Interact CardioVasc Thorac Surg 2009;9:654-658. doi:10.1510/icvts.2008.199521
© 2009 European Association of Cardio-Thoracic Surgery
Institutional report - Cardiac general |
Predicting prolonged intensive care unit length of stay in patients undergoing coronary artery bypass surgery – development of an entirely preoperative scorecard
Christine Herman,
Wojtek Karolak,
Alexandra M. Yip,
Karen J. Buth,
Ansar Hassan and
Jean-Francois Légaré*
Division of Cardiac Surgery, Queen Elizabeth II Health Sciences Center, 1796 Summer Street, Rm. 2629, Halifax, Nova Scotia, B3H 3A7, Canada
Received 30 November 2008;
received in revised form 13 July 2009;
accepted 14 July 2009
 Poster Presentation: American Heart Association, 2007 presented Nov 6th, 2007.
*Corresponding author. Tel.: +1-902-473-3808; fax: +1-902-473-4448.
E-mail address: jean.legare{at}cdha.nshealth.ca (J.-F. Légaré).
 |
Abstract
|
|---|
We sought to develop a predictive model based exclusively on preoperative factors to identify patients at risk for PrlICULOS following coronary artery bypass grafting (CABG). Retrospective analysis was performed on patients undergoing isolated CABG at a single center between June 1998 and December 2002. PrlICULOS was defined as initial admission to ICU exceeding 72 h. A parsimonious risk-predictive model was constructed on the basis of preoperative factors, with subsequent internal validation. Of 3483 patients undergoing isolated CABG between June 1998 and December 2002, 411 (11.8%) experienced PrlICULOS. Overall in-hospital mortality was higher among these patients (14.4% vs. 1.2%, P 0.0001). The following variables were found to be independent predictors of PrlICULOS: increased age, recent myocardial infarction, preoperative renal failure, cerebral and/or peripheral vascular disease, chronic obstructive pulmonary disease, ejection fraction <40%, previous CABG, triple-vessel and/or left main disease, NYHA class IV symptoms and urgent or emergent status. Subsequent validation of this model demonstrated a c-statistic of 78%. An internally-validated, risk predictive model of PrlICULOS in patients undergoing CABG was constructed. Based on preoperative clinical factors, a scorecard was developed allowing identification of these patients prior to surgery and allowing for strategies aimed at optimizing hospital resources.
Key Words: Cardiac surgery; Intensive care units; Coronary artery bypass
|
1. Introduction
|
|---|
Prolonged intensive care unit (ICU) length of stay (LOS), or PrlICULOS, following cardiac surgery is a source of great expense to the health care system [1]. Increased costs resulting from the consumption of excess ICU resources emphasize the importance of being able to identify those patients at risk of experiencing PrlICULOS in advance of the planned intervention.
EuroSCORE and Parsonnet tested the ability of established risk-predictive to predict PrlICULOS, or have retrospectively identified independent risk factors of PrlICULOS [2–8]. Few, however, have developed internally-validated, risk-predictive models with scorecards designed to easily identify patients at risk of PrlICULOS. Ghotkar et al. developed a scorecard predicting PrlICULOS which included an intra-operative variable, thus limiting its use as an exclusively preoperative predictive tool [7]. Others have developed predictive models based solely on preoperative variables but have neglected to develop scorecards, thereby restricting their real-time applicability [3].
The purpose of this study was to identify CABG patients at risk of PrlICULOS using a predictive model, and to develop a clinically-relevant patient risk scorecard capable of functioning as a preoperative resource planning tool.
|
2. Materials and methods
|
|---|
2.1. Data sources and study population
All patients who underwent isolated CABG at the Queen Elizabeth II Health Sciences Centre (QEII HSC) in Halifax, Canada, between June 1998 and December 2002, were identified using the Maritime Heart Center (MHC) registry. The MHC registry is a prospectively-collected, detailed clinical database containing pertinent pre-, intra- and postoperative data on all cardiac surgical cases performed at the QEII HSC from March 1995 until the present.
In this study, the ICU admits exclusively cardiac surgical patients where all patients are discharged to an intermediate care unit. Discharge criteria are determined by the Intensivist on call and all patients are exposed to early extubation (<6 h) and fast-track parameters if clinical stability allows. PrlICULOS was defined as ICU LOS of >72 h, while normal ICU LOS was defined as ICU LOS lasting 72 h. Only LOS resulting from the patient's initial admission to ICU following surgery was used in determining overall ICU LOS. Time spent in ICU as the result of readmission to ICU was not considered. In order to ensure that patients included in the study were eligible to experience either PrlICULOS or normal ICU LOS, patients who had not yet been discharged from ICU following their initial admission but who died within 72 h of surgery were excluded from the analysis.
2.2. Variable selection
Preoperative variables of interest included age, sex, smoking history, diabetes, hypercholesterolemia, hypertension, renal failure (preoperative serum creatinine >176 µmol/l), cerebrovascular and/or peripheral vascular disease (CVD/PVD), chronic obstructive pulmonary disease (COPD), left ventricular dysfunction (ejection fraction <40%), recent MI (within 21 days), prior percutaneous coronary intervention (PCI), prior CABG, number of diseased, functional class as defined by the New York Heart Association (NYHA) classification system, and urgency status.
Intra-operative variables of interest included whether the surgery was performed with or without cardiopulmonary bypass (on-pump vs. off-pump), left internal mammary artery (LIMA) use, whether or not total arterial revascularization was employed, number of distal anastomoses, cross-clamp time and total bypass time.
Finally, adverse postoperative outcomes considered included prolonged ventilation (>24 h), blood product transfusion, permanent stroke, deep sternal wound infection, peri-operative myocardial infarction, re-operation for either bleeding, graft occlusion or any cardiac cause and in-hospital mortality.
2.3. Statistical analysis
Patients with PrlICULOS were compared to those with normal ICU LOS on the basis of pre-, intra- and postoperative characteristics. Continuous variables were compared using the two-sided t-test, whereas categorical variables were compared using the  2 and Fisher's exact test. Number of distal anastomoses, as an ordinal variable, was compared using the Mann–Whitney–Wilcoxon test. Univariate preoperative predictors of PrlICULOS with a P-value <0.20 were entered into a multivariate logistic regression model. Using a bootstrapping procedure, backward elimination was conducted on 200 subsamples, and variables retained at P-value <0.05 in at least 50% of these subsamples were identified as independent risk factors of PrlICULOS. A c-statistic was calculated as a measure of the sensitivity and specificity of the final logistic regression model, applied to the entire sample. The 95% confidence interval (CI) was obtained from the 2.5th and 97.5th percentiles of the bootstrap distribution. The final predictive model was subsequently validated on 2024 consecutive isolated CABG surgeries performed between 1 January 2003 and 30 September 2005, at the QEII HSC employing the same exclusion criteria as outlined above. Statistical significance was indicated by a P-value <0.05 in our analyses, all of which were performed using the SAS software package version 9.1.3 (SAS, Cary, North Carolina).
2.4. Scorecard development
We applied the predictive model to the derivation sample to calculate a predicted probability of PrlICULOS for each patient and then sorted the patients by increasing probability. Next, we assigned a weighted score to each risk factor and we adjusted these weights up or down by increments of 0.5 to ensure that total scores increased correspondingly with categories of predicted probabilities.
|
3. Results
|
|---|
3.1. Patient population
During the study period, a total of 3523 patients underwent isolated CABG. Of these, 40 died within 72 h prior to being discharged from ICU and were thus excluded. The remaining 3483 patients formed the final study population, of whom 11.8% (n=411) experienced PrlICULOS. Patients who experienced PrlICULOS had significantly greater co-morbid disease (Table 1).
A greater percentage of patients with PrlICULOS underwent on-pump CABG with longer cross-clamp and cardiopulmonary bypass times (Table 2). They were less likely to have received a LIMA graft or to have undergone total arterial grafting. Patients who eventually experienced PrlICULOS were more likely to be transferred from operating room with inotrope support.
3.2. In-hospital outcomes
The patients with PrlICULOS spent a median of 141 h in ICU (IQR 95–265 h) in comparison to patients with normal ICU LOS who spent a median of 22 h in ICU (IQR 20–26). Overall in-hospital mortality was 14.4% in patients with PrlICULOS as compared to 0.2% in patients with normal ICU LOS (P0.0001) (Table 3).
3.3. Multivariable analysis
Significant, independent risk factors for PrlICULOS are listed in Table 4. The Hosmer–Lemeshow goodness of fit P-value for this model was 0.41, and the c-statistic of the logistic regression model was 0.78 (95% CI 0.76–0.81).
3.4. Model validation
The patient characteristics of the validation cohort were similar to the model derivation group with notable exceptions (Table 5). Using the coefficients from the original derivation model, the c-statistic of the model, when employed in the validation group, was 0.78 (95% CI 0.75–0.81).
3.5. Scorecard
The resulting preoperative risk scorecard is presented in Table 6. We categorized the total scores and risk probabilities into clinically significant risk intervals allowing for simple and easy assignment of percent risk for PrlICULOS.
|
4. Discussion
|
|---|
The purpose of this study was to develop and validate a risk-predictive model for PrlICULOS in patients undergoing isolated CABG solely on the basis of preoperative characteristics. The following preoperative risk factors for PrlICULOS were identified: age, recent MI, preoperative renal failure, CVD/PVD, COPD, EF<40%, prior CABG, left main and/or three vessel disease, NYHA class IV symptoms, and urgency status. A c-statistic of 0.78 in both the derivation and validation patient subsets demonstrated the model's ability to adequately discriminate between patients who were likely to experience PrlICULOS and those who were not. The model was used to develop a scorecard intended for the prospective preoperative identification of patients at risk for PrlICULOS.
While PrlICULOS following cardiac surgery comes at a tremendous financial cost [1], its negative effects on short- and long-term survival, functional capacity and health-related quality of life have been shown to be significant as well [9–14]. In this study, we demonstrated increased overall hospital LOS and increased in-hospital outcomes among patients with PrlICULOS. In a time of rising health care expenditures, efforts to reduce the financial and clinical impact of PrlICULOS are needed. The ability to identify patients at risk for PrlICULOS is a critical step in developing ways to optimize utilization of limited health care resources.
Of the previous studies that either tested existing risk-predictive models to assess risk of PrlICULOS or retrospectively identified independent risk factors for PrlICULOS [2–8], few created risk-predictive models with accompanying scorecards. The scorecard put forth by Ghotkar et al. included an intra-operative variable in addition to baseline characteristics, thus limiting its use as a preoperative predictive tool [7]. The scorecard developed in this study was based on baseline characteristics alone and is intended for use as a preoperative screening tool. By identifying high-risk patients prior to surgery and restricting how many of these patients would be operated on during a given period of time, ICU beds are less likely to be occupied with patients requiring long-term intensive care. Through application of prediction models designed to better estimate ICU LOS in patients undergoing esophagectomy for carcinoma, Van Houdenhoven et al. were able to prevent cancellations in 15% of their surgical cohort [14].
This study was not without its limitations. First, the risk model and scorecard created applied only to isolated CABG patients. Isolated CABG patients were chosen as they represented a homogeneous population which enabled the creation of a more stable risk-predictive model. In addition, >60% of patients undergoing cardiac surgery at our institution over the past two years and >50% of patients experiencing PrlICULOS during that time underwent isolated CABG, thus supporting the utility of this model in effecting significant changes in resource utilization. Secondly, only time spent during the initial admission to ICU was considered when defining ICU LOS, a definition that did not take into account time spent in ICU as the result of readmission. This may have resulted in an underestimation of patients who experienced an overall ICU LOS of >72 h. However, in a previous study from our institution, only 3.6% of patients undergoing CABG had ICU readmission, and the median initial ICU LOS was 6 days in these patients as compared to 1 day in patients who were not readmitted [15]. Thus, the percentage of patients who were considered to have experienced PrlICULOS was not likely affected by excluding time spent in ICU during a readmission. Thirdly, the predictive model was validated using a cohort of patients from our center. External validation would establish the predictive ability and utility of this model across a variety of cardiac surgical populations.
In conclusion, we developed a multivariate risk-predictive model for PrlICULOS based exclusively on preoperative variables. The resulting scorecard will allow for the simple identification of patients at risk for PrlICULOS prior to surgery. It is anticipated that this tool may be used to devise strategies aimed at optimizing resource utilization.
|
References
|
|---|
- Kurki TS, Hakkinen U, Lauharanta J, Ramo J, Leijala M. Evaluation of the relationship between preoperative risk scores, postoperative and total length of stays and hospital costs in coronary bypass surgery. Eur J Cardiothorac Surg 2001 Dec;20:1183–1187.[Abstract/Free Full Text]
- Nilsson J, Algotsson L, Hoglund P, Luhrs C, Brandt J. EuroSCORE predicts intensive care unit stay and costs of open heart surgery. Ann Thorac Surg 2004 Nov;78:1528–1534.[Abstract/Free Full Text]
- Tu JV, Jaglal SB, Naylor CD. 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 Feb 1;91:677–684.[Abstract/Free Full Text]
- Hein OV, Birnbaum J, Wernecke K, England M, Konertz W, Spies C. Prolonged intensive care unit stay in cardiac surgery: risk factors and long-term-survival. Ann Thorac Surg 2006 Mar;81:880–885.[Abstract/Free Full Text]
- Janssen DP, Noyez L, Wouters C, Brouwer RM. Preoperative prediction of prolonged stay in the intensive care unit for coronary bypass surgery. Eur J Cardiothorac Surg 2004 Feb;25:203–207.[Abstract/Free Full Text]
- Ranucci M, Bellucci C, Conti D, Cazzaniga A, Maugeri B. Determinants of early discharge from the intensive care unit after cardiac operations. Ann Thorac Surg 2007 Mar;83:1089–1095.[Abstract/Free Full Text]
- Ghotkar SV, Grayson AD, Fabri BM, Dihmis WC, Pullan DM. Preoperative calculation of risk for prolonged intensive care unit stay following coronary artery bypass grafting. J Cardiothorac Surg 2006;1:14.[CrossRef][Medline]
- Lawrence DR, Valencia O, Smith EE, Murday A, Treasure T. Parsonnet score is a good predictor of the duration of intensive care unit stay following cardiac surgery. Heart (British Cardiac Society) 2000 Apr;83:429–432.[Medline]
- Bapat V, Allen D, Young C, Roxburgh J, Ibrahim M. Survival and quality of life after cardiac surgery complicated by prolonged intensive care. J Cardiac Surg 2005, May–Jun, 20:212–217.[CrossRef][Medline]
- Bashour CA, Yared JP, Ryan TA, Rady MY, Mascha E, Leventhal MJ, Starr NJ. Long-term survival and functional capacity in cardiac surgery patients after prolonged intensive care. Crit Care Med 2000 Dec;28:3847–3853.[CrossRef][Medline]
- Gaudino M, Girola F, Piscitelli M, Martinelli L, Anselmi A, Della Vella C, Schiavello R, Possati G, Schiavello R, Possati G. Long-term survival and quality of life of patients with prolonged postoperative intensive care unit stay: unmasking an apparent success. J Thoracic Cardiovascul Surg 2007 Aug;134:465–469.[CrossRef]
- Heimrath OP, Buth KJ, Legare JF. Long-term outcomes in patients requiring stay of more than 48 hours in the intensive care unit following coronary bypass surgery. J Crit Care 2007 Jun;22:153–158.[CrossRef][Medline]
- Mazzoni M, De Maria R, Bortone F, Parolini M, Ceriani R, Solinas C, Arena V, Parodi O. Long-term outcome of survivors of prolonged intensive care treatment after cardiac surgery. Ann Thorac Surg 2006 Dec;82:2080–2087.[Abstract/Free Full Text]
- Van Houdenhoven M, Nguyen DT, Eijkemans MJ, Steyerberg EW, Tilanus HW, Gommers D, Wullink G, Bakker J, Kazemier G. Optimizing intensive care capacity using individual length-of-stay prediction models. Crit Care 2007 Mar 27;11:R42.[CrossRef][Medline]
- Bardell T, Legare JF, Buth KJ, Hirsch GM, Ali IS. ICU readmission after cardiac surgery. Eur J Cardiothorac Surg 2003 Mar;23:354–359.[Abstract/Free Full Text]
|
|
Top
Abstract
1. Introduction
