International Journal for Quality in Health Care 16:157-164 (2004)
International Journal for Quality in Health Care vol. 16 no. 2 © International Society for Quality in Health Care and Oxford University Press 2004; all rights reserved
Risk adjustment for coronary artery bypass graft surgery: an administrative approach versus EuroSCORE
1 Department of Economics, University of Bologna, Bologna,
2 Regional Health Care Agency of Emilia Romagna, Bologna, Italy
Objective. To determine the ability of administrative data in predicting in-hospital mortality for patients undergoing coronary artery bypass graft surgery.
Methods. Patient data were obtained from the administrative databases on hospital discharge abstracts of the Italian region Emilia Romagna for the years 2000–2001. We used a multivariate logistic regression analysis to compare an ICD-9-CM risk adjustment approach based on administrative variables (such as age, gender, principal diagnosis, combined operation, previous cardiac surgery, emergency admission, and Charlson comorbidity index) with a risk adjustment approach based on the clinical European System for Cardiac Operative Risk Evaluation (EuroSCORE) to predict in-hospital mortality and to assess hospital performance. In order to distinguish complications of care from comorbidities, we linked hospital data across multiple episodes of care up to 1 year before the admission for coronary artery bypass graft (CABG).
Results. The risk adjustment approach based on ICD-9-CM data provides good explanatory ability in models assessing in-hospital mortality (the c statistics obtained are very close: c = 0.76 in 2000 and c = 0.80 in 2001 for the administrative model versus 0.78 in 2000 and 0.77 in 2001 for the clinical one) and in those ranking the centres (c = 0.78 in 2000 in both approaches, and c = 0.82 for the administrative model versus c = 0.78 for the clinical one in 2001).
Conclusions. Adding some administrative variables considered proxy for clinical complexity to the administrative model and linking hospital data across patients’ multiple episodes of care eliminated much of the difference in effectiveness between the clinical and administrative risk adjustment approach. Focusing on the health policy context of measuring CABG death rates, our study strengthened the thesis that, with the growing improvement in accurate coding practice, administrative databases could provide a valuable and economical source for health planning and research.
Keywords: administrative data, cardiac surgery, Charlson Index, EuroSCORE, risk adjustment
Address reprint requests to C. Ugolini, Piazza Scaravilli 2, 40126 Bologna, Italy. E-mail: cugolin{at}economia.unibo.it
Accepted for publication November 10, 2003.
Risk adjustment plays an increasing role in evaluating the quality of health care, and the selection of appropriate score systems for the evaluation of hospital performance has become an important issue. From this point of view, administrative databases represent a very valuable source of information for research on the outcomes of hospital care, even though in attempting to use information from this source for purposes of comparison, researchers encounter considerable methodological and conceptual problems [1,2]. One of the major problems with administrative data is that differences in outcomes between producers might be due to differences in their case mix. This may reflect factors such as demographic characteristics or severity of illness, including principal diagnosis and complexity of comorbidities. Statistical adjustment can improve the estimate’s reliability of the treatment’s effect, although how adequately confounding factors can be taken into account is still controversial [3]. Other concerns arise from inaccuracies and omissions in hospital discharge data. Despite these limits, the advantages of using a risk adjuster based on administrative data include lower cost and reduced data collection time. Furthermore, administrative data bases still represent the most important decision-making information sources routinely available to policy makers: indeed, they allow analyses of epidemiological changes, expenditure patterns, hospital performance, utilization review within each hospital, and longitudinal research on health reform. For these reasons, analysis based on administrative databases represents a useful first step towards detecting the existence of situations worthy of closer examination.
Comorbidity has been studied for many years for its influence on many different outcomes of hospital care and economic outcomes, including resource utilization, discharge destination, and intensity of treatment, resulting in the development of several comorbidity scales. Administrative databases are increasingly being used to derive indices of comorbidity, most often the Charlson index, and previous work has questioned the accuracy of using this information source for the purpose of measuring comorbidity, comparing the chart-based data with administrative data [4–7]. In most studies aimed at determining the accuracy of administrative data for measuring comorbidity, the Charlson index derived from medical record data was shown to be superior to the same index derived from administrative data, principally because administrative data based on hospital discharge codes consistently underestimate the presence of comorbid conditions and are unable to distinguish perfectly between comorbidities and complications.
More generally, Hannan and colleagues [8,9] compared the ability of administrative and clinical data to predict mortality for coronary artery bypass graft (CABG) surgery. They concluded that the statistical model based on clinical data had a substantially better predictive ability than the administrative approach. However, they also found that adding some clinical data elements to the administrative data eliminated much of the difference in effectiveness between the two systems. Their studies also pointed out the need to make an effort to distinguish between comorbidities and complications among secondary diagnoses in administrative data systems. Similarly, Iezzoni and colleagues [10] tested the predictive accuracy of five generic clinical and administrative severity of illness measures for CABG patients. They found that the discrimination abilities of these measures were similar and this result was accounted for by the fact that administrative models often included postoperative complication data as preoperative risk predictors. Our study expands on this work by comparing the explanatory power of an administrative model (containing several predictive variables obtainable from patient hospital discharge abstract and other variables considered as proxies of clinical complexity, and known or suspected a priori to influence the survival of CABG patients) with a model using EuroSCORE (European System for Cardiac Operative Risk Evaluation; http://www.euroscore.org). Among preoperative risk scores for heart surgery, the EuroSCORE is the most recently developed one, involving the highest number of patients and institutions for its development, with data collected for >19 000 consecutive patients undergoing open heart surgery from 132 centres in eight European countries [11–13]. The database generated was subjected to multiple regression analysis to determine which factors were associated with operative mortality. Weights were allocated to each risk factor on the basis of the odds-ratios and a risk model was constructed in which the percentage predicted mortality for a patient could be calculated by adding the weighted values of risk factors that are present. As EuroSCORE is derived from a cross-section of current European cardiac surgery, it is proposed as a standard Europe-wide method allowing institutional quality control as well as pan-European epidemiological analysis. Studies comparing different score systems previously confirmed that the EuroSCORE has been shown to work well in coronary surgery, valve surgery, and overall cardiac surgery across many European countries [14–17]. For example, Geissler and colleagues [14] compared Parsonnet, Cleveland Clinic [18], French, Euro, Pons, and Ontario Province [19] risk scoring systems by calculating the area under the receiver operating characteristic (ROC) curve, and found that EuroSCORE yielded the highest predictive value (78.6%) in their patient population. Asimakopoulos et al. [15] compared the Parsonnet Score, the EuroSCORE, the American College of Cardiology/American Heart Association (ACC/AHA) score, and three UK Bayes models on 5741 patients who underwent isolated CABG at two UK cardiac centres between 1993 and 1999, and found that both EuroSCORE and the simple Bayes models predicted the overall level of mortality reasonably well. Recently, Nashef and colleagues [20] evaluated the performance of EuroSCORE on the national database of the Society of Thoracic Surgeons (STS), demonstrating that it also performs remarkably well in North American cardiac surgery, across years, types of surgery, and risk groups, despite substantial differences in demographic, risk, and surgical characteristics between European and American patients.
No studies have compared the ability and accuracy of administrative data for evaluating differences in hospital case-mix directly with EuroSCORE so far. The present study was undertaken to address these concerns.
Data were drawn from administrative databases containing hospital discharge abstracts (SDO) of the Italian region Emilia Romagna for the years 2000 and 2001. In this region the implementation of a diagnosis-related groups (DRG)-based payment system for hospital care during the second half of the 1990s has strongly increased the completeness of secondary diagnoses contained within administrative data sources, especially for cardiac surgery, which is considered a function of regional interest. Besides, since 2000 Emilia Romagna has been promoting the insertion in the hospital discharge abstracts of EuroSCORE for patients undergoing heart surgery. This provides an opportunity to compare the accuracy of an administrative risk adjustment approach with a clinical risk index based on medical records and introduces a useful improvement for policy makers who usually approach a subject on the basis of administrative databases.
Even if the administrative databases, inexpensive and readily available, lack an important amount of clinical information necessary for risk adjustment in the prediction of patient mortality, we tested the possibility of reaching a reasonable predictive power by integrating the variables routinely obtainable from patient hospital discharge abstracts with other administrative variables potentially able to reduce unobserved clinical heterogeneity. Following the findings of Hannan et al. and Iezzoni et al. [9,10], in order to distinguish complications of care from comorbidities, we linked hospital data across multiple episodes of care for the same patient up to 1 year before the admission for CABG. This kind of link is possible in any closed health care system, such as Emilia Romagna, but also within, say, Medicare. As certain diagnostic and procedure codes may represent complications when they appear on the index discharge abstract, but clearly represent comorbidities when they appear on a previous discharge abstract for the same patient, controlling for the patient’s multiple episodes of care permits a reduction in the number of codes needed to represent probable postoperative events.
The Methods section discusses the data, the empirical specification and the statistical analysis adopted. The Results section presents the main results of the logistic regression models used after adjusting for patient-specific, risk-related characteristics.
 |
Methods |
|---|
 Top Methods ResultsDiscussionReferences |
|---|
Data source
The Italian region of Emilia Romagna is located in the north-east and has a total population of
4 million inhabitants. The supply of cardiovascular treatments in this region is characterized by a strict regulatory control and great emphasis is placed upon coordination and cooperation between public and private producers. In 1996 the Region concentrated by-pass services in six cardiac surgical centres (CSCs); four private and two public [21–23]. Our study included 3192 patients in the year 2000 and 3265 in 2001 who had been discharged from one of the six CSCs after coronary artery bypass surgery, including both emergency and elective admissions. Transfers and repeated admissions were traced using the patient identification number. Patient data were obtained from the regional administrative SDO data, which includes demographic information (age and sex), principal diagnosis, an unlimited number of secondary diagnosis and procedure codes (ICD-9-CM), discharge status, and the clinical risk EuroSCORE for patients undergoing heart surgery. EuroSCORE contains 17 risk factors that are weighted for the definitive scoring system: nine patient-related factors, four factors derived from preoperative cardiac status, and four related to the timing and nature of the operation performed. The risk factors and the weights allocated to them are as follows: age (1); sex (1); chronic pulmonary disease (1); extracardiac arteriopathy (2); neurological dysfunction (2); previous cardiac surgery (3); serum creatinine (2); active endocarditis (3); critical preoperative state (3); cardiac-related factors such as unstable angina (2), LV dysfunction etc. [1 for moderate (LVEF 30–50%) or 3 for poor (LVEF <30%)]; recent myocardial infarct (2); pulmonary hypertension (2); operation-related factors such as emergency (2) etc.; other than isolated CABG (3); surgery on thoracic aorta (3); and postinfarct septal rupture (4). The system is additive: the predicted risk for a patient is calculated by adding the scores for existing risk factors to obtain an approximate predicted mortality index.
Statistical analysis
The statistical analysis was performed using SAS version V8.1 [24]. We used multiple logistic regression (SAS procedure LOGISTIC) to evaluate the accuracy with which patient mortality could be predicted according to each different risk adjustment approach used. As the binary dependent variable we considered in-hospital mortality from CABG surgery. Also having at our disposal the regional death registry for the same period, we performed the 60-day outcome analysis, but as the results of both outcome analyses were actually quite similar we concentrated on the short-term outcome. We estimated three multivariate models. The first model proposed a risk adjustment based exclusively on the Charlson comorbidity index, one of the most widely used comorbidity rating systems. Originally developed on the basis of hospital chart data, it represents a measure of patient comorbidity [25]. Based on medical record review, Charlson and colleagues developed a weighted index measure of comorbidity that was shown to predict the adjusted relative risk of 1-year mortality, using a proportional hazards model. The index considers 19 conditions, each of which receives a weight from 1 to 6 according to its potential for influencing mortality, giving the index a possible range of 0–33. Romano et al. [26,27] adapted the index for use with the ICD-9-CM diagnostic and procedure codes available in administrative datasets, searching a patient’s hospital claims data for the presence of a specific ICD-9-CM diagnosis and procedures corresponding to Charlson comorbid conditions [28,29]. As it was constructed to predict outcomes following coronary artery bypass surgery, this adaptation was applied in our study.
The second model presented a risk adjustment approach using the Charlson comorbidity index plus other information obtainable exclusively from administrative data sources and associated with an increased mortality risk, particularly in-hospital mortality: patient-specific factors such as older age and sex, emergency admission, a combined operation (heart surgery other than isolated CABG during the same operation), and previous cardiac surgery within 1 year from CABG hospitalization. As we observed different coding methods among the six CSCs regarding patients admitted with an acute myocardial infarction—some of them recorded a primary diagnosis acute myocardial infarction (AMI) with ICD-9-CM code 410, even for patients with recent myocardial infarction—we constructed a variable for a primary diagnosis of AMI within 8 weeks of admission for CABG surgery (AMI within 8 weeks).
In the third model we included the clinical index EuroSCORE as a unique variable of risk adjustment.
In order to evaluate whether a different risk adjustment approach affected the hospital ranking, we re-ran the three models, including, among the independent variables, the dummies representing the five CSCs with reference to the sixth highest volume hospital (D), a private accredited provider. Consequently, we were able to determine the level of performance of each individual surgical centre, comparing the type of ranking obtainable through the three models.
The c-statistic was used to assess the discrimination of the model, i.e. the extent to which the model predicts higher probabilities of death for patients who die. In particular, it characterizes the predictive ability of the logistic model from c = 0.5 (the model has no predictive ability) to c = 1.0 (perfect prediction). To measure the model’s calibration, we used the Hosmer–Lemeshow statistic. It is useful to remember that a model is said to be calibrated to a data set when the average of the predicted values is close to the average of the actual outcomes: models with an Hosmer–Lemeshow statistic value >15.51 (P < 0.05) are rejected for poor fit [30,31].
|
Results |
|---|
|
TopMethodsResultsDiscussionReferences |
|---|
Table 1 illustrates patient characteristics with respect to CSC.
|
The number of CABG procedures performed in 2000 (and 2001) in each hospital ranged from 252 (257) to 855 (866), and the mean annual volume was 532 (544). The crude in-hospital mortality rates were 3.1% in 2000 and 3.3% in 2001, with a large variability among centres. Most patients were men (82% in both years); 3.7% in 2000 and 3.2% in 2001 had previous heart surgery within 1-year from the CABG hospitalization; 20% in 2000 and 13.5% in 2001 had a principal diagnosis of AMI within 8 weeks of the CABG surgery; and 17.4% in 2000 and 19.1% in 2001 have other than an isolated CABG surgery.
The patient case-mix differed between hospitals and across years. For both years CSC F presented a higher proportion of females, while CSC A had the highest proportion of patients aged 
70 years. The Charlson comorbidity score signalled more comorbid illness for patients treated in CSCs C and F for the year 2000 and in CSCs D and E for 2001. Patients admitted due to emergency were concentrated in CSCs E and F, both of which are public providers with emergency rooms.
Table 2 presents the univariate analysis of the associations between all variables and in-hospital mortality, calculating odds-ratios and P-value significance for the years 2000 and 2001.
|
The risk factors with by far the highest odds-ratios were related to: older patients; patients who underwent heart surgery other than isolated CABG (combined surgery); patients who had other heart surgery during the previous year; and patients admitted due to emergency.
Table 3 presents the multivariate logistic goodness of fit statistics for the three models using different risk adjustment approaches.
|
In both years, Model 1 ranked the lowest in terms of discriminatory ability, whereas Model 2 yielded the highest predictive value in 2001. The level of predictive performance for mortality achieved using our administrative risk approach was comparable to that achieved using the chart-based risk adjustment model. In particular, the c-statistic signalled a growing discrimination capability through the three models: c = 0.68 in 2000 and c = 0.70 in 2001 with the Charlson comorbidity index as the only risk adjuster; c = 0.76 in 2000 and c = 0.80 in 2001 with the general administrative risk adjustment approach; and c = 0.78 in 2000 and c = 0.77 in 2001 with EuroSCORE. Model 1 showed poor calibration, whereas the other two models predicted mortality reasonably accurately.
Table 4 compares the results of logistic regression models, including dummy variables for the centres and displays, for each model, including the significant risk factors for in-hospital mortality after adjustment for patient Characteristics along with their odds-ratios and the P-values.
|
In comparison with the results shown in Table 3, the inclusion of the dummy variables for the centres in the first model increased the c-statistic (from c = 0.68 in 2000 to c = 0.71 in 2000; and from c = 0.70 to c = 0.75 in 2001), showing that in this model the dummy variables represented not only the centre effect but still contained the effects of other confounding factors that the Charlson index alone was not fully able to extrapolate.
More interesting for our purpose was a direct comparison of the third model using EuroSCORE with the second one that presented the risk adjustment obtainable using other available administrative information.
The hospitals’ relative performances (compared with CSC D) were generally consistent, regardless of which risk adjustment model was used. As we found no changes in ranking of hospitals, our results confirmed those of Iezzoni et al. [10], i.e. that for comparing CABG death rates, either data source is potentially able to produce comparable hospital rankings.
The most powerful predictors of death in the second model were combined surgery (odds-ratio 3.02, P < 0.0001, in 2000; odds-ratio 3.24, P < 0.0001, in 2001), older age (odds-ratio 2.78, P < 0.0001, in 2000; odds-ratio 2.08, P < 0.001, in 2001), emergency (odds-ratio 2.33, P < 0.05, in 2000; odds-ratio 2.24, P < 0.05, in 2001), and previous cardiac surgery (odds-ratio 1.99, P = 0.08, in 2000; odds-ratio 3.71, P < 0.0005, in 2001). Regarding model discrimination and calibration, both the c-statistic and the Hosmer–Lemeshow statistic indicated that the model’s fit was satisfactory. The c-statistic for Models 2 and 3 was 0.78 for 2000, and 0.82 and 0.78, respectively, for 2001. Both models presented a good calibration capability.
|
Discussion |
|---|
|
TopMethodsResultsDiscussionReferences |
|---|
Relating mortality to risk remains the mainstay of any system that assesses the quality of hospital care. In the present study we have investigated the predictive performance for mortality of a risk adjustment approach based on administrative variables for CABG patients. In particular, we tested the performance of a risk adjustment approach using the Charlson comorbidity index plus other information obtainable exclusively from administrative data sources and associated with an increased risk of patients’ in-hospital mortality. Our results suggested that, with the growing completeness and accuracy of administrative data, a risk adjustment approach based on such information yields predictive performance for mortality that is comparable to that obtainable with a risk adjustment approach using EuroSCORE, whose risk factors are derived mostly from the clinical status of the patient. These results confirmed those reached in previous studies [8–10], in which the discrimination ability of clinical and administrative risk adjustment models was found to be quite similar. Iezzoni and colleagues [10] even found that code-based severity measures had better statistical performance in predicting death than severity measures based on clinical information. This paradoxical better performance of the administrative models over clinical-based ones was accounted for by the dependence of code-based measures on ICD-9-CM codes representing life-threatening conditions occurring after surgery. In order to adequately take into account the possible inclusion of codes presenting probable postoperative events, we linked hospital data across multiple episodes of care up to 1 year before the admission for CABG for the same patient. In our analysis, major efforts to distinguish between comorbidities and complications among secondary diagnoses in administrative data systems did not alter the paradoxical result stressed in previous analyses, and the statistical model based on clinical data did not have a substantially better predictive ability than the administrative approach. As suggested by Hannan et al. [8,9], our findings confirmed that adding some administrative variables considered proxies for clinical complexity to the administrative risk adjustment approach eliminated much of the difference in effectiveness of the two systems. Focusing on the health policy context of measuring CABG death rates, our study strengthened the theory that, with increasing improvements in accurate coding practice, administrative databases could provide a valuable and economical source for health planning and research. These results seem to be of particular importance if we consider the possibility of adapting and applying such an administrative approach to illnesses where a clinical risk index like EuroSCORE has not yet been developed.
We are grateful to Gianluca Fiorentini, Francesco Taroni, Roberto Grilli, Andrea Donatini and Matteo Lippi Bruni for helpful discussions. We thank the Agenzia Sanitaria of Emilia Romagna for kindly providing the databases used for this research. Any opinions expressed in this paper are those of the authors and not those of the Agenzia Sanitaria of Emilia Romagna.
|
References |
|---|
|
TopMethodsResultsDiscussionReferences |
|---|
- Black C, Roos N. Administrative data: baby or bathwater? Med Care 1998; 36: 3–5.[CrossRef][Web of Science][Medline]
- Wang P, Walker A, Tsuang M et al. Strategies for improving comorbidity measures based on Medicare and Medicaid claims data. J Clin Epidemiol 2000; 53: 571–578.[CrossRef][Web of Science][Medline]
- Sowden A, Deeks J, Sheldon T. Volume and outcome in coronary artery bypass graft surgery: true association or artefact? Br Med J 1995; 311: 151–155.
[Abstract/Free Full Text] - Romano P, Roos L, Luft H, Jollis J, Doliszny K. A comparison of administrative versus clinical data: coronary artery bypass surgery as an example. J Clin Epidemiol 1994; 47: 249–260.[CrossRef][Web of Science][Medline]
- Ghali WA, Hall RE, Rosen AK, Ash AS, Moskowitx MA. Searching for an improved clinical comorbidity index for use with ICD-9-CM administrative data. J Clin Epidemiol 1996; 49: 273–278.[CrossRef][Web of Science][Medline]
- Kieszak S, Flanders W, Kosinski A et al. A comparison of the Charlson Comorbidity Index derived from medical record data and administrative billing data. J Clin Epidemiol 1999; 52: 137–142.[CrossRef][Web of Science][Medline]
- Powel H, Lim L, Heller R. Accuracy of administrative data to assess comorbidity in patients with heart disease: an Australian perspective. J Clin Epidemiol 2001; 54: 687–693.[CrossRef][Web of Science][Medline]
- Hannan EL, Kilburn H, Lindsey L, Lewis R. Clinical versus administrative data bases for CABG surgery: does it matter? Med Care 1992; 30: 892–907.[Web of Science][Medline]
- Hannan EL, Racz MJ, Jollis JG, Peterson ED. Using Medicare claims data to assess provider quality for CABG surgery; does it work well enough? Health Serv Res 1997; 31: 659–678.[Web of Science][Medline]
- Iezzoni LI, Ash AS, Schwartz M, Landon B, Mackernan YD. Predicting in-hospital deaths from coronary artery bypass graft surgery: do different severity measures give different predictions? Med Care 1998; 36: 28–39.[CrossRef][Web of Science][Medline]
- Roques F, Gabrielle F, Michel P et al. Quality of care in adult heart surgery: proposal for a self-assessment approach based on a French multicenter study. Eur J Cardiothorac Surg 1995; 9: 433–440.[Abstract]
- Roques F, Nashef S, Michel P et al. Risk factors and outcome in European cardiac surgery: analysis of the EuroSCORE multinational database of 19030 patients. Eur J Cardiothorac Surg 1999; 15: 816–823.
[Abstract/Free Full Text] - Nashef S, Roques F, Michel P et al. European system for cardiac operative risk evaluation (EuroSCORE). Eur J Cardiothorac Surg 1999; 16: 9–13.
[Abstract/Free Full Text] - Geissler H, Hoelzl P, Marohl S et al. Risk stratification in heart surgery: comparison of six score systems. Eur J Cardiothorac Surg 2000; 17: 400–406.
[Abstract/Free Full Text] - Asimakopoulos G, Al-Ruzzeh S, Ambler G et al. An evaluation of existing risk stratification models as a tool for comparison of surgical performances for coronary artery bypass grafting between institutions. Eur J Cardiothorac Surg 2003; 23: 935–942.
[Abstract/Free Full Text] - Kawachi Y, Nakashima A, Toshima Y, Arinaga K, Kawano H. Risk stratification analysis of operative mortality in heart and thoracic aorta surgery: comparison between Parsonnet and EuroSCORE additive model. Eur J Cardiothorac Surg 2001; 20: 961–966.
[Abstract/Free Full Text] - Roques F, Nashef S, Michel P et al. Does EuroSCORE work in individual European countries? Eur J Cardiothorac Surg 2000; 18: 27–30.
[Abstract/Free Full Text] - Higgins TL, Estefanous FG, Loop FD et al. Stratification of morbidity and mortality outcome by preoperative risk factors in coronary artery bypass patients. J Am Med Assoc 1992; 267: 2344–2348.
[Abstract/Free Full Text] - Tu J, Jaglal S, Naylor D. Multicenter validation of a risk index for mortality, intensive care unit stay, and overall hospital length of stay after cardiac surgery. Circulation 1995; 91: 677–684.
[Abstract/Free Full Text] - Nashef S, Roques F, Hammill BD et al. Validation of the European system for cardiac operative risk evaluation (EuroSCORE) in North American cardiac surgery. Eur J Cardiothorac Surg 2002; 22: 101–105.
[Abstract/Free Full Text] - Nobilio L, Ugolini C. Vertical integration and contractual network in the cardiovascular sector: the experience of the Italian region Emilia Romagna. Int J Integr Care 2003; 3: 1–11. (Full text available online at http://www.ijic.org).
- Nobilio L, Ugolini C. Selective referrals in a hub and spoke institutional setting: the case of coronary angioplasty procedures. Health Policy 2003; 63: 95–107.[CrossRef][Web of Science][Medline]
- Nobilio L, Ugolini C. Different regional organisational models and quality of health care: the case of coronary artery bypass graft surgery. J Health Serv Res Policy 2003; 8: 25–32.[CrossRef][Medline]
- SAS User’s Guide, Version 5. Cary, NC: SAS Institute, 1991.
- Charlson M, Pompei P, Ales K, McKenzie C. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis 1987; 40: 373–383.[CrossRef][Web of Science][Medline]
- Romano P, Roos L, Jollis J. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol 1993; 46: 1075–1079.[CrossRef][Web of Science][Medline]
- Romano P, Roos L, Jollis J. Further evidence concerning the use of a clinical comorbidity index with ICD-9-CM administrative data. J Clin Epidemiol 1993; 46: 1085–1090.[CrossRef][Web of Science]
- Elixhauser A, Steiner C, Harris R, Coffey R. Comorbidity measures for use with administrative data. Med Care 1998; 36: 1, 8–27.[CrossRef][Web of Science][Medline]
- O’Connell R, Lim Y. Utility of the Charlson Comorbidity Index computed from routinely collected hospital discharge diagnosis code. Methods Info Med 2000; 39: 7–11.
- Hosmer DW, Lemeshow S. Applied Logistic Regression. New York, NY: John Wiley & Sons, 1989.
- Iezzoni L. Risk Adjustment for Measuring Healthcare Outcomes. Chicago, IL: Health Administrative Press, 1997.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  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]
|
|
|
|
 |
|
 F. Seccareccia, C. A. Perucci, P. D'Errigo, M. Arca, D. Fusco, S. Rosato, D. Greco, and on behalf of the research group of the Italian CAB The Italian CABG Outcome Study: short-term outcomes in patients with coronary artery bypass graft surgery Eur. J. Cardiothorac. Surg., January 1, 2006; 29(1): 56 - 62. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||