Health-Related Quality of Life as a Predictor of Mortality Following Coronary Artery Bypass Graft Surgery FREE

Coronary artery bypass graft (CABG) surgery is one of the most common surgical procedures in the United States and requires considerable resources. In 1995, approximately 573,000 CABG surgeries were performed in the United States at an average cost of $44,820 per procedure.¹ Clinical decision making about whether to perform CABG surgery usually includes physician assessment of a patient's mortality risk from the procedure. Clinical predictors of mortality following CABG surgery can help guide physicians and patients in risk stratification.^2- 4 Established clinical risk factors include a history of previous heart surgery, the severity of coronary artery disease, and the degree of comorbidity.

There is an increasing interest in exploring health-related quality of life (HRQL) in relation to a given health care episode, such as CABG surgery.^5- 6 Multiple self-report (by the patient) health-status measurement tools to assess HRQL, such as the Short-Form 36 (SF-36), have been developed in the past 2 decades.⁷ While these tools vary in the constructs they measure, all have the common goal of capturing "health status as perceived by the patient in areas of health identified to be of value to the patient."⁸ The most common constructs proposed for generic health surveys to capture HRQL are physical functioning, psychological functioning, social functioning, role functioning, and general health perceptions.^8- 10 Measurements of self-perceived health status can be used to evaluate the broad impact of a disease on a patient and the effectiveness of interventions. They can play a role in the clinical management of patients with cardiac disorders by extending the assessment process beyond traditional clinical parameters and tracking the multidimensional impact of treatments over time.¹¹

To date, preoperative HRQL has not been evaluated as a predictor of mortality following CABG surgery. Evaluating HRQL as a mortality predictor may add to the understanding of how the outcome of a given health care episode is influenced by a patient's preoperative health status. Furthermore, it is possible that patient self-report may be a valuable tool for risk stratification before CABG surgery. The present study tests the hypothesis that preoperative HRQL variables reflecting the physical and mental components of health status independently predict 6-month mortality following CABG surgery, even after adjusting for traditional clinical risk variables.

The study population was derived from the Department of Veterans Affairs (VA) Cooperative Study in Health Services No. 5, entitled Processes, Structures, and Outcomes of Care in Cardiac Surgery (PSOCS), a multicenter, prospective, observational study investigating the links between processes and structures of care and risk-adjusted outcomes. Details of the PSOCS study are published elsewhere^12- 13 and are only reported here as relevant to the current analyses. Approximately 1500 variables representing patient-related risk factors, processes, structures, and outcomes of care were collected for a representative sample of 4969 patients who underwent cardiac surgery at 14 VA medical centers between September 1992 and December 1996. The 14 VA medical centers (Ann Arbor, Mich; Asheville, NC; Charleston, SC; Denver, Colo; Hines, Ill; Little Rock, Ark; Milwaukee Wis; Minneapolis, Minn; Nashville, Tenn; New York, NY; San Antonio, Tex; Tucson, Ariz; West Roxbury, Mass; and San Diego, Calif) were chosen from the total of 43 VA medical centers that perform open heart surgery to be representative of the spectrum of risk-adjusted operative mortality by center. Data were prospectively collected by full-time, trained research nurses located at each of the 14 sites. Risk data were obtained by patient interview and chart review within 72 hours prior to surgery. The baseline SF-36 was given to the patients for self-administration within 72 hours of surgery. If a patient was unable to complete the SF-36, the research nurse conducted a personal interview for administration of the baseline SF-36.

All patients enrolled in the PSOCS who only had CABG surgery and who completed a preoperative SF-36 were included. Of the 3956 patients who underwent CABG surgery who were enrolled in the PSOCS study, 2480 (63%) completed a baseline SF-36. The primary reason for not completing the baseline SF-36 was urgent or emergency surgical priority that precluded time to obtain the assessment. Therefore, the study population predominantly included patients undergoing elective surgery.

The primary independent, or predictor, variables of interest were the Physical Component Summary (PCS) and Mental Component Summary (MCS) scores from the preoperative SF-36 (Figure 1). The SF-36 measures 8 health constructs using 8 scales with 2 to 10 items per scale (total of 36 questions).¹⁴ Several advantages of the PCS and MCS over the original 8 scales of the SF-36 have been reported.^15- 16 The summary scores are standardized to the general US population (mean score, 50; SD, 10), allowing easier norm-based interpretation. Scoring of the SF-36 for the original 8 scales and the PCS and MCS summary scores followed the methods described by Ware et al.¹⁵ For the PCS, very high scores indicate no physical limitations, disabilities, or decrements in well-being as well as high energy level. Very low scores indicate substantial limitations in self-care, physical, social, and role activities; severe bodily pain; or frequent tiredness.¹⁵

For the MCS, very high scores indicate frequent positive affect, absence of psychological distress and of limitations in usual social/role activities due to emotional problems. Very low scores indicate frequent psychological distress, and substantial social and role disability due to emotional problems.¹⁵

Candidate preoperative clinical variables for risk adjustment (Table 1) were derived from the published literature on clinical risk variables for mortality following CABG surgery in both VA and non-VA populations.^2- 4

The dependent, or outcome, variable was all-cause 6-month mortality, defined as death due to any cause within 180 days following CABG surgery. Mortality status was determined by the research nurses and confirmed using the VA Beneficiary Identification and Record Locator system, shown to be comparable with the National Death Index for mortality assessment in a VA population.¹⁷ The cause of death was determined by consensus of the multidisciplinary PSOCS Cause of Death Committee, composed of 6 clinician investigators, using all available information including final hospital summaries and death certificates.

All analyses were performed using the SAS statistical software package.¹⁸ Baseline characteristics were compared between patients who underwent CABG surgery included in the study population and excluded patients (ie, those without a baseline SF-36) using the Wilcoxon rank sum test for continuous variables and the χ² test for categorical variables. Univariate analyses were performed between the candidate independent variables (PCS, MCS, and traditional clinical variables) and 6-month mortality using simple logistic regression. Independent variables associated with 6-month mortality that had P≤.10 in the univariate logistic regression analysis were considered in the multivariate modeling of 6-month mortality using multiple logistic regression (SAS Proc Logistic). Odds ratios were calculated for each independent variable in the multivariate models, and 95% confidence intervals (CIs) were calculated using maximum likelihood methods.¹⁹ Because the magnitude of the odds ratio (OR) for continuous variables depends on the increment of the variable, we chose approximately 1-SD increments for all continuous variables to standardize the comparison.

The goal of our multivariate analyses was to determine the relative magnitude of PCS and MCS scores as predictors of mortality after adjusting for traditional clinical risk variables. To this end, our primary risk model included both HRQL (PCS and MCS scores) and traditional clinical risk variables as candidate independent variables. To confirm the association between PCS and MCS scores and mortality, we developed an alternative risk model using only clinical risk variables as candidate predictors. We then added PCS and MCS separately and tested their significance as predictors of mortality (above and beyond the traditional clinical risk variables) using the likelihood ratio test statistic.²⁰

Discrimination and reliability (calibration) of the multivariate models were assessed by c-index and the Hosmer-Lemeshow goodness-of-fit test statistic, respectively.^20- 21 The c-index is the percentage of time that patients who died had a higher predicted probability of death than patients who lived for all possible pairs of patients, one of whom lived for at least 6 months and the other of whom died within 6 months.

In Table 1 the baseline characteristics of the 2480 patients who underwent CABG surgery are compared with the 1476 patients excluded because they had no baseline SF-36. In general, patients included in this study had a lower-risk profile than those not included, likely due to the logistical requirements required to obtain a preoperative SF-36. For example, 11% of the study population had urgent or emergency surgical priority vs 40% of the excluded patients.

The mean preoperative PCS score for the study population was 32.6 or 1.74 SDs below the mean for the general US population. The mean preoperative MCS score was 44.0 or 0.6 SDs below the mean for the general US population.

Six-month mortality for the study population was 4.7% (117 deaths) compared with 6.4% (92 deaths) for the excluded patients. In the study population, 69 deaths (60%) occurred within 30 days of surgery. Of the 117 deaths, the Cause of Death Committee classified 69 deaths (59%) as cardiac, 45 (38%) as noncardiac, and 3 (3%) as unknown cause.

Univariate predictors (P≤.10) of 6-month mortality are shown in Table 2. The PCS score was associated with 6-month mortality, with an OR of 1.59 (95% CI, 1.27-2.00; P<.001) for a 10-point lower baseline PCS score. However, MCS score was not associated with 6-month mortality; the OR was 1.04 (95% CI, 0.89-1.22; P=.59) for a 10-point lower baseline MCS score.

Our primary multivariate risk model, which included both HRQL and clinical variables as candidate predictors of mortality, showed that PCS was a statistically significant predictor of 6-month mortality even after adjusting for traditional clinical risk variables. A 10-point (1 SD) decrement in a baseline PCS score had an OR of 1.39 (95% CI, 1.11-1.77; P=.006). Statistically significant predictor variables in the multivariate risk model included recent myocardial infarction, prior heart surgery, left ventricular ejection fraction of less than 0.25, left ventricular ejection fraction of 0.25 to 0.34, diuretic use, age, 3-vessel coronary artery disease, PCS score, serum creatinine level, and smoking history (Table 3). This model had good discriminative power (c-index, 0.774) and calibration (P=.96 for Hosmer-Lemeshow test statistic).

Our alternative risk model, in which we added the HRQL variables to a completed model of traditional clinical risk variables, confirmed that PCS scores predicted mortality above and beyond the traditional clinical risk variables. In this model, a 10-point lower PCS score had an OR of 1.32 (95% CI, 1.04-1.69; P=.03). However, when a baseline MCS score was added to the alternative risk model, it was not associated with 6-month mortality, with an OR of 1.09 (95% CI, 0.92-1.29; P=.31) for a 10-point lower baseline MCS score. The power to conclude that MCS did not predict 6-month mortality was estimated to be 82% to detect a difference in the β coefficient of .02.

Preoperative PCS and MCS scores for the study population were below norms for the US population. The mean preoperative PCS and MCS scores correspond to approximately the 9th and 24th percentiles, respectively, of the US general population.¹⁵ However, the preoperative scores for the study population were higher than the 1996 VA national averages of 31.1 for PCS and 40.7 for MCS, likely reflecting the relatively lower-risk profile of the patients who underwent CABG surgery in the study population.²²

In this study, a preoperative PCS score was an important independent risk factor for 6-month mortality following CABG surgery, even after adjusting for traditional clinical risk factors. The magnitude of risk associated with change in a PCS score was greater than for change in serum creatinine level and smoking history, established risk factors for mortality following CABG surgery.^2,4 In multivariate analysis, a 10-point decrement in a baseline PCS score was associated with a 39% increase in the risk of 6-month mortality following CABG surgery. On the other hand, the baseline MCS score did not predict 6-month mortality following CABG surgery, and this study had more than 80% power to draw this conclusion. Therefore, patient self-report of the baseline physical component of health, but not the mental component of health, predicted mortality following CABG surgery in this study population.

The findings of this study are strengthened by its prospective design, large population, and comprehensive set of clinical risk variables considered for risk adjustment. To our knowledge, this study is the first to use the SF-36, a widely used, reliable, and valid health status questionnaire, in relation to CABG surgery. Furthermore, this study quantitates the magnitude of risk associated with the HRQL variables, thereby placing them in context with traditional clinical risk variables.

Although HRQL variables have not been evaluated as predictors of mortality in patients who undergo CABG surgery, several studies provide support for the findings of this study. Functional status measurements, which include some similar items to components of the PCS, have been reported to predict mortality after hospitalization in the elderly.²³ In addition, activities of daily living and general health have been reported to predict mortality in patients with congestive heart failure.²⁴ Also, lower PCS scores alone predicted 5-year mortality in patients enrolled in the Medical Outcomes Study.¹⁵

Several studies have established that patient-reported HRQL measurements provide information beyond measurement of clinical variables alone in patients with cardiac disease. Permanyer-Miralda et al¹¹ suggested that performance measured by an exercise treadmill may be better predicted by self-perception of health than by conventional interpretation of functional capacity by physicians. Krumholz et al²⁵ reported that the physical function scale of the SF-36 was more responsive to change with angioplasty than Canadian Cardiovascular Society anginal class, concluding that the SF-36 was "capable of depicting the burden of disease and the benefits of treatment across a number of highly valued health states." Chen et al²⁶ evaluated anginal status, the SF-36, and comorbidity in patients with coronary artery disease, and noted that, "our results argue for the inclusion of measurements of patient health-state preferences in addition to those of symptom severity in studies of treatment outcomes."²⁶

The finding that the baseline MCS score was not predictive of mortality was somewhat surprising as previous evidence has linked mental health and cardiac outcomes, including mortality. Depression is an independent risk factor for the development of ischemic heart disease, and has been shown to predict long-term mortality in patients with known coronary artery disease.^27- 29 Furthermore, an MCS score of less than 42 has been associated with clinical depression.¹⁵ Even though approximately 47% of the study population had a baseline MCS score of less than 42, the MCS score was not predictive of mortality in this study.

A question may be raised about the necessity and practicality of measuring baseline HRQL in addition to the traditional clinical variables such as left ventricular ejection fraction or serum creatinine. We feel there are several reasons why finding that the baseline PCS score predicts mortality following CABG surgery is important and should be considered for preoperative risk assessment. First, from an epidemiologic standpoint, the PCS score represents an independent risk factor for mortality following CABG surgery, even after adjustment for traditional clinical risk variables. Therefore, a baseline PCS score helps to explain the variance in the outcome above and beyond clinical variables. Second, the magnitude of effect associated with a change in PCS score (39% increase in risk for an approximately 1-SD lower PCS score) is clinically important. Third, the SF-36 is a noninvasive, patient self-report measurement. All previous established risk factors for mortality following CABG surgery are either test results (some invasive) or subjective measurements by a physician. Finding a patient self-report measurement that predicts mortality shows that patients can, in effect, assist in the classification of their risk status. Finally, health status measurements, such as the SF-36, are increasingly available to physicians. For example, the VA now collects annual assessments on veterans using modified versions of the SF-36 and SF-12.^22,30 Data like these may become increasingly important to clinicians and patients making clinical decisions when health status can be linked to care outcomes.

Determination of a preoperative PCS score may be useful when counseling patients regarding their risk of mortality in conjunction with traditional clinical measurements. For example, if a patient is at moderate risk for operative mortality from CABG surgery based on traditional clinical variables, he/she may decide against surgery if he/she knows the incremental risk for mortality based on a low PCS score. Alternatively, a patient with significant comorbidity, but a good PCS score, may be considered a better candidate compared with a similar patient with a lower PCS score. Moreover, it is possible that patients with low preoperative PCS scores need additional clinical evaluation or therapy targeting factors that may be contibuting to their lower PCS scores.

Several limitations of this study should be addressed. First, the study population included only patients with a baseline SF-36, conferring a selection bias toward more elective, lower-risk cases. It is often logistically difficult to obtain an SF-36 in urgent/emergency cases. Furthermore, most patients who require urgent/emergency intervention have pressing indications for CABG surgery, such as life-threatening ischemia, which may make patient-reported HRQL information less important in clinical decision making. Therefore, our largely elective surgical study population may reflect a population in which one may consider baseline HRQL in preoperative evaluation. Second, this study does not address determinants of the baseline PCS score. Understanding the specific causes of a low preoperative PCS score may improve clinical interpretation of the score and help to determine if the score can be improved. Traditional clinical risk measurements cannot simply be a substitute for patient self-report of physical health. Identification of the determinants of self-report of physical health remains an important goal for future research. Third, this study was conducted in a VA patient population, which is largely made up of men, and veteran populations appear to have worse health status than nonveteran populations. Both of these factors may limit the generalizability of our findings to non-VA settings.^6,22,30

In summary, this study demonstrated that the PCS score from the preoperative SF-36 is an independent predictor of mortality following CABG surgery. The baseline MCS score does not appear to be associated with mortality. In this era of limited resources and expensive testing, preoperative patient self-report of the physical component of health status may be useful in risk stratification and decision making before undergoing CABG surgery.