Eur J Cardiothorac Surg 2008;34:953-959. doi:10.1016/j.ejcts.2008.07.061
Copyright © 2008, European Association for Cardio-thoracic Surgery. Published by Elsevier. All rights reserved.
Logistic risk model predicting postoperative respiratory failure in patients undergoing valve surgery
Farzan Filsoufi*,
Parwis B. Rahmanian,
Javier G. Castillo,
Joanna Chikwe,
David H. Adams
Department of Cardiothoracic Surgery, Mount Sinai School of Medicine, 1190 Fifth Avenue, New York, NY 10029-1028, United States
Received 16 November 2007;
received in revised form 10 July 2008;
accepted 14 July 2008.
* Corresponding author. Tel.: +1 212 659 6820; fax: +1 212 659 6818. (Email: farzan.filsoufi{at}mountsinai.org).
 |
Abstract
|
|---|
Background: Previous studies have been unable to identify independent valve-related risk factors for postoperative respiratory failure (RF) in patients undergoing valve surgery. This study was designed to determine the incidence and predictors of RF in these patients. We also aimed to create a model based on these risk factors that could serve as a tool for the prediction of this complication. Methods: We analyzed prospectively collected data of 2808 patients (mean age 63±15 years, 43% female) who underwent valve surgery from January 1998 to December 2006. Isolated valve surgery was performed in 2007 (72%) patients whereas 801 (28%) received concomitant coronary artery bypass grafting (CABG) procedures. The main outcome investigated was RF (ventilation >72 h). Other postoperative parameters included in the analysis were hospital mortality, morbidity, length of hospital stay, discharge and late survival. Results: Respiratory failure occurred in 12.2% (n
= 342) of patients. The incidence of RF varied according to the procedures (single valve: 7.4–15.8%; multiple valves: 21.7–23.4%). The addition of CABG significantly increased the rate of RF (isolated valves: 10.8%, combined valve/CABG 15.7%, p
< 0.001). Multivariate analysis revealed preoperative renal failure, ejection fraction <30%, age >70 years, active endocarditis, emergent procedures, reoperation, diabetes, congestive heart failure, previous myocardial infarction, female gender, double aortic and mitral valve procedures, and cardiopulmonary bypass time >180 min as independent predictors of RF. Hospital mortality among patients with RF was 22.2% (n
= 76) versus 2.7% (n
= 66) in the control group (p
< 0.001). A logistic equation including the coefficients of the regression analysis was generated to calculate an individual patient’s risk for the development of RF. Predictive accuracy of the model and validation was measured (ROC area under the curve: 0.751). Long-term survival of discharged patients with RF was significantly decreased compared to those without RF. Conclusion: Respiratory failure is a common complication particularly in patients undergoing complex valve operations such as endocarditis or multiple valve procedures. The independent predictors of RF including valve-related factors allowed us to create a predictive model with great accuracy. The poor long-term survival of patients with RF underlines the need to direct more resources towards prevention and treatment of this complication.
Key Words: Respiratory failure • Valve surgery • Mortality • Morbidity • Long-term survival • Predicting model
|
1. Introduction
|
|---|
Respiratory failure (RF) is the commonest complication of cardiac surgery, with a reported prevalence of up to 15% [1,2]. It is associated with substantially increased mortality, morbidity, and decreased long-term survival [3–7]. Most previous studies reporting on the incidence and predictors of this complication have focused on patients undergoing CABG [1,3,8–10]. Over the last decade the trend in referral patterns in cardiac surgery has been toward greater numbers of more elderly patients with multiple comorbidity undergoing valve surgery, including multiple valve or combined valve/CABG procedures [11]. An understanding of predictors and outcomes of RF in this high-risk patient group is therefore increasingly relevant.
This study was designed to determine the incidence and predictors of RF in patients undergoing valve surgery. We also aimed to create a model based on these risk factors that could serve as a tool for the prediction of this complication. Finally, we studied early and late outcome of this patient population.
|
2. Patients and methods
|
|---|
2.1 Study population
Between January 1998 and December 2006, 2808 consecutive patients underwent valve or combined valve/CABG surgery at Mount Sinai Medical Center. Preoperative demographic information, intraoperative and postoperative variables and events were extracted from a computerized database using the New York State Department of Health data registry. Clinical variables were prospectively entered and retrospectively analyzed (Table 1
). The protocol was approved by our institutional review board.
The main outcome variable was postoperative RF. This complication was defined as pulmonary insufficiency requiring intubation and ventilation for a period of 
72 h, at any time during the postoperative stay. Patients who required reintubation were counted as having RF when the total ventilation time was 72 h.
Other postoperative parameters included in the analysis and their definitions are described in Table 1. Information regarding long-term survival was obtained using the web-based social security death index (www.ancestry.com). In addition, the logistic EuroSCORE was used for risk stratification (www.euroscore.org) [12].
2.2 Intra- and postoperative management
All procedures were performed through a median (full or partial) sternotomy. The detail of intraoperative management has been previously reported [13]. Following surgery, all patients were transferred to the intensive care unit (ICU). In the absence of contraindications and using pressure support methods, patients were gradually weaned off ventilation. Extubation was performed when adequate oxygenation (PaO2 >80 mmHg, SaO2 >95%) and ventilation (PaCO2 <45 mmHg, pH >7.35) with no evidence of metabolic acidosis were achieved.
2.3 Statistical analysis
Normally distributed continuous variables are presented as mean ± standard deviation (SD) and otherwise as median and interquartile range (IQR). Categorical variables are shown as the percentage of the sample.
2.4 Model development
The 
2-test, Fisher’s exact test, Student’s t-test, and Mann–Whitney U test were used as appropriate to evaluate the relationship between preoperative variables and the occurrence of postoperative RF in univariate analysis. Stepwise multivariate logistic regression was then performed to assess the influence of these variables as independent risk factors for RF [14]. The potential confounders included age, gender, ejection fraction (EF) 
30%, congestive heart failure (CHF), previous cardiac procedure, diabetes mellitus, hypertension, myocardial infarction (MI), chronic obstructive pulmonary disease (COPD), peripheral vascular disease (PVD), renal failure, previous stroke, hemodynamic instability, priority of procedure, and endocarditis. A logistic equation including the coefficients of the regression analysis was then constructed to calculate an estimation of individual patient’s risk for the development of postoperative RF. [probability of RF = (Exp
(X
x
B) + intercept(
))/(1 + Exp(X
x
B) + intercept()); where X
x
B is the coefficient B for each single confounding factor]. After the model was created, the probability of RF was calculated for each patient.
2.5 Validation of the model
For validation of the model, we included a separate cohort of patients who underwent valve surgery between January 2007 and March 2007 at our institution. The same data elements used for the creation of the model were available for the validation cohort. The probability for the development of RF was calculated for each individual patient of this cohort. We then generated receiver operating characteristic (ROC) curves and compared their area under the curves (AUC, c-statistics) to measure the predictive accuracy of the model in the study population and validation cohort. An area under the ROC curve >0.7 was assumed to be accurate [15]. Finally, the area under the ROC curve of the model was compared to the corresponding curve that resulted when the EuroSCORE was used to predict the risk of respiratory failure among the study population.
A p value <0.05 was considered as significant for all used statistical methods. Long-term survival was analyzed using Kaplan–Meier survival curves. Differences in patient characteristics were adjusted by Cox proportional hazard analysis. The statistical analyses were performed with SPSS 15 (SPSS Inc., Chicago, IL, USA) and MedCalc 9 (MedCalc Software, Mariakerke, Belgium).
|
3. Results
|
|---|
Characteristics of the 2808 patients and information regarding operative procedures are shown in Table 2
. Respiratory failure occurred in 12.2% (n
= 342) of patients, with a varying incidence according to the valves involved (single valve: 7.4–15.8%; multiple valves: 21.7–23.4%) (Fig. 1
).
3.1 Predictors of respiratory failure
Univariate analysis identified differences in patient characteristics among patients with and without RF. These variables are shown in Table 2. With respect to operative procedures, patients with RF were more likely to undergo mitral valve replacement (n
= 51, 15%), and double valve procedures (n
= 57, 17%). Patients with RF had a longer mean CPB time (212 ± 83 min vs 180 ± 69 min, p
< 0.001) and aortic cross-clamp time (156 ± 60 min vs 136 ± 50 min, p
< 0.001). The mean EuroSCORE was also significantly higher in RF patients compared to those without this complication (24% vs 13%, p
< 0.001). Stepwise multivariate logistic regression analysis revealed 12 variables as independent predictors for the occurrence of RF (Table 3
).
3.2 Predictive model and performance
The predictive model was created based on the 12 independent predictors identified by multivariate regression analysis. We first calculated the risk of RF of individual patients and compared the results with the observed rate of this complication using the c-statistic. The ROC area under the curve for the study population was 0.751 (Fig. 2
). We then applied the same model to our validation cohort and compared the predicted rate of RF with the observed occurrence of this complication. The validation cohort consisted of 90 patients (mean age 61 ± 16 years, 33% (n
= 30) female). The c-statistic for the validation cohort was 0.742. The ROC areas under the curve were not significantly different between study population and validation cohort (p
= 0.105) confirming the accuracy of the predictive model. When the EuroSCORE was used to predict respiratory failure among the study population, the area under the ROC curve was 0.71. There was a statistical significant difference when comparing our model with the results of the EuroSCORE (p
= 0.013).
3.3 Outcome of patients with respiratory failure
The overall hospital mortality among RF patients was 22.2% (n
= 76) compared to a mortality rate of 2.7% (n
= 66) in patients without RF (p
< 0.001). The causes of death among the 76 patients were multi system organ failure (n
= 53, 70%), sepsis (n
= 15, 20%), and ischemic bowel disease (n
= 8, 10%). Using multivariate analysis we were able to identify predictors of hospital mortality in the overall population (Table 4
). When the development of respiratory failure was added, these risk factors remained significant. Respiratory failure had its own independent influence on mortality (OR = 6.5, CI = 4.4–9.6, p
< 0.001).
Major complications occurred significantly more often in RF patients compared to the control group (Fig. 3
). The median length of hospital stay was significantly increased in patients with RF compared to the control group (30 days, IQR 16–51 days vs 7 days, IQR 6–11 days, p
< 0.001). The majority of patients without RF were discharged home (86%). In contrast, 56% patients with this complication were discharged to rehabilitation facilities, skilled nursing homes and acute care facilities (p
< 0.001).
3.4 Late survival
Long-term survival of discharged patients with RF was significantly decreased compared to those without RF. One-, three- and five-year crude survival rates were 72.2 ± 2.8%, 61.8 ± 3.4%, and 49.4 ± 4.3% for patients with RF compared to 96.1 ± 0.4%, 90.2 ± 0.7%, and 84.5 ± 1.0% for the control group (p
< 0.001) (Fig. 4A and B). Independent predictors of late mortality were age >70 years (hazard risk (HR = 2.9, CI = 2.3–3.7, p
< 0.001), CHF (HR = 1.9, CI = 1.5–2.5, p
< 0.001), diabetes (HR = 1.4, CI = 1.1–1.9, p
= 0.011), and COPD (HR = 1.8, CI = 1.3–2.5, p
= 0.001). The addition of RF (HR = 2.3, CI = 1.7–3.1, p
< 0.001) did not change the negative impact of these factors on late survival. Long-term survival of patients with single versus double valve procedures is shown in Fig. 4C.
View larger version (12K):
[in this window]
[in a new window]
|
Fig. 4. Unadjusted (A) and adjusted (B) Kaplan–Meier survival curves after valve surgery for patients who presented with postoperative respiratory failure compared to the control group and Kaplan–Meier survival curves for patients with single versus double valve procedures who developed postoperative respiratory failure (C).
|
|
|
4. Discussion
|
|---|
4.1 Predictors of respiratory failure
Prolonged intubation is a marker, not only of pulmonary dysfunction, but also of other major postoperative complications including low cardiac output state, re-exploration, stroke, and gastro-intestinal problems requiring laparotomy. Previous studies have identified age, female gender, previous stoke, renal failure, CHF, reoperation, urgency, and CPB time [1,3,8,16,17] as predictors of respiratory failure in patients undergoing CABG, but there is no similar information available in patients undergoing valve surgery. Our analysis revealed 12 independent factors associated with RF, among which renal failure and ejection fraction <30% were the most important.
Renal failure was the strongest predictor of postoperative RF, more than doubling the risk of prolonged ventilation. This reflects similar findings in patients undergoing CABG, where renal failure has been reported to be an independent predictor of RF with an odds ratio of between 2 and 3 [9,18]. The importance of renal failure reflects its impact at many levels. Firstly, patients with renal failure are reported to be at increased risk of adverse postoperative outcome because of a higher likelihood of delayed investigation, referral, and intervention. Secondly renal failure is associated with an increased atherosclerotic burden which predisposes these patients to ischemic complications such as stroke and gastrointestinal ischemia, which may require prolonged intubation [13,19]. Finally renal failure adversely affects pulmonary function through an increased risk of pulmonary edema, sepsis and metabolic dysfunction [20].
Preoperative left ventricular dysfunction (EF <30%) more than doubled the risk of postoperative RF in this cohort. The adverse effect of left ventricular dysfunction may reflect its impact at several levels, not all of which are accounted for in multivariate analysis. Patients with low EF are more likely to be older, and have advanced valvular heart disease with associated pulmonary hypertension. They are also more likely to suffer from a prolonged postoperative low cardiac output state, during which time extubation may be delayed until hemodynamic status can be optimized [1,18,21].
In our analysis, emergent procedures were associated with almost double the risk of postoperative RF. Rankin et al. reported that acute presentation in patients undergoing valvular surgery was associated with a 2-fold increase in operative mortality and a significant increase in the rate of other major morbidities [2]. They observed a 42% increase in the incidence of acute presentation of valve surgery patients during the last decade, from just over a fifth to almost a third of patients. This is one of the few potentially avoidable risk factors for respiratory failure. A proactive approach to early investigation, referral and surgery has contributed to the low rate of emergency presentation of fewer than 3% in our cohort.
In addition to risk factors for RF previously documented in patients undergoing CABG, we were able to identify valve-related independent predictors of this complication for the first time. Active endocarditis and double valve procedures increased the risk of RF significantly with an odds ratio of 2.1 and 1.5, respectively. Patients with active endocarditis are more likely to have a course complicated by systemic inflammatory response syndrome and associated organ dysfunction including acute lung injury. Furthermore, because of acute onset of valvular incompetence, acute bacterial endocarditis is more likely to be associated with pulmonary edema, which may predispose patients to postoperative respiratory dysfunction. Patients with double valve procedures often present complex pathology requiring prolonged cardiopulmonary bypass and cross-clamp times. Additionally, pulmonary hypertension is more prevalent in these patients compared to patients with single valve disease [2].
4.2 Predictive model
We were able to generate and validate a unique predictive model for the occurrence of respiratory failure based on 12 independent predictors. Because our study included a large and heterogeneous group of valve patients, the predictive model is applicable to the entire spectrum of adults undergoing valvular surgery. The model is simple to use and is a useful tool for assisting clinical decision making, informing consent, and planning resource allocation. For example, a 50-year-old male with no additional risk factors undergoing an isolated AVR has a probability of RF of 2.5%, compared to a 76-year-old female with previous cardiac surgery, diabetes, and a preoperative creatinine of 2.9 mg/dl undergoing the same procedure who has a 40% risk of RF. If she required an additional mitral valve procedure with a CPB time exceeding 180 min, her risk increases to 57.8%.
4.3 Long-term survival
The majority of previous studies of RF following cardiac surgery do not provide long-term results in this patient population. In our study, long-term survival was significantly reduced in RF patients (1-year and 5-year survival: 72.2 ± 2.8% and 49.4 ± 4.3% vs 96.1 ± 0.4% and 84.5 ± 1.0% (p
< 0.001)). The burden of this complication continued beyond the initial hospitalization with a significantly higher rate of death during the first postoperative year. By the end of the first year, there was a 25% absolute survival difference between patients who did not have RF and those who developed this complication. One other study reported similar findings in patients with RF following cardiac surgery with a 5-year survival reduced from 89% in the control group to 56% in patients with RF [18]. The survival of patients who underwent double valve procedures and developed RF was even worse: less than a third of this patient subgroup was still alive 5 years after surgery. Although information on cause of late death was not available to us, it is likely that the relationship between respiratory failure and poor long-term survival reflects the complications and associated comorbidities that resulted in prolonged intubation described above, rather than the direct impact of respiratory failure per se.
4.4 Strengths and limitations
The study includes a large cohort of patients who underwent all types of adult cardiac valve procedures and the findings are therefore applicable to most valve surgery patients. Our predictive model takes into consideration several preoperative characteristics and valve-related factors and can be reliably applied to a heterogeneous group of patients with valvular heart disease. The data provided in this study were extracted from the NYSDH registry, a state mandated database with external audit, and therefore provides very accurate information about perioperative variables. However, this was a retrospective observational study and therefore conclusions are necessarily limited in their application. Our study did not examine some previously reported risk factors for the occurrence of RF such as smoking [1]. Furthermore, clinical outcome analysis focused on postoperative mortality and morbidity and we were not able to provide information on late complications, quality of life, and cause of death following discharge.
|
References
|
|---|
- Branca P, McGaw P, Light R. Factors associated with prolonged mechanical ventilation following coronary artery bypass surgery. Chest 2001;119:537-546.[CrossRef][Medline]
- Rankin JS, Hammill BG, Ferguson Jr. TB, Glower DD, O’Brien SM, DeLong ER, Peterson ED, Edwards FH. Determinants of operative mortality in valvular heart surgery. J Thorac Cardiovasc Surg 2006;131:547-557.[Abstract/Free Full Text]
- Canver CC, Chanda J. Intraoperative and postoperative risk factors for respiratory failure after coronary bypass. Ann Thorac Surg 2003;75:853-857[Discussion 857–8].[Abstract/Free Full Text]
- Hein OV, Birnbaum J, Wernecke KD, Konertz W, Jain U, Spies C. Three-year survival after four major post-cardiac operative complications. Crit Care Med 2006;34:2729-2737.[CrossRef][Medline]
- Kollef MH, Wragge T, Pasque C. Determinants of mortality and multiorgan dysfunction in cardiac surgery patients requiring prolonged mechanical ventilation. Chest 1995;107:1395-1401.[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;28:3847-3853.[CrossRef][Medline]
- Combes A, Costa MA, Trouillet JL, Baudot J, Mokhtari M, Gibert C, Chastre J. Morbidity, mortality, and quality-of-life outcomes of patients requiring > or = 14 days of mechanical ventilation. Crit Care Med 2003;31:1373-1381.[CrossRef][Medline]
- Cohen AJ, Katz MG, Frenkel G, Medalion B, Geva D, Schachner A. Morbid results of prolonged intubation after coronary artery bypass surgery. Chest 2000;118:1724-1731.[CrossRef][Medline]
- Legare JF, Hirsch GM, Buth KJ, MacDougall C, Sullivan JA. Preoperative prediction of prolonged mechanical ventilation following coronary artery bypass grafting. Eur J Cardiothorac Surg 2001;20:930-936.[Abstract/Free Full Text]
- Yende S, Wunderink R. Causes of prolonged mechanical ventilation after coronary artery bypass surgery. Chest 2002;122:245-252.[CrossRef][Medline]
- Ferguson Jr. TB, Hammill BG, Peterson ED, DeLong ER, Grover FL. A decade of change--risk profiles and outcomes for isolated coronary artery bypass grafting procedures, 1990–1999: a report from the STS National Database Committee and the Duke Clinical Research Institute. Society of Thoracic Surgeons. Ann Thorac Surg 2002;73:480-489[Discussion 489–90].[Abstract/Free Full Text]
- Nashef SA, Roques F, Michel P, Gauducheau E, Lemeshow S, Salamon R. European system for cardiac operative risk evaluation (EuroSCORE). Eur J Cardiothorac Surg 1999;16:9-13.[Abstract/Free Full Text]
- Filsoufi F, Rahmanian PB, Castillo JG, Scurlock C, Legnani PE, Adams DH. Predictors and outcome of gastrointestinal complications in patients undergoing cardiac surgery. Ann Surg 2007;246:323-329.[CrossRef][Medline]
- Hosmer DW, Lemeshow S. Applied logistic regression. New York: Wiley; 1989.
- Lemeshow S, Le Gall JR. Modeling the severity of illness of ICU patients. A systems update. JAMA 1994;272:1049-1055.[Abstract/Free Full Text]
- Bando K, Sun K, Binford RS, Sharp TG. Determinants of longer duration of endotracheal intubation after adult cardiac operations. Ann Thorac Surg 1997;63:1026-1033.[Abstract/Free Full Text]
- Spivack SD, Shinozaki T, Albertini JJ, Deane R. Preoperative prediction of postoperative respiratory outcome. Coronary artery bypass grafting. Chest 1996;109:1222-1230.[CrossRef][Medline]
- Rajakaruna C, Rogers CA, Angelini GD, Ascione R. Risk factors for and economic implications of prolonged ventilation after cardiac surgery. J Thorac Cardiovasc Surg 2005;130:1270-1277.[Abstract/Free Full Text]
- Mangi AA, Christison-Lagay ER, Torchiana DF, Warshaw AL, Berger DL. Gastrointestinal complications in patients undergoing heart operation: an analysis of 8709 consecutive cardiac surgical patients. Ann Surg 2005;241:895-901[Discussion 901–4].[CrossRef][Medline]
- Nagachinta T, Stephens M, Reitz B, Polk BF. Risk factors for surgical-wound infection following cardiac surgery. J Infect Dis 1987;156:967-973.[Medline]
- Topkara VK, Cheema FH, Kesavaramanujam S, Mercando ML, Cheema AF, Namerow PB, Argenziano M, Naka Y, Oz MC, Esrig BC. Coronary artery bypass grafting in patients with low ejection fraction. Circulation 2005;112:I344-I350.[Medline]
Top
Abstract
1. Introduction
