent, prospective validation studies are often lower [17]. Also, a prospective multicenter study for validation of this ICU model is necessary to establish a benchmark. It will also enhance the management strategies that improve outcome on the high-risk patients [9].

	Limitations of the Study

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 
This study was performed in a single hospital with a high surgical volume and may not reflect the experience of hospitals performing a small number of CABG operations, as outcome can be related to surgical volume [24].

The model was developed and validated prospectively on large groups of patients; therefore, this score will be useful in comparing outcomes in groups of patients. The use of the model for risk prediction in individual patients can be limited.

A logistic regression model cannot include every risk factor, especially factors with low incidence. In the present study, factors such as asystolic arrest, open chest on the day of operation, and tricuspid valve operation were highly significant univariate predictors of poor outcome but of such low prevalence (all <0.8%) that they were not retained in the logistic regression models. There were too few outcomes to develop reliable univariate risk estimates for several predictive factors, such as aortic valve repair, ventricular aneurysm repair, or carotid repair performed concurrently with CABG.

The simplification of continuous variables into scoring categories introduces large increments or steps of risk in the clinical model that contrasts with the gradual increases apparent in the logistic regression models. For example, creatinine clearance and renal reserve differ only slightly between a patient with a preoperative creatinine level of 1.9 mg/dL and one with a creatinine level of 1.8 mg/dL, but the clinical scoring system assigns increased risk to one and not the other.

	Conclusions

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 
The ability to predict outcome with an ICU admission score enhances preoperative risk stratification and allows sequential prognosis, isolation of preoperative status from operative care in outcome assessment, and the ability to use a revised prognosis at ICU admission to plan further medical care and interventions.

Morbidity and mortality in CABG patients are influenced by factors unique to the cardiac surgical patient, which may limit the applicability of general scoring systems to this specialized population. This ICU admission clinical model assesses the impact of comorbid disease, and includes factors that may not be adequately assessed in models derived from discharge or other administrative data [25–27]. The impact of ICU admission physiology and operative events should be considered in assessing postoperative outcome.

	Acknowledgments

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 
Partial support for this study was provided by APACHE Medical Systems.

	Footnotes

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 
Address reprint requests to Dr Estafanous, Division of Anesthesiology and Critical Care, The Cleveland Clinic Foundation, 9500 Euclid Ave, Cleveland, OH 44195.

	References

Top Footnotes Abstract Introduction Material and Methods Development of the Logistic... Development of the Clinical... Results Comment Limitations of the Study Conclusions Acknowledgments References

 

1. Higgins TL, Estafanous FG, Loop FD, et al. Stratification of morbidity and mortality outcome by preoperative risk factors in coronary artery bypass patients. JAMA 1992;267:2344–8.[Abstract]
2. Paiement B, Pelletier C, Dyrda I, et al. A simple classification of the risk in cardiac surgery. Can Anaesth Soc J 1983;30:61–8.[Medline]
3. Hattler BG, Madia C, Johnson, et al. Risk stratification using The Society of Thoracic Surgeons program. Ann Thorac Surg 1994;58:1348–52.[Abstract]
4. Hannan EL, Kilburn 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;264:2768–74.[Abstract]
5. O'Connor GT, Plume SK, Olmstead EM. Multivariate prediction of in-hospital mortality associated with coronary artery bypass graft surgery. Circulation 1992;85:2111–8.
6. Kouchoukos NT, Oberman A, Kirlin JW, et al. Coronary bypass surgery: analysis of factors affecting hospital mortality. Circulation 1980;72(Suppl 1):84–9.
7. Merry AF, Ramage MC, Whitlock RML, et al. First-time coronary artery bypass grafting: the anaesthestist as a risk factor. Br J Anaesth 1992;68:6–12.[Abstract/Free Full Text]
8. Cosgrove DM, Loop FD, Lytle BW, et al. Primary myocardial revascularization. Trends in surgical mortality. J Thorac Cardiovasc Surg 1984;88:673–84.[Abstract]
9. Loop FD, Higgins TL, Panda R, et al. Myocardial protection during cardiac operations. J Thorac Cardiovasc Surg 1992;104:608–14.[Abstract]
10. Slogoff S, Keats AS. Does perioperative myocardial ischemia lead to postoperative myocardial infarction? Anesthesiology 1985;62:107–14.[Medline]
11. Higgins TL, Starr NJ. Risk stratification and outcome assessment of the adult cardiac surgical patient. Semin Thorac Cardiovasc Surg 1991;3:88–94.[Medline]
12. Cleveland WS. Robust locally weighted regression and smoothing scatterplots. J Am Stat Assoc 1979;74:829–36.
13. Harrell FE Jr, Lee KL, Matchar DB, Reichert TA. Regression models for prognostic prediction: advantages, problems, and suggested solutions. Cancer Treatment Rep 1985;69:1071–7.[Medline]
14. Hosmer DW, Lemeshow S. Goodness of fit tests for the multiple logistic regression model. Communication of Statistics: Theory and Methods 1980;A9:1043–69.
15. Metz CE. Basic principles of ROC analysis. Semin Nucl Med 1978;8:283–93.[Medline]
16. Metz CE, Wang PL, Kronman MB. A new approach for testing the significance of differences between ROC curves for correlated data. In: Decornick F, ed. Information processing in medical imaging. The Hague: Nijhoff, 1984.
17. Orr RK, Maini BS, Sottile FD, Dumas EM, O'Mara P. A comparison of four severity-adjusted models to predict mortality after coronary artery bypass graft surgery. Arch Surg 1995;130:301–6.[Abstract]
18. Pierpont GL, Kruse M, Ewald S, Weir EK. Practical problems in assessing risk for coronary artery bypass grafting. J Thorac Cardiovasc Surg 1985;89:673–82.[Abstract]
19. Kennedy JW, Kaiser GC, Fisher LD, et al. Multivariate discriminant analysis of the clinical and angiographic predictors of operative mortality from the Collaborative Study in Coronary Artery Surgery (CASS). J Thorac Cardiovasc Surg 1980;80:876–87.[Abstract]
20. Goenen M, Jacquemart JL, Galvez S, et al. Preoperative left ventricular dysfunction and operative risks in coronary bypass surgery. Chest 1987;92:804–6.[Abstract/Free Full Text]
21. Knaus WA, Wagner DP, Draper EA, et al. The APACHE III prognostic system: risk prediction of hospital mortality for critically ill hospitalized adults. Chest 1991;100:1619–23.[Abstract/Free Full Text]
22. Becker RB, Zimmermann JE, Knaus WA, et al. The use of APACHE III to evaluate ICU length of stay, resource use, and mortality after coronary artery bypass surgery. J Thorac Cardiovasc Surg 1995;36:1–11.
23. Mossad E, Estafanous F. Blood use in cardiac surgery and the limitations of hemodilution. Curr Opin Cardiol 1995;10:584–90.[Medline]
24. Flood AB, Scott WR, Ewy W. Does practice make perfect? Part I: The relation between hospital volume and outcomes for selected diagnostic categories. Med Care 1984;22:98–114.[Medline]
25. Greenfield S, Aronow HU, Elashoff RM, et al. Flaws in mortality data. The hazards of ignoring comorbid disease. JAMA 1988;260:2253–5.[Abstract]
26. Jencks SF, Williams DK, Kay TL. Assessing hospital-associated deaths from discharge data. The role of length of stay and comorbidities. JAMA 1988;260:2240–6.[Abstract]
27. Jencks SF, Daley J, Draper D, Thomas N, Lenhart G, Walker J. Interpreting hospital mortality data. The role of clinical risk adjustment. JAMA 1988;260:3611–6.[Abstract]

This article has been cited by other articles:

				B. H. Nathanson and T. L. Higgins An Introduction to Statistical Methods Used in Binary Outcome Modeling Seminars in Cardiothoracic and Vascular Anesthesia, September 1, 2008; 12(3): 153 - 166. [Abstract] [PDF]

				J. Granton and D. Cheng Risk Stratification Models for Cardiac Surgery Seminars in Cardiothoracic and Vascular Anesthesia, September 1, 2008; 12(3): 167 - 174. [Abstract] [PDF]

				D. Jegger, J.-P. Revelly, J. Horisberger, L. K. von Segesser, and P. Ruchat Establishing an association between a peri-operative perfusion score system (PerfSCORE) and post-operative patient morbidity/mortality during CPB cardiac surgery Perfusion, September 1, 2007; 22(5): 311 - 316. [Abstract] [PDF]

				T. L. Higgins Quantifying Risk and Benchmarking Performance in the Adult Intensive Care Unit J Intensive Care Med, May 1, 2007; 22(3): 141 - 156. [Abstract] [PDF]

				L. Abrahamyan, A. Demirchyan, M. E Thompson, and H. Hovaguimian Determinants of Morbidity and Intensive Care Unit Stay after Coronary Surgery Asian Cardiovasc Thorac Ann, April 1, 2006; 14(2): 114 - 118. [Abstract] [Full Text] [PDF]

				M. Berman, A. Stamler, G. Sahar, G. P. Georghiou, E. Sharoni, R. Brauner, B. Medalion, B. A. Vidne, and A. Kogan Validation of the 2000 Bernstein-Parsonnet Score Versus the EuroSCORE as a Prognostic Tool in Cardiac Surgery Ann. Thorac. Surg., February 1, 2006; 81(2): 537 - 540. [Abstract] [Full Text] [PDF]

				J. Alex, R. Shah, S. C Griffin, A. R. Cale, M. E Cowen, and L. Guvendik Intensive Care Unit Readmission after Elective Coronary Artery Bypass Grafting Asian Cardiovasc Thorac Ann, December 1, 2005; 13(4): 325 - 329. [Abstract] [Full Text] [PDF]

				N. Serrano, C. Garcia, J. Villegas, S. Huidobro, C. C. Henry, R. Santacreu, M. L. Mora, and for the Epidemiological Project for ICU Research a Prolonged Intubation Rates After Coronary Artery Bypass Surgery and ICU Risk Stratification Score Chest, August 1, 2005; 128(2): 595 - 601. [Abstract] [Full Text] [PDF]

				K. Hekmat, A. Kroener, H. Stuetzer, R. H.G. Schwinger, S. Kampe, G. B.W.E. Bennink, and U. Mehlhorn Daily Assessment of Organ Dysfunction and Survival in Intensive Care Unit Cardiac Surgical Patients Ann. Thorac. Surg., May 1, 2005; 79(5): 1555 - 1562. [Abstract] [Full Text] [PDF]

				C. V. Thakar, S. Arrigain, S. Worley, J.-P. Yared, and E. P. Paganini A Clinical Score to Predict Acute Renal Failure after Cardiac Surgery J. Am. Soc. Nephrol., January 1, 2005; 16(1): 162 - 168. [Abstract] [Full Text] [PDF]

				S.-A. Hassantash, K. Mirpoor, and M. Afrakhteh Cardiac Surgery in an Iranian Teaching Hospital: Outcome and Risk Factors Asian Cardiovasc Thorac Ann, December 1, 2004; 12(4): 312 - 315. [Abstract] [Full Text] [PDF]

				M. P. Siegenthaler, K. Brehm, T. Strecker, T. Hanke, A. Notzold, M. Olschewski, M. Weyand, H. Sievers, and F. Beyersdorf The Impella Recover microaxial left ventricular assist device reduces mortality for postcardiotomy failure: a three-center experience J. Thorac. Cardiovasc. Surg., March 1, 2004; 127(3): 812 - 822. [Abstract] [Full Text] [PDF]

				H. Granfeldt, B. Koul, L. Wiklund, B. Peterzen, U. Lonn, A. Babic, and H. C. Ahn Risk factor analysis of Swedish Left Ventricular Assist Device (LVAD) patients Ann. Thorac. Surg., December 1, 2003; 76(6): 1993 - 1998. [Abstract] [Full Text] [PDF]

				R. V.H.P. Huijskes, P. M.J. Rosseel, and J. G.P. Tijssen Outcome prediction in coronary artery bypass grafting and valve surgery in the Netherlands: development of the Amphiascore and its comparison with the Euroscore Eur. J. Cardiothorac. Surg., November 1, 2003; 24(5): 741 - 749. [Abstract] [Full Text] [PDF]

				J. Dunning, J. Au, M. Kalkat, and A. Levine A validated rule for predicting patients who require prolonged ventilation post cardiac surgery Eur. J. Cardiothorac. Surg., August 1, 2003; 24(2): 270 - 276. [Abstract] [Full Text] [PDF]

				R. Ceriani, M. Mazzoni, F. Bortone, S. Gandini, C. Solinas, G. Susini, and O. Parodi Application of the Sequential Organ Failure Assessment Score to Cardiac Surgical Patients Chest, April 1, 2003; 123(4): 1229 - 1239. [Abstract] [Full Text] [PDF]

				V. A. Ferraris and S. P. Ferraris Risk Stratification and Comorbidity Card. Surg. Adult, January 1, 2003; 2(2003): 187 - 224. [Full Text]

				P. Pinna Pintor, S. Colangelo, and M. Bobbio Evolution of case-mix in heart surgery: from mortality risk to complication risk Eur. J. Cardiothorac. Surg., December 1, 2002; 22(6): 927 - 933. [Abstract] [Full Text] [PDF]

				M. P. Fillinger, S. D. Surgenor, G. S. Hartman, C. Clark, T. M. Dodds, A. J. Rassias, W. C. Paganelli, P. Marshall, D. Johnson, D. Kelly, et al. The Association Between Heart Rate and In-Hospital Mortality After Coronary Artery Bypass Graft Surgery Anesth. Analg., December 1, 2002; 95(6): 1483 - 1488. [Abstract] [Full Text] [PDF]