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): W. Dudley Johnson Permission Requests Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Hartz, A. J. Articles by Johnson, W. D. Search for Related Content PubMed PubMed Citation Articles by Hartz, A. J. Articles by Johnson, W. D. Related Collections Related Article

Ann Thorac Surg 1995;60:398-404
© 1995 The Society of Thoracic Surgeons

---

Original Articles: Cardiovascular

BUN as a Risk Factor for Mortality After Coronary Artery Bypass Grafting

Department of Family and Community Medicine, Medical College of Wisconsin, and Milwaukee Heart Surgery Associates, Milwaukee, Wisconsin

Accepted for publication March 13, 1995.

	Abstract

Top Footnotes Abstract Introduction Material and Methods Results Comment Acknowledgments References

 
Background. Although information on blood urea nitrogen (BUN) is universally available for patients who undergo coronary artery bypass grafting, BUN has not often been considered as a risk factor for mortality. This study assessed BUN as a risk factor for CABG patients.

Methods. Four data sets were evaluated that differed with respect to the types of patients and available patient information. In each of these data sets logistic regression analysis was used to examine the relationship between BUN and mortality after adjusting for other risk factors.

Results. Blood urea nitrogen level was strongly associated with mortality in each of the data sets. After adjustment for the available risk factors other than creatinine level, patients with BUN levels greater than 30 mg/dL had a relative odds of mortality ranging between 1.86 and 2.49 (p < 0.0001 in three of the data sets). Even after adjustment for creatinine level as well as the other variables, BUN was statistically significant at the p less than 0.01 level for three of the data sets.

Conclusions. The results suggest that BUN provides additional information on cardiac function that supplements the information provided by other risk factors.

	Introduction

Top Footnotes Abstract Introduction Material and Methods Results Comment Acknowledgments References

 
See also page 404.

Beginning in 1980 [1] and continuing until recently [2], there have been numerous studies to derive mathematical models for describing the relationship between patient characteristics and the risk of mortality after coronary artery bypass grafting (CABG). The purpose of these studies is to determine patient risk [3–7], assess whether some patients may be at a prohibitive risk for operation [1], perform risk-adjusted outcomes comparisons of providers [6, 8–15], or test the efficacy of a therapeutic modality [13–15].

Even though there have been many studies of the relationship between patient characteristics and patient risk, there is no consensus as to the patient characteristics that should be included in the mathematical models. One reason for this lack of consensus may be extensive redundancy of clinical data, ie, there may be alternative subsets of variables that could report on the same physiologic problems. A second reason is that the studies do not all have the same information on patient characteristics. Some studies must use data that is collected for other purposes, eg, the MedisGroups data set [9], and other studies have the option of collecting what they consider to be an optimum set of risk factors for CABG patients. Even for these later studies, however, there is no consensus as to what patient information should be collected [3–8, 10–16].

The differences among these studies and the increasing importance and invasive use of mathematical modeling suggest a need for a systematic study of risk factors. Studies should determine what set of risk factors is optimum and what sets of risk factors should be used if certain test results or other patient information are not available. In the present study we assessed one variable and demonstrated its value in several very different data sets. The variable investigated was the blood urea nitrogen (BUN) level. It is one of the most readily available risk factors for CABG patients, and it is affected by both cardiac and renal function. However, it is rarely included as a risk factor for mortality.

	Material and Methods

Top Footnotes Abstract Introduction Material and Methods Results Comment Acknowledgments References

 
Data Sets
Four different data sets were used in this study. The data sets differed with respect to the types of patients, the organization and purpose for collecting the data, and the specific information collected. For each data set the only risk factors for mortality considered were those collected before operation. When there was more than one measurement of a risk factor collected in the hospital before operation, only information on the most abnormal value was recorded for the first and second data sets, only the first value was recorded for the third data set, and only the last value before operation was recorded in the fourth data set.

Two of the data sets were collected using the MedisGroups data system [17]. This system, developed by MediQual, is a proprietary method for abstracting patient characteristics or ``key clinical findings'' (KCFs) from the medical record. The KCFs include admission symptoms, medical history, the results of 33 preadmission tests documented in the medical record (eg, angiography), physical examinations (including vital signs), and laboratory and other diagnostic tests performed during the hospitalization. No KCFs are abstracted from emergency room records or data obtained during resuscitation efforts, and this study only used KCFs collected before the operation. Information is only recorded for KCFs that are abnormal. If a laboratory test was repeated before operation, only the most abnormal value was recorded for the MedisGroups data and the first value was recorded for the CABG Quality data. Therefore, there was no information on the value of BUN levels less than 30 mg/dL, creatinine levels less than 1.7 mg/dL, and ejection fractions greater than 0.40. For some of the computations in this study we imputed values for all subjects in the range considered normal by MedisGroups. The values used for normals were 17.6 mg/dL for BUN, 1.1 mg/dL for creatinine, and 0.554 for ejection fraction. These values were the means of all values in the normal range for the third data set, which is described below. We used these imputed values for normal subjects in analyses of the relationship of BUN to other risk factors but not of the relationship of the risk factors to mortality. Previous studies of MedisGroup patients also have used a single value for subjects in the normal range [18].

A variable that was found to be a strong risk factor for the MedisGroups data bases was cardiac impairment. Patients were considered to have cardiac impairment if they had pleural effusion, ejection fraction less than 0.40, wedge pressure less than 20 mm Hg, or systolic blood pressure less than 90 mm Hg.

One of the MedisGroups data sets used in this study was obtained by the Health Care Financing Administration from 2,063 randomly selected Medicare admissions who had CABG in seven states. There were 550 KCFs available from the patients in this data set [19]. A second MedisGroups data set of 6,973 CABG patients from 28 hospitals was obtained from the MedisGroups Comparative Data Base. Information was available on 400 KCFs from these patients. For each of the MedisGroups data bases more than 50 preoperative patient risk factors were used in the present study, as described previously [9, 19].

Patients who had a cardiac arrest before operation were removed from some analyses from the MedisGroups data sets because these patients had a probability of mortality of 51% in one data base (n = 41) and 25% in the other (n = 44) compared with average mortality rates of 5.8% and 5.3% for the two data bases. It is likely that patients who have had a cardiac arrest have different prognostic factors than other patients. The sample sizes for these high-risk patients were so small, however, that they did not affect the results. The final sample sizes were 2,022 patients in the Health Care Financing Administration MedisGroups data and 6,929 patients in the national MedisGroups data.

A third data set to measure quality of care for CABG patients included 4,238 patients from all 16 hospitals in Wisconsin that performed CABG and one hospital outside of Wisconsin. Data from the Wisconsin hospitals were abstracted by the Wisconsin Peer Review Organization on all 1,998 charts of Medicare CABG patients reviewed from April 1990 to June 1991. The development of the data collection form and the data collection procedure for the Wisconsin patients has been described previously [8]. The data from the hospital outside of Wisconsin were abstracted by hospital medical records personnel from 2,340 Medicare and non-Medicare patients hospitalized in fiscal years 1990 through 1992. The same data abstraction form was used for these patients as for the Wisconsin patients. The data for all patients includes 25 patient characteristics that had been found in previous literature to be risk factors for patient outcome in CABG and were commonly available on the medical record [1, 4–7, 20]. Patients in shock before operation (the probability of mortality for these 58 patients was 43%) and patients with missing values for BUN were omitted from the analysis, leaving 4,065 patients. As with the high-risk patients in the first two data sets, eliminating these patients did not affect the results of the study.

In the third data set the ejection fraction was not recorded in the medical record for 10.4% of the patients. We made the assumption that if ejection fraction was missing it was normal because patients with missing and normal (>0.55) ejection fractions had similar mortality rates and because the ejection fraction should have been recorded in the chart somewhere for patients known to have low ejection fractions. Subjects missing ejection fractions were assigned an ejection fraction of 0.554 for some analyses, which was the mean value for all subjects with ejection fractions greater than 0.40. All analyses for this data set were repeated after eliminating patients with missing ejection fractions, but the differences were minimal and not reported.

The fourth data set was from the Milwaukee Heart Surgery Data Base, a registry of patients from a private practice specializing in high-risk patients. This data set has been described in several previous studies [15, 16, 21]. Data for the present study were obtained from the 4,135 patients undergoing CABG between February 1986 and March 1994 who had information on hospital discharge status. There were 4,094 patients available from this data set who were not missing data on any patient risk factors used in the analyses.

Statistical Analysis
The same analytic methods were used for each of the four data sets. We examined the relationship between BUN and the other risk factors using t tests if the other risk factors were binary (eg, congestive heart failure or diabetes) and the correlation coefficient if the other risk factors were continuous (eg, creatinine level or ejection fraction).

To graph the relationship between BUN and mortality we grouped together all patients with a value of BUN within a 10 mg/dL interval. Groups of patients of fewer than 100 were combined with the group with the next highest value of BUN. Each data point in the graph represents the mean value of BUN and the mortality rate for the subjects in the group.

The association of BUN and the other risk factors with in-hospital mortality was tested using logistic regression analysis. The logistic regression equation expresses the functional relationship between the values of the risk factors and the probability of mortality. It makes it possible to evaluate each risk factor after taking into account all of the other risk factors. Associated with each variable in the logistic equation is a risk for the variable. This risk is the relative odds of mortality for patients with an abnormal value of the variable compared with the odds of mortality for patients with a normal value. To make the analyses consistent across the four data sets we used the MedisGroups definitions of normal as less than 30 mg/dL for BUN and less than 1.7 mg/dL for creatinine.

We used two measures of the ability of the logistic regression equation to predict in-hospital mortality. One measure was the area under the receiver operating characteristic curve [22]. This area ranges from a value of 0.50 if the equation has no ability to predict outcome to 1.00 if the equation is a perfect predictor of outcome. The second measure was the sensitivity and specificity of the logistic regression equation. We defined the sensitivity of the equation as the percentage of patients who died who had a probability of dying from the logistic regression equation of greater than 10%, and we defined the specificity of the equation as the percentage of patients who did not die who had a probability of dying of less than 10%. We chose 10% as the cut off point because very few patients had probabilities of dying of greater than 10%. If higher cut off points would have been selected the specificity would have improved, but the sensitivity would have been very low.

A method other than logistic regression analysis to adjust for the other variables was to analyze the data in the subgroup of patients who did not have any of the following indicators of ventricular or renal impairment before operation: creatinine level greater than 1.7 mg/dL, congestive heart failure, ejection fraction less than 0.40, ventricular resection, valve procedures, unstable angina, or insulin-dependent diabetes. Logistic regression analysis was not used to make further adjustments for other variables in this subgroup of patients.

	Results

Top Footnotes Abstract Introduction Material and Methods Results Comment Acknowledgments References

 
The four data sets are compared in Table 1. The patients are older in the first and third data sets because all of the patients in the first data set and 83% of the patients in the third data set were on Medicare. The mortality rate was much greater for data sets 3 and 4 than for the first two data sets. This higher mortality rate may have resulted from a selection of higher risk patients in the third and fourth data bases as shown by higher percentages of patients with low ejection fractions and reoperations. The fourth data set also was unusual in that 60% of the patients had coronary endarterectomy compared with 10% in the third data set. Information on endarterectomies was not recorded in the first two data sets.

View larger version (20K): [in this window] [in a new window]   Fig 1. . Relationship between blood urea nitrogen (BUN) level and mortality. The sample sizes for points marked with an asterisk are between 133 and 143 patients, and the sample sizes for points marked with a plus are between 193 and 246 patients. The sample sizes for all other points are between 1,056 and 6,518 patients.

Despite the high levels of statistical significance and the high relative risks of the variables in the equation, the performance measures for the full equation suggest that the equation was not good at identifying patients who would die after operation. The areas under the receiver operating characteristic curves ranged from 0.66 to 0.78, the sensitivities were 35%, 36%, 56%, and 67%, and the corresponding specificities were 88%, 91%, 81%, and 71%.

In Table 4 the risk associated with BUN before and after adjusting for other patient risk factors is summarized for each data set. The unadjusted risk was large and highly significant for each of the data sets but somewhat lower for the first data set than for the others. After adjusting for other risk factors except for creatinine, the risk was still significantly associated with mortality for each of the data sets. The risk was higher for the third data set than for the others. Adjusting for creatinine in addition to the other risk factors substantially reduced the risk associated with BUN, but the risk was still statistically significant in three of the data sets. Creatinine was statistically significant at the p less than 0.05 level after adjusting for BUN in the first and fourth data sets. When the cutoff point for BUN was set at 25 mg/dL instead of 30 mg/dL, the risk for BUN adjusted for creatinine was 1.80 (p = 0.0002) in the third data set and 1.73 (p < 0.0001) in the fourth data set.

The third data set contained information for 963 patients on the highest creatinine value after operation as an indication of renal function. The partial correlation of BUN with postoperative creatinine level was 0.05 (p < 0.10) after adjusting for the preprocedure creatinine value. This suggests that the presurgical BUN value provides little information on postoperative renal function beyond that provided by the presurgical creatinine level.

	Comment

Top Footnotes Abstract Introduction Material and Methods Results Comment Acknowledgments References

 
Four data sets with very different patients and available patient information were assessed in this study. In each of these data sets BUN was found to provide information on the risk of mortality that was independent of the information provided by all other available risk factors. In particular, BUN significantly improved prediction equations that contained creatinine. Although BUN was one of the most statistically significant risk factors in three of the data sets, it was only a small piece in the complex puzzle of predicting mortality. For this study, as in previous studies, most patients who died did not have high risk according to the prediction equations.

A possible explanation for the association between BUN and mortality is that BUN is more responsive than creatinine to prerenal blood flow. The two major influences on prerenal blood flow are cardiac output and the degree of prerenal stenosis. If the relationship between BUN and prerenal stenosis accounted for our findings, we would have expected that patients with high values of BUN would have had more postoperative renal impairment as indicated by large increases in postoperative creatinine. However, we found only a weak relationship between BUN and postoperative creatinine after taking into account preoperative creatinine. Therefore, it is unlikely that BUN is primarily a marker for prerenal stenosis.

It is more likely that BUN is a marker for cardiac output. Ejection fraction also provides an indication of cardiac output. If BUN only provided information on cardiac output and ejection fraction were a very good indicator of cardiac output, then BUN would not be statistically significant when ejection fraction was in the equation. However, BUN was significant when ejection fraction was in the equation and for three of the data sets BUN was more significant than ejection fraction. There are several reasons that ejection fraction may not be a good measure of cardiac output for some patients:

1. Ejection fraction often is not measured accurately. The measurements depend on only the right anterior oblique projection, and they will be inaccurate if the cardiac contraction is not uniform due to focal muscle damage. In addition, ejection fraction is not measured with the same methods or care in all hospitals; any errors in measurement or differences in measurement from one site to another will weaken the association between ejection fraction and mortality.
   
2. Ejection fraction from cardiac catheterization only measures heart function at one point in time, during rest and often without usual medications. The ejection fraction varies, particularly when the cardiac demand increases.
   
3. Cardiac output is determined by end-diastolic volume and heart rate as well as ejection fraction.
   

In contrast to the limitations of ejection fraction, BUN is an indication of the patient's average cardiac output over a period of time. In addition, it is more precisely and uniformly measured than ejection fraction. For these reasons BUN may provide information on cardiac output that is not provided by ejection fraction. It is also possible the BUN may be an indication of physiologic parameters important for predicting survival other than cardiac output. Further exploration of the reason that BUN is a strong risk factor may add to the understanding of the underlying physiologic factors that influence the risk of patient mortality after bypass operations.

Regardless of the reason that BUN is important, it is in many ways an ideal risk factor. It can be measured accurately, inexpensively, and objectively, and it is a good predictor of mortality in diverse data sets after adjusting for most other commonly used risk factors. Because BUN is an excellent risk factor and has been consistently overlooked, this study raises the larger issue of how to determine what risk factors are of the most value. Although there have been numerous studies by large, well-funded research groups for more than 30 years, there is still no consensus of what factors should be used to determine risk for CABG patients. Such a consensus would improve patient risk assessment and decrease the resources used to collect unnecessary data.

	Acknowledgments

Top Footnotes Abstract Introduction Material and Methods Results Comment Acknowledgments References

 
We appreciate the help of MediQual Systems, Inc, in supplying some of the data for this study.

	Footnotes

Top Footnotes Abstract Introduction Material and Methods Results Comment Acknowledgments References

 
Address reprint requests to Dr Hartz, Department of Family and Community Medicine, Medical College of Wisconsin, 8701 Watertown Plank Rd, Milwaukee, WI 53226.

	References

Top Footnotes Abstract Introduction Material and Methods Results Comment Acknowledgments References

 

1. 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]
2. Pennsylvania Health Care Cost Containment Council. Coronary artery bypass graft surgery: technical report. Vol III:1992. Harrisburg, PA, December 1994.
3. Higgins TL, Fawzy EG, Loop F, et al. Stratification of morbidity and mortality outcome by preoperative risk factors in coronary artery bypass patients: a clinical severity score. JAMA 1992;267:2344–8.[Abstract/Free Full Text]
4. Myers WO, Davis K, Foster ED, Maynard C, Kaiser GC. Surgical survival in the Coronary Artery Surgery Study (CASS) registry. Ann Thorac Surg 1985;40:245–60.[Abstract]
5. Kennedy JW, Kaiser GC, Fisher LD, et al. Clinical and angiographic predictors of operative mortality from the Collaborative Study in Coronary Artery Surgery (CASS). Circulation 1981;63:793–802.[Abstract/Free Full Text]
6. O'Connor GT, Plume SK, Olmstead EM, et al. Multivariate prediction of in-hospital mortality associated with coronary artery bypass graft surgery. Circulation 1992;85:2110–8.[Abstract/Free Full Text]
7. Parsonnet V, Dean D, Bernstein AD. A method of uniform stratification of risk for evaluating the results of surgery in acquired adult heart disease. Circulation 1989;79(Suppl 1):3–12.
8. Hartz AJ, Kuhn EM, Kayser K, Pryor DP, Green R, Rimm AA. Assessing providers of coronary revascularization: a method for peer review organization. Am J Public Health 1992;82:1631–40.[Abstract/Free Full Text]
9. Hartz AJ, Kuhn EM, Green R, Rimm AA. The use of routinely collected data to assess quality of care for coronary artery bypass surgery and angioplasty. Int J Technol Assess Health Care 1992;8:524–38.[Medline]
10. O'Connor GT, Plume SK, Olmstead EM, et al. A regional prospective study of in-hospital mortality associated with coronary artery bypass grafting. JAMA 1991;266:803–9.[Abstract/Free Full Text]
11. Hannan EL, Kilburn H Jr, O'Donnell JF, Lukacik G, Shields EP. Adult open heart surgery in New York State: an analysis of risk factors and mortality rates. JAMA 1990;264:2768–74.[Abstract/Free Full Text]
12. Hannan EL, Kilburn H, Racz M, Shields E, Chassin MR. Improving the outcomes of coronary artery bypass surgery in New York State. JAMA 1994;271:761–6.[Abstract/Free Full Text]
13. Hannan EL, Siu AL, Kumar D, Kilburn H, Chassin MR. The decline in coronary artery bypass graft surgery mortality in New York State: the role of surgeon volume. JAMA 1995;273:209–13.[Abstract/Free Full Text]
14. Hlatky MA, Califf RM, Harrell FE, Lee KL, Mark DB, Pryor DB. Comparison of predictions based on observational data with the results of randomized controlled clinical trials of coronary artery bypass surgery. J Am Coll Cardiol 1988;11:237–45.[Abstract]
15. Johnson WD, Kayser KL, Hartz AJ, Saedi SF. Aspirin use and survival after coronary bypass surgery. Am Heart J 1992;123:603–8.[Medline]
16. Johnson WD, Brenowitz JB, Kayser KL. Factors influencing long-term (10-year to 15-year) survival after a successful coronary bypass operation. Ann Thorac Surg 1989;48:19–24.[Abstract]
17. Brewster AC, Karlin BG, Hyde LA, Jacobs CM, Bradbury RC, Young MC. MEDISGROUPS: a clinically based approach to classifying hospital patients at admission. Inquiry 1985;22:377–87.[Medline]
18. Krakauer H, Bailey RC, Skellan KJ, et al. Evaluation of the HCFA model for the analysis of mortality following hospitalization. Health Services Res 1992;27:317–35.
19. Hartz AJ, Kuhn EM, Pryor DP, et al. Mortality after angioplasty and coronary artery bypass surgery: the national medicare experience. Am J Cardiol 1992;70:179–85.[Medline]
20. Kouchoukos NT, Ebert PA, Grover FL, Lindesmith GG. Report of the Ad Hoc Committee on Risk Factors for Coronary Artery Bypass Surgery. Ann Thorac Surg 1988;45:348–9.[Abstract]
21. Johnson WD, Kayser KL, Pedraza PM. Angina pectoris and coronary bypass surgery: patterns of prevalence and recurrence in 3105 consecutive patients followed up to 11 years. Am Heart J 1984;108:1190–7.[Medline]
22. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating curve. Diagn Radiol 1982;143: 29–36.

This article has been cited by other articles:

				V. A. Ferraris, F. H. Edwards, D. M. Shahian, and S. P. Ferraris Risk Stratification and Comorbidity Card. Surg. Adult, January 1, 2008; 3(2008): 199 - 246. [Full Text]

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

				R. S. MacWalter, S. Y.S. Wong, K. Y.K. Wong, G. Stewart, C. G. Fraser, H. W. Fraser, Y. Ersoy, S. A. Ogston, and R. Chen Does Renal Dysfunction Predict Mortality After Acute Stroke?: A 7-Year Follow-Up Study Stroke, June 1, 2002; 33(6): 1630 - 1635. [Abstract] [Full Text] [PDF]

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): W. Dudley Johnson Permission Requests Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Hartz, A. J. Articles by Johnson, W. D. Search for Related Content PubMed PubMed Citation Articles by Hartz, A. J. Articles by Johnson, W. D. Related Collections Related Article