Ann Thorac Surg 2004;77:557-562
© 2004 The Society of Thoracic Surgeons
Original article: cardiovascular
Health status and social risk correlates of extended length of stay following coronary artery bypass surgery
Gilbert Johnston, MDa*,
J. Richard Goss, MD, MPHb,
Judith A. Malmgren, PhDc,d,
John A. Spertus, MD, MPHe
a St. Joseph Medical Center, Tacoma, Washington, USA
b Division of General Internal Medicine, Department of Internal Medicine,Seattle, WA, USA
c Department of Epidemiology, University of Washington, Seattle, Washington, USA
d HealthStat Consulting, Inc, Seattle, Washington, USA
e Department of Cardiovascular Research, Mid-America Heart Institute/Saint Luke's Hospital and the University of Missouri–Kansas City, Kansas City, Missouri, USA
Accepted for publication August 6, 2003.
* Address reprint requests to Dr Johnston, Cardiac Surgery Group, 1802 S. Yakima #102, Tacoma, WA 98405, USA.
e-mail: gilj{at}nwheartcenter.com
 |
Abstract
|
|---|
BACKGROUND: Preoperative severity of illness in patients undergoing coronary artery bypass grafting (CABG) surgery is a major determinant of clinical postoperative outcomes and surgical length of stay (SLOS). Preoperative patient reported health status and social risk have not been quantified as predictors of SLOS post-CABG. Our hypothesis was that poorer self-reported health and greater social risk, as measured by standardized instruments, are significantly associated with extended SLOS defined as greater than or equal to 7 days.
METHODS: In the pilot phase of the Washington State Clinical Outcomes Assessment Program (COAP) patients in a case series between 1995 and 1996 at all hospitals with a cardiac surgery program were administered preoperative SF-36 and Seattle Angina Questionnaires (SAQ) in addition to the collection of prospective clinical data with Society of Thoracic Surgeons' compatible definitions (n = 1073). Factors found significant from bivariate analysis were incorporated into a logistic regression model to assess relative association with extended SLOS (
7 days).
RESULTS: The final model included the following elements in descending order of significance: site, SF-36 health perceptions (HP) scale, social risk factors, age, intraaortic balloon pump, congestive heart failure, comorbidity score more than 2, preoperative days more than 2, emergency operation, prior CABG, and gender.
CONCLUSIONS: The HP subscore of the SF-36 and the composite social risk factors score were significantly associated with extended SLOS after controlling for other standard clinical variables. "Hospital site" remained the factor with the greatest variance independent of patient severity of illness.
|
Introduction
|
|---|
The time of hospitalization from coronary by bypass surgery to discharge (surgical length of stay [SLOS]) is an important determinant of cost variability among patients and has been utilized as a metric for quality and efficiency of care. Accordingly, considerable effort has been made to study and understand the factors that influence and improve our ability to predict "expected" length of stay. Particular interest centers on identifying patients who may be at high risk for extended lengths of stay to better anticipate the needs of the patient. Both demographic variables and variables pertaining to severity of illness have been demonstrated to be predictive of SLOS. Self-reported health status has been identified as an important independent predictor of mortality after coronary artery bypass graft (CABG) surgery [1], but it has not been well studied as a predictor of prolonged SLOS. If patient reported health and socioeconomic status are independent predictors of extended SLOS, then efforts to obtain this information at baseline may help case management activities, and may permit better case-mix adjustment when assessing the variability in costs and SLOS across institutions.
Beginning in the early 1990s, a physician-led cardiac outcomes reporting program was instituted in Washington State. The first phase of this program involved a pilot study involving 1073 patients from 14 surgical centers in the state. This effort focused on the predictive power of self-reported information on outcomes after CABG surgery. Upon completion of the pilot program, the Clinical Outcomes Assessment Program (COAP; see Appendix for more information), was inaugurated as a coordinated quality improvement program for CABG surgery and percutaneous coronary interventions (PCI) performed in Washington State hospitals [2]. This analysis was conducted with data from the pilot phase of the COAP program to test the hypothesis that poorer self-reported health and greater social risk, as measured by standardized instruments, are significantly associated with extended SLOS defined as greater than or equal to 7 days.
|
Patients and methods
|
|---|
Patients
A prospective, multicenter cohort study was conducted using patiens enrolled between February 1995 and June 1996 at 14 medical centers in Washington State. It was approved by the Institutional Review Board of the University of Washington Medical Center in August 1994. Each hospital was asked to prospectively enlist 100 consecutive patients undergoing isolated CABG surgery. Patients who had concomitant procedures, such as valve repair or replacement, aneurysm repair, pacemaker/implantable cardioverter defibrillator placement, and those unable to complete a baseline survey were excluded from the analysis. Patients were asked to participate in the study before or soon after surgery. Sixty-four percent of patients were enrolled before the surgical procedure. The remaining 36% signed the informed consent and completed baseline surveys (reflecting their preoperative status) 3 to 7 days postoperatively, their acuity status being too severe to allow preoperative administration. Validation that this method obtained data comparable to preoperative collection has been described previously [2].
Study variables
The specific variables used in this report are grouped as follows: (a) site of procedure; (b) patient demographics (age, gender, income, marital status, education, and insurance status); (c) social risk factors (feels lonely often, not having someone to trust and confide in, lives alone, and not enough social contact) [3]; and (d) health status defined by the Medical Outcomes Short Form 36 (SF-36) [4] and the Seattle Angina Questionnaire (SAQ) [5]. In addition, the following clinical variables were included in the model and defined to be compatible with the Society of Thoracic Surgeons (STS) national database (www.sts.org): diabetes, creatinine, hypertension, peripheral vascular disease, cerebral vascular accident, chronic obstructive pulmonary disease, smoking history, previous myocardial infarction, unstable angina, surgical priority, prior CABG or other cardiac surgery, ejection fraction, number of diseased vessels, cardiogenic shock, preoperative intraaortic balloon pump (IABP), preoperative use of intravenous nitrates, preoperative use of inotropic agents, and presence of aortic or mitral valve disease. In addition, we collected hospital course data characterizing the surgical procedure, postprocedure events, and length of stay. Elevated creatinine was defined as more than or equal to 2.0 mg/dL, and low ejection fraction as less than 40%. The number of diseased vessels was computed using a criteria of more than 70% stenosis with left main disease counting as two diseased vessels. A comorbidity score was computed as the sum of diabetes, elevated creatinine, hypertension, peripheral vascular disease, cerebrovascular accident, and chronic obstructive pulmonary disease. The number of preoperative hospital days was computed.
Statistical methods
The primary outcome variable was extended SLOS (number of days from the date of surgery to the date of discharge) defined as postoperative length of stay greater than or equal to 7 days for patients discharged alive. The decision to dichotomize the variable was made based on the finding that SLOS in this (and other studies) was a highly skewed variable (Fig 1).
Our findings indicated a mean SLOS of 5.65 days, a median SLOS of 5.0 days, and a positively skewed distribution with a skewness value of 4.13, greater than 50 times the standard error of 0.075. There is precedent in the literature for transforming or dichotomizing skewed length of stay data [6–10]. We chose to dichotomize the element at more than or equal to 7 days consistent with other studies [6, 7] and because it reflected a clinically meaningful subset of patients (22.7% of all patients) with an average of $15,000 more in charges and a fourfold higher incidence of postprocedure events.
Analyses were performed using SPSS (SPSS Version 10.1.3, Chicago, IL). Potential variables were reviewed for association with extended SLOS by 
2 comparison for dichotomous variables and t test mean comparison for continuous variables. Variables significantly associated in bivariate analysis with extended SLOS at the p less than 0.05 level were selected for inclusion in the logistic regression model. Three variables with greater than 10% missing values (ejection fraction, income, and social risk factors) were significantly associated with extended SLOS and were coded to include the "missing values" as a separate category. Age was examined as a categorical variable defined as less than 65 years old, 65 to 75 years old, and greater than 75 years old. Baseline SF-36 and SAQ scores were used as dichotomous variables scored above and below the median. The number of preoperative days spent in the hospitals were categorized as 1 to 2 days versus greater than 2 days. The comorbidity score was categorized as zero to one conditions versus two or more conditions.
Multivariate analysis was performed first using elements from the medical record including site, age, gender, and preoperative clinical variables. Next, a separate multivariate analysis was developed using data obtained from the patient survey data (SF-36, SAQ, social risk factors, and socioeconomic questions) to determine which elements were independently predictive of extended SLOS. The final analytic model combined the base model with the patient survey elements by performing a forward stepwise logistic regression to determine which elements added independent predictive value. The final model was recomputed after eliminating one institution determined to be a SLOS outlier where 77% of patients had a SLOS greater than or equal to 7 days.
|
Results
|
|---|
One thousand seventy-three patients were enrolled in the study. Of these, 5 patients died during hospitalization and 19 were known to have died during the first 12 months following surgery (unadjusted hospital surgical mortality 0.5%; unadjusted surgical mortality at 1 year 1.1%). The number of patients enrolled per hospital ranged from 37 to 107 patients with a mean of 76 patients. The mean age was 65 years old, 82% were male, 22% had diabetes, and 10% had had a prior CABG (Table 1).
These values compare closely with the national STS database (www.sts.org) for the corresponding years.
Bivariate analyses for three groups of data were conducted including hospital site, clinical characteristics, and self-reported survey data. The site with the lowest proportion of extended SLOS patients was selected as the referent group. Compared with the referent site, the odds ratios (OR) for extended SLOS ranged from 1.34 to 79.84. Eliminating the site with the extreme value, the OR ranged from 1.34 to 10.03. Seventeen preoperative clinical variables tested were significantly associated with extended SLOS. The elements with the greatest statistical significance were congestive heart failure (CHF; OR 54.5, p < 0.001), preop IABP (OR 24.1, p < 0.001), ejection fraction less than 40% (OR 22.9, p < 0.001), comorbidity score more than or equal to 2 (OR 18.1, p < 0.001), cardiogenic shock (OR 16.4, p < 0.001), and preoperative stay of more than 2 days (OR 12.3, p < 0.001). Self-reported variables that were significantly associated with extended SLOS included: not being married (OR 12.5, p < 0.001), unemployment (OR 18.0, p < 0.001), one or more social risk factors (OR 19.3, p < 0.001), and multiple SF-36 and SAQ domains.
In the multivariate analysis, site of procedure was the element most highly associated with SLOS after controlling for site and clinical risk factors. Other clinical factors that were significantly associated with extended SLOS were age, gender, comorbidity score, prior CABG, preoperative IABP, CHF, emergent or salvage procedure, and preoperative hospital stay more than 2 days. The SF-36 health perceptions (HP) value and the social risk factors variable were independently predictive of extended SLOS when entered after site and all other clinical elements were examined (Table 2).
The SF-36 HP scale and the social risk factors scale both improved the model performance measured by a 2 improvement value of 23.2 and 12.2, respectively (Table 3). The receiver operator characteristic (ROC) c-statistics values increased from 0.802 to 0.826 with the addition of the self-reported elements. After removing the outlier site, the model's overall performance was poorer as reflected by the lower model X2. However, the relative contribution of the HP scale and the social risk factors remained similar in magnitude and statistically significant (Table 4).
View this table:
[in this window]
[in a new window]
|
Table 4. Step Summary for Base Model and Each Step of the Final Model After Eliminating One Site With a Several-Fold Higher Proportion of SLOS 7 Days
|
|
|
Comment
|
|---|
We tested the hypothesis that standardized, patient self-reported measures of health status and socioeconomic risk factors could be used to increase the precision of predicting SLOS after CABG surgery. We found that two elements in particular, the SF-36 HP scale and the social risk factors composite variable, were highly associated with extended SLOS after controlling for site and other clinical variables. In the final model only the site variable had greater statistical significance than these two elements, demonstrating their robust predictive power for this outcome of interest.
Numerous other studies have examined the relationship between preoperative clinical severity of illness and post-CABG SLOS [6, 8–12]. Despite differences in study samples, methodologies and years in which the data were collected a similar pattern of predictor variables has emerged. Age, gender, poor left ventricular function, selected comorbidities, prior cardiac surgery, and surgical priority are consistently associated with extended SLOS. This current study identified a similar set of independent clinical predictors.
Investigators have also found that measures of postoperative complications add to the predictive power of models predicting both length of stay and cost, but differ on the degree to which the models are significantly enhanced by including this information [9, 10]. We did not include procedural or postoperative events in our study design because of our interest in determining the relative strength of self-reported data before patients' entry into the surgical procedure, and because complication rates are also a function of preoperative severity of illness and perioperative quality of care, which creates a significant problem with colinearity [13]. We did, however, test our hypothesis on data from all centers when the models were rerun using any major complication, and again using any major complication plus new onset atrial fibrillation. We found that although the model X2 increased by approximately 108 points (p = 0.000) with the addition of perioperative data, the relative strength of the two self-reported measures were similar to the existing model the with 2 improvement values of 21 and 10 for HP and social risk, respectively. We also observed a similar degree of SLOS variability across cardiac institutions, suggesting that the addition of procedure associated data did not further explain the significant differences across the sites.
This study extends previous research with the finding that markers of health status and social risk added significantly to models predicting SLOS after CABG surgery. There is a growing amount of literature demonstrating that psychosocial factors and self-reported health status are important independent predictors of mortality in cardiac patients [1, 3, 14–19]. Reiley and Howard [20] found that clinical severity and functional status were the two strongest predictors of length of stay in patients with CHF. In another study of patients admitted to an acute care hospital, the investigators found that that the severity of a patient's psychosocial problem was a more significant predicator of length of stay than the diagnostic related group variable [21].
Even with the added explanatory power that self-reported data provides, the variance between hospitals in this region still exceeded that predicted by models based on the burden of patient's illness. Multi-institutional studies of postoperative length of stay have consistently observed this same effect [9, 10, 16]. As with any outcome measure that varies significantly by institution, it is important to consider the extent to which all measurable patient characteristics that may confound the relationship are included. This study has indicated that there may indeed be measurable variables that add to predictive models of SLOS. However, the likelihood that additional information on patient characteristics will fully explain the wide variability across sites is unlikely. Other theories that have been suggested include the possibility that risk-adjusted institutional differences are due differences in quality of care or in efficiency of care, the latter being most supported by the literature [9, 10, 22, 23]. Rosen and coworkers [9] noted that more than 75% of the variability across sites remained unexplained after postprocedural complications and death (which may represent quality of care issues) were included in the models. They also found that shorter stays were not associated with increased readmission rates (another marker of poor quality). They proposed that the most likely reason for hospital variability may be due to differences in efficiency of care and identify literature supporting the safe and effective role of fast-track protocols, care maps, and clinical pathways. We too found that even when controlling for postoperative complications as a possible marker of quality of care, significant variability existed between hospitals. We concur that preexisting institutional practice patterns and efficiency may be the most important determinant of SLOS, but that case-mix also has a important role in predicting which patients are likely to stay longer than average.
Strengths of this study include its multi-institutional design and prospective clinical data collection, and it involved the use of a number of patient self-reported measures of health and social risk. Prospectively collected clinical data has been demonstrated to be superior to administrative data when controlling for severity of illness [24, 25]. The major limitation was the lack of complete capture of consecutive patients by all centers, raising the possibility of selection bias. Comparisons with the state discharge abstract database reported previously [2] identified that patients enrolled in this study were in fact a somewhat healthier subset of the entire concurrent Washington State CABG population. Yet, this fact makes our findings even more interesting in that the patient-centered characteristics were able to explain additional variation among a more homogenously healthy cohort. We speculate that sicker patients at baseline would have greater health status and social risk burden, and greater variability in SLOS reflecting the heterogeneity in their clinical courses. These patterns suggest an association between self-reported health and extended SLOS that is equal to or greater than the relationships identified in this study. Further research would be needed to confirm this hypothesis.
In conclusion, this study demonstrated that, after controlling for site and many standard and well-validated clinical predictors of prolonged SLOS, patient self-report of overall health status (SF-36 HP scale) and a composite social risk variable provided a robust and independent power to predict extended SLOS. These findings suggested a potential role for administering health status surveys to patients preparing for cardiac surgery as a way to prompt more aggressive case management activities, particularly among those with relatively unremarkable clinical severity of illness. Like other studies, site remained the greatest determinant of SLOS in this region. This recurring theme likely reflects differences in institutional protocols, surgeon practice patterns and institutional culture, and identifies a need to better understand the relationships between patient characteristics, institutional practice patterns, and quality of care.
|
Acknowledgments
|
|---|
This work was supported by the Robert Wood Johnson Clinical Scholars Program; the Veteran's Affairs Health Services Research and Development Northwest Field Program; the Foundation for Health Care Quality; and the Franciscan Health System, Washington.
|
Appendix
|
|---|
The Clinical Outcomes Assessment Program (COAP) is designed to facilitate collaborative quality improvement for cardiac revascularization in Washington State. Program activities are supported through participant fees. All data collection, analyses, and dissemination are conducted in accordance with Washington State RCW 43.70.510, the University of Washington Human Subjects Review Board, and the Department of Health Human Subjects Review Board. The specific content of this manuscript may not reflect the opinions or conclusions of all members of the COAP organization.
The COAP organization is described on the COAP website (www.coap.org) including the management committee members, subcommittee leaders, and participating institutions. The following institutions contributed data to COAP that was used in this report:
- Deaconess Medical Center, Spokane
- Overlake Hospital Medical Center, Bellevue
- Providence General Medical Center, Everett
- Providence St. Peter Medical Center, Olympia
- Providence Medical Center, Yakima
- Sacred Heart Medical Center, Spokane
- St. Joseph Hospital, Bellingham
- St. Joseph Medical Center, Tacoma
- Swedish Health Services First Hill Campus, Seattle
- Swedish Health Services Providence Campus, Seattle
- Tacoma General Hospital, Tacoma
- University of Washington Medical Center, Seattle
- Veterans Affairs Medical Center, Seattle
- Virginia Mason Medical Center, Seattle
|
References
|
|---|
- Rumsfeld J.S., MaWhinney S., McCarthy M., Jr, et al. Health-related quality of life as a predictor of mortality following coronary artery bypass graft surgery. Participants of the Department of Veterans Affairs cooperative study group on processes, structures, and outcomes of care in cardiac surgery. JAMA 1999;281:1298-1303.[Abstract/Free Full Text]
- Goss J.R., Whitten R.W., Phillips R.C., et al. Washington State's model of physician leadership in cardiac outcomes reporting. Ann Thorac Surg 2000;70:695-701.[Abstract/Free Full Text]
- Williams R.B., Barefoot J.C., Califf R.M., et al. Prognostic importance of social and economic resources among medically treated patients with angiographically documented coronary artery disease. JAMA 1992;267:520-524.[Abstract/Free Full Text]
- Ware J.E., Jr, Sherbourne C.D. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care 1992;30:473-483.[Medline]
- Spertus J.A., Winder J.A., Dewhurst T.A., et al. Development and evaluation of the Seattle Angina Questionnaire: a new functional status measure for coronary artery disease. J Am Coll Cardiol 1995;25:333-341.[Abstract]
- Lazar H.L., Fitzgerald C., Gross S., Heeren T., Aldea G.S., Shemin R.J. Determinants of length of stay after coronary artery bypass graft surgery. Circulation 1995;92(9 Suppl):II20-24.
- Mounsey J.P., Griffith M.J., Heaviside D.W., Brown A.H., Reid D.S. Determinants of the length of stay in intensive care and in hospital after coronary artery surgery. Br Heart J 1995;73:92-98.[Abstract/Free Full Text]
- Tu J.V., Jaglal S.B., Naylor C.D. Multicenter validation of a risk index for mortality, intensive care unit stay, and overall hospital length of stay after cardiac surgery. Steering Committee of the Provincial Adult Cardiac Care Network of Ontario. Circulation 1995;91:677-684.[Abstract/Free Full Text]
- Rosen A.B., Humphries J.O., Muhlbaier L.H., Kiefe C.I., Kresowik T., Peterson E.D. Effect of clinical factors on length of stay after coronary artery bypass surgery: results of the cooperative cardiovascular project. Am Heart J 1999;138(1 Pt 1):69-77.[Medline]
- Ghali W.A., Hall R.E., Ash A.S., Moskowitz M.A. Identifying pre- and postoperative predictors of cost and length of stay for coronary artery bypass surgery. Am J Med Qual 1999;14:248-254.[Abstract/Free Full Text]
- Lahey S.J., Borlase B.C., Lavin P.T., Levitsky S. Preoperative risk factors that predict hospital length of stay in coronary artery bypass patients > 60 years old. Circulation 1992;86(5 Suppl):II181-185.
- Aldea G.S., Gaudiani J.M., Shapira O.M., et al. Effect of gender on postoperative outcomes and hospital stays after coronary artery bypass grafting. Ann Thorac Surg 1999;67:1097-1103.[Abstract/Free Full Text]
- Hammermeister K.E., Burchfiel C., Johnson R., Grover F.L. Identification of patients at greatest risk for developing major complications at cardiac surgery. Circulation 1990;82(5 Suppl):IV380-389.
- Ruberman W., Weinblatt E., Goldberg J.D., Chaudhary B.S. Psychosocial influences on mortality after myocardial infarction. N Engl J Med 1984;311:552-559.[Abstract]
- Case R.B., Moss A.J., Case N., McDermott M., Eberly S. Living alone after myocardial infarction. Impact on prognosis. JAMA 1992;267:515-519.[Abstract/Free Full Text]
- Butterworth J., James R., Prielipp R., Cerese J., Livingston J., Burnett D. Female gender associates with increased duration of intubation and length of stay after coronary artery surgery. CABG clinical benchmarking database participants. Anesthesiology 2000;92:414-424.[Medline]
- Jenkinson C.M., Madeley R.J., Mitchell J.R., Turner I.D. The influence of psychosocial factors on survival after myocardial infarction. Public Health 1993;107:305-317.[Medline]
- Brummett B.H., Barefoot J.C., Siegler I.C., et al. Characteristics of socially isolated patients with coronary artery disease who are at elevated risk for mortality. Psychosom Med 2001;63:267-272.[Abstract/Free Full Text]
- Bosworth H.B., Siegler I.C., Brummett B.H., et al. The association between self-rated health and mortality in a well-characterized sample of coronary artery disease patients. Med Care 1999;37:1226-1236.[Medline]
- Reiley P., Howard E. Predicting hospital length of stay in elderly patients with congestive heart failure. Nurs Econ 1995;13:210-216.[Medline]
- Keefler J., Duder S., Lechman C. Predicting length of stay in an acute care hospital: the role of psychosocial problems. Soc Work Health Care 2001;33:1-16.
- Every N.R., Hochman J., Becker R., Kopecky S., Cannon C.P. Critical pathways: a review. Committee on acute cardiac care, council on clinical cardiology, American Heart Association. Circulation 2000;101:461-465.[Free Full Text]
- Silber J.H., Rosenbaum P.R., Schwartz J.S., Ross R.N., Williams S.V. Evaluation of the complication rate as a measure of quality of care in coronary artery bypass graft surgery. JAMA 1995;274:317-323.[Abstract/Free Full Text]
- Hannan E.L., Kilburn H., Jr, Lindsey M.L., Lewis R. Clinical versus administrative data bases for CABG surgery. Does it matter?. Med Care 1992;30:892-907.[Medline]
- Hartz A.J., Kuhn E.M. Comparing hospitals that perform coronary artery bypass surgery: the effect of outcome measures and data sources. Am J Public Health 1994;84:1609-1614.[Abstract/Free Full Text]
Top
Abstract
Introduction
