In-Hospital Mortality After Cardiac Surgery: Patient Characteristics, Timing, and Association With Postoperative Length of Intensive Care Unit and Hospital Stay Michael Mazzeffi, MD, MPH, Joel Zivot, MD, Timothy Buchman, MD, PhD, and Michael Halkos, MD, MS Department of Anesthesiology, University of Maryland, Baltimore, Maryland; Departments of Anesthesiology, Cardiothoracic Surgery, and Emory Center for Critical Care, Emory University Hospital, Emory University, Atlanta, Georgia
Background. It is important to characterize in-hospital mortality after cardiac surgery and understand the relationships between postoperative length of intensive care unit stay, postoperative length of hospital stay, and the likelihood of in-hospital mortality. Methods. We retrospectively identified all cardiac surgery cases that resulted in in-hospital mortality over an 8-year period at a single center. For these subjects we collected demographic data, preoperative comorbidities, and postoperative complications. We performed stepwise multivariate linear regression to determine which postoperative complications were associated with mortality timing. We also analyzed the relationships between postoperative length of intensive care unit stay, postoperative length of hospital stay, and in-hospital mortality in all patients (including survivors) who had cardiac surgery during the same time period. Finally, we calculated the daily incremental observed mortality rate for patients in the hospital up to postoperative day 50. Results. Six hundred twenty-one in-hospital mortalities occurred among 18,348 patients during the study
period (3.4%). Four postoperative complications were associated with mortality timing. Cardiac arrest had a negative association with the number of days until mortality, while deep sternal wound infection, stroke, and pneumonia had a positive association (all p < 0.05). Postoperative complications explained 15% of the variability in mortality timing (R2 model [ 0.15). The odds ratio for in-hospital mortality was 1.033 for each postoperative day in the hospital and 1.071 for each postoperative day in the intensive care unit (both p < 0.05). Conclusions. Most in-hospital mortality occurs during the first week after cardiac surgery with few mortalities occurring after a protracted hospital course. Postoperative complications have a limited ability to explain the variability in mortality timing. Increased length of postoperative intensive care unit stay and hospital stay after cardiac surgery are associated with an increased likelihood of in-hospital mortality.
A
surgeons, ICU physicians, and palliative care physicians as it may allow for better information to be given to family members when making complex decisions about end of life care. One common perception is that longer ICU or hospital stays are directly related to in-hospital mortality; however this supposition is not fully validated in cardiac surgery patients and it remains unclear on what ICU or hospital day number the risk for mortality reaches a significant level. The principal objective of our study was to better characterize in-hospital mortality after cardiac surgery by identifying characteristics of patients who had a mortality, describing its timing, and determining the relationships between postoperative length of ICU stay, postoperative length of hospital stay, and the likelihood of in-hospital mortality. We also sought to determine whether the occurrence of common postoperative complications could explain much of the variability in mortality timing.
ccording to the Society for Thoracic Surgeons (STS) database, the 30-day mortality rate for over 774,000 coronary artery bypass surgeries was 2.3% and for over 109,000 isolated valve surgeries was 3.4% [1, 2]. Similar 30-day mortality rates (from 1% to 4%) have been reported in other large cardiac surgery data sets [3, 4]. Despite these 30-day mortality data, in-hospital mortality after cardiac surgery has not been well characterized, including patient characteristics, timing, and the relationships between postoperative length of intensive care unit (ICU) stay, length of hospital stay, and the likelihood of in-hospital mortality. Better characterizing in-hospital mortality should be of great interest to cardiac
Accepted for publication Oct 11, 2013. Address correspondence to Dr Mazzeffi, Department of Anesthesiology, University of Maryland, 22 S Greene St, Baltimore, MD 21201; e-mail: mmazzeffi@anes.umm.edu.
Ó 2013 by The Society of Thoracic Surgeons Published by Elsevier Inc
(Ann Thorac Surg 2013;-:-–-) Ó 2013 by The Society of Thoracic Surgeons
0003-4975/$36.00 http://dx.doi.org/10.1016/j.athoracsur.2013.10.040
2
MAZZEFFI ET AL IN-HOSPITAL DEATH AND CARDIAC SURGERY
Patients and Methods Data Collection The Institutional Review Board at Emory University approved the study and granted a waiver of informed consent for data collection. All cardiac surgery cases performed at Emory University and its associated hospitals between February 1, 2004 and March 31, 2012 were identified by retrospective query of the Emory institutional STS database. For all patients we recorded the total number of postoperative ICU days, total number of postoperative hospital days, in-hospital mortality status, and discharge location. As a subset of the initial cohort we identified patients who had in-hospital mortality after their surgery. For these patients we collected multiple variables including demographics, common preoperative comorbidities, and preoperative lab values (age, body mass index, gender, chronic lung disease, diabetes, dyslipidemia, endocarditis, hypertension, New York Heart Association class, peripheral arterial disease, STS predicted mortality rate, previous cardiac surgery, previous myocardial infarction, preoperative creatinine, preoperative hematocrit, preoperative ejection fraction, cardiopulmonary bypass time, and procedure type). We also collected common postoperative complications (acute renal failure, cardiac arrest, cardiac reoperation for bleeding, cardiac reoperation other than bleeding, deep sternal wound infection, stroke, and pneumonia). Definitions for these variables were based on definitions used by the national STS database, which can be found in the STS training manual (www. sts.org/sts-national-database/database-managers/adultcardiac-surgery¼database/data-collection). Multiple versions of the STS database were used during initial data collection including 2.41, 2.52, 2.61, and 2.73. Definitions for variables were those that were used by the STS database at the time of the initial data collection. A significant number of patients were missing some data in the STS database (approximately 16%). In addition to the STS database variables, we also recorded the number of days until mortality occurred and its principal cause. The principal cause of mortality was classified as cardiac, respiratory, neurologic, infectious, or other, based upon the primary cause of mortality listed on the patients’ death certificate. For all analyses postoperative day 0 was defined as ending at 2400 on the same calendar day as the patients’ surgery. Subsequent postoperative days began at 0000 and ended at 2400.
Ann Thorac Surg 2013;-:-–-
Next, we created an explanatory model to describe the associations between postoperative complications and mortality timing. Both Pearson correlation analysis and stepwise multivariate linear regression were performed to determine which complications had a significant association with the number of days until death. Pearson correlation values were reported with respective p values. For the regression model, all previously mentioned postoperative complications (see Patients and Methods: Data collection) were included as independent variables and the dependent variable was the number of days until mortality occurred. A p value of 0.05 was required for entry into the model and was considered significant in the final model. Because there were a significant number of missing data points in the cohort, we used the missing data analysis; expectation maximization function in SPSS 21 to complete the data set prior to analysis. Missing data were considered to be missing completely at random because there was no logical association between the occurrence of postoperative complications (independent variables) and missing data points. Beta values with 95% confidence intervals were reported for regression coefficients included in the final model and the coefficient of determination was reported for the model. In order to determine the relationships between postoperative length of ICU stay, postoperative length of hospital stay, and in-hospital mortality we performed univariate logistic regression. In the first analysis, postoperative length of ICU stay was treated as a continuous independent variable and in-hospital mortality was treated as a dichotomous dependent variable. In the second analysis, postoperative length of hospital stay was treated as a continuous independent variable and inhospital mortality was treated as a dichotomous dependent variable. For these analyses, odds ratios with 95% confidence intervals were reported which reflect the odds of death per postoperative day spent in the ICU or hospital. Finally, we created a risk table containing the daily incremental observed mortality rate for each postoperative day in the hospital. This was done by first counting the number of patients remaining in the hospital on a given hospital day (denominator) and then looking at the number of those patients who subsequently died during their hospitalization (numerator). We calculated the daily incremental observed mortality rate for every day up to postoperative day 50 and then for individual days 75, 100, 125, and 150.
Statistical Methods
Results
All statistical analyses for the study were carried out using SPSS 21 (SPSS Inc, Chicago IL). First, we performed descriptive statistics on the in-hospital mortality cohort to determine the time distribution of mortality as well as the distributions for other study variables (demographics, preoperative lab values, preoperative comorbidities, and postoperative complications). Non-normally distributed continuous data were described using median [Q1–Q3] and nominal data were described using number (%).
The overall in-hospital mortality rate for the entire cohort was 3.4% (621 of 18,348). The most common principal causes of death were cardiac death 62.1%, respiratory failure 11.8%, infectious 7.7%, and neurologic injury 6.0%. The mode day (most common day) of death was postoperative day 0 with 15.0% of all deaths occurring on that day. The median day of death was 7. Of all deaths, 51.2% occurred in the first 7 days, while only 59 deaths or 9.5% occurred after postoperative day 30; 89.0% of patients
Ann Thorac Surg 2013;-:-–-
MAZZEFFI ET AL IN-HOSPITAL DEATH AND CARDIAC SURGERY
who were discharged went home, 7.8% went to transitional care facilities, 1.6% went to other hospitals, 1.4% went to nursing homes, and 0.2% went to hospice. Table 1 shows the characteristics of patients who had an inhospital death and Table 2 shows the common postoperative complications that occurred in them. Table 3 shows the results of the correlation studies and multivariate linear regression analysis for the in-hospital mortality cohort. Four variables (cardiac arrest, deep sternal wound infection, stroke, and pneumonia) were associated with mortality timing in the final model (all had p < 0.05). The coefficient of determination for the linear regression model was 0.15. Table 4 shows the results of the univariate logistic regression analyses. Both postoperative length of ICU
Table 1. Characteristics of Patients Who Had In-Hospital Mortality Variable Demographics: Age (years) Body mass index (kg/m2) Male gender, no (%) Comorbidities: Chronic lung disease – no (%) Diabetes mellitus – no (%) Dyslipidemia – no (%) Endocarditis – no (%) Hypertension – no (%) NYHA class – no (%) I II III IV Peripheral arterial disease – no (%) Predicted mortality (%) Previous cardiac surgery – no (%) Previous myocardial infarction – no (%) Lab values: Baseline creatinine (mg/dL) Hematocrit (g/dL) Left ventricular ejection fraction Surgical variables: CPB time (minutes) Type of procedure – no (%): CABG: AVR: AVR þ CABG: Isolated MVR: Other:
Median (Q1, Q3) or no (%)
66 (56, 76) 26 (23, 32) 368/621 (59.3) 159/605 214/609 150/263 44/601 524/617
(26.3) (35.1) (57.0) (7.3) (84.9)
55/487 (11.3) 93/487 (19.1) 126/487 (25.9) 213/487 (43.7) 132/601 (22.0) 7 (3, 16) 318/608 (52.3) 110/267 (41.2) 1.3 (1.0, 1.8) 35 (31, 40) 0.50 (0.35, 0.60) 159 (112, 219) 133/605 52/605 34/605 21/605 365/605
(22.0) (8.6) (5.6) (3.5) (60.3)
Total n ¼ 621, data were incomplete for some variables. Non-normally distributed continuous variables are represented by median (Q1–Q3) and nominal variables are listed as n (%). AVR ¼ aortic valve replacement; CABG ¼ coronary artery bypass graft; CPB ¼ cardiopulmonary bypass; NYHA ¼ New York Heart Association; MVR ¼ mitral valve repair or replacement.
3
Table 2. Complications in Patients Who Had an In-Hospital Death Variable Day of death (postoperative) Total ICU days Acute renal failure – no (%) Cardiac arrest – no (%) Cardiac reoperation for bleeding – no (%) Cardiac reoperation other than bleeding – no (%) Deep sternal wound infection – no (%) Stroke – no (%) Pneumonia – no (%)
Median (Q1, Q3) or no (%) 7 (1, 18) 5 (1, 15) 234/517 (45.3) 250/517 (48.4) 101/517 (19.5) 23/517 (4.4) 9/476 (1.9) 57/517 (11.0) 57/517 (11.0)
Total n ¼ 621, data were incomplete for some variables. Non-normally distributed continuous data are listed as median (Q1–Q3) and nominal variables are listed as n (%). ICU ¼ intensive care unit.
stay and postoperative length of hospital stay were positively associated with in-hospital mortality. The odds ratios (odds of death per postoperative day) for these 2 variables were 1.071 (1.064–1.078) and 1.033 (1.027–1.039), respectively (both p < 0.001). Finally, Figure 1 shows the daily incremental observed mortality rate for every postoperative hospital day up to day 50. On postoperative hospital day 0, the incremental mortality rate was 3.4%. By day 7 it increased to 4.2% and by day 30 it was 14.9%. The mortality rates for days 75, 100, 125, and 150 are not shown in Figure 1, but were 25.8%, 40.0%, 50.0%, and 100.0% respectively. After 30 postoperative ICU days the daily incremental observed mortality rate was 25.6%. A total of 234 patients in the cohort of 18,348 patients stayed longer than 30 postoperative days in the ICU.
Comment In a cohort of 18,348 cardiac surgery patients we found an in-hospital mortality rate of 3.4%. Most mortality occurred early after surgery with few deaths occurring after a protracted hospital course. There were high incidences of postoperative complications in patients who had a mortality event, but the occurrence of these complications explained only a small part of the variability in mortality timing. Further, we demonstrated that increasing postoperative length of ICU stay and postoperative length of hospital stay are associated with an increased likelihood of in-hospital mortality. Understanding the timing of in-hospital mortality after cardiac surgery is critical for physicians who take care of patients in the ICU and their families. Decisions about end of life care are complex and often practitioners and family members are forced to make decisions with only a crude estimation of mortality probability. In addition, medicine is practiced in a setting of scare resources and it
4
MAZZEFFI ET AL IN-HOSPITAL DEATH AND CARDIAC SURGERY
Ann Thorac Surg 2013;-:-–-
Table 3. Correlations and Linear Regression: Postoperative Complications and Mortality Timing Variable Acute renal failure Cardiac arrest Cardiac reoperation for bleeding Cardiac reoperation other than bleeding Deep sternal wound infection Pneumonia Stroke
Correlation
p Value
Beta (95% CI)
p Value
0.16 0.15 0.003 0.07 0.17 0.30 0.19
Background. It is important to characterize in-hospital mortality after cardiac surgery and understand the relationships between postoperative length of intensive care unit stay, postoperative length of hospital stay, and the likelihood of in-hospital mortality. Methods. We retrospectively identified all cardiac surgery cases that resulted in in-hospital mortality over an 8-year period at a single center. For these subjects we collected demographic data, preoperative comorbidities, and postoperative complications. We performed stepwise multivariate linear regression to determine which postoperative complications were associated with mortality timing. We also analyzed the relationships between postoperative length of intensive care unit stay, postoperative length of hospital stay, and in-hospital mortality in all patients (including survivors) who had cardiac surgery during the same time period. Finally, we calculated the daily incremental observed mortality rate for patients in the hospital up to postoperative day 50. Results. Six hundred twenty-one in-hospital mortalities occurred among 18,348 patients during the study
period (3.4%). Four postoperative complications were associated with mortality timing. Cardiac arrest had a negative association with the number of days until mortality, while deep sternal wound infection, stroke, and pneumonia had a positive association (all p < 0.05). Postoperative complications explained 15% of the variability in mortality timing (R2 model [ 0.15). The odds ratio for in-hospital mortality was 1.033 for each postoperative day in the hospital and 1.071 for each postoperative day in the intensive care unit (both p < 0.05). Conclusions. Most in-hospital mortality occurs during the first week after cardiac surgery with few mortalities occurring after a protracted hospital course. Postoperative complications have a limited ability to explain the variability in mortality timing. Increased length of postoperative intensive care unit stay and hospital stay after cardiac surgery are associated with an increased likelihood of in-hospital mortality.
A
surgeons, ICU physicians, and palliative care physicians as it may allow for better information to be given to family members when making complex decisions about end of life care. One common perception is that longer ICU or hospital stays are directly related to in-hospital mortality; however this supposition is not fully validated in cardiac surgery patients and it remains unclear on what ICU or hospital day number the risk for mortality reaches a significant level. The principal objective of our study was to better characterize in-hospital mortality after cardiac surgery by identifying characteristics of patients who had a mortality, describing its timing, and determining the relationships between postoperative length of ICU stay, postoperative length of hospital stay, and the likelihood of in-hospital mortality. We also sought to determine whether the occurrence of common postoperative complications could explain much of the variability in mortality timing.
ccording to the Society for Thoracic Surgeons (STS) database, the 30-day mortality rate for over 774,000 coronary artery bypass surgeries was 2.3% and for over 109,000 isolated valve surgeries was 3.4% [1, 2]. Similar 30-day mortality rates (from 1% to 4%) have been reported in other large cardiac surgery data sets [3, 4]. Despite these 30-day mortality data, in-hospital mortality after cardiac surgery has not been well characterized, including patient characteristics, timing, and the relationships between postoperative length of intensive care unit (ICU) stay, length of hospital stay, and the likelihood of in-hospital mortality. Better characterizing in-hospital mortality should be of great interest to cardiac
Accepted for publication Oct 11, 2013. Address correspondence to Dr Mazzeffi, Department of Anesthesiology, University of Maryland, 22 S Greene St, Baltimore, MD 21201; e-mail: mmazzeffi@anes.umm.edu.
Ó 2013 by The Society of Thoracic Surgeons Published by Elsevier Inc
(Ann Thorac Surg 2013;-:-–-) Ó 2013 by The Society of Thoracic Surgeons
0003-4975/$36.00 http://dx.doi.org/10.1016/j.athoracsur.2013.10.040
2
MAZZEFFI ET AL IN-HOSPITAL DEATH AND CARDIAC SURGERY
Patients and Methods Data Collection The Institutional Review Board at Emory University approved the study and granted a waiver of informed consent for data collection. All cardiac surgery cases performed at Emory University and its associated hospitals between February 1, 2004 and March 31, 2012 were identified by retrospective query of the Emory institutional STS database. For all patients we recorded the total number of postoperative ICU days, total number of postoperative hospital days, in-hospital mortality status, and discharge location. As a subset of the initial cohort we identified patients who had in-hospital mortality after their surgery. For these patients we collected multiple variables including demographics, common preoperative comorbidities, and preoperative lab values (age, body mass index, gender, chronic lung disease, diabetes, dyslipidemia, endocarditis, hypertension, New York Heart Association class, peripheral arterial disease, STS predicted mortality rate, previous cardiac surgery, previous myocardial infarction, preoperative creatinine, preoperative hematocrit, preoperative ejection fraction, cardiopulmonary bypass time, and procedure type). We also collected common postoperative complications (acute renal failure, cardiac arrest, cardiac reoperation for bleeding, cardiac reoperation other than bleeding, deep sternal wound infection, stroke, and pneumonia). Definitions for these variables were based on definitions used by the national STS database, which can be found in the STS training manual (www. sts.org/sts-national-database/database-managers/adultcardiac-surgery¼database/data-collection). Multiple versions of the STS database were used during initial data collection including 2.41, 2.52, 2.61, and 2.73. Definitions for variables were those that were used by the STS database at the time of the initial data collection. A significant number of patients were missing some data in the STS database (approximately 16%). In addition to the STS database variables, we also recorded the number of days until mortality occurred and its principal cause. The principal cause of mortality was classified as cardiac, respiratory, neurologic, infectious, or other, based upon the primary cause of mortality listed on the patients’ death certificate. For all analyses postoperative day 0 was defined as ending at 2400 on the same calendar day as the patients’ surgery. Subsequent postoperative days began at 0000 and ended at 2400.
Ann Thorac Surg 2013;-:-–-
Next, we created an explanatory model to describe the associations between postoperative complications and mortality timing. Both Pearson correlation analysis and stepwise multivariate linear regression were performed to determine which complications had a significant association with the number of days until death. Pearson correlation values were reported with respective p values. For the regression model, all previously mentioned postoperative complications (see Patients and Methods: Data collection) were included as independent variables and the dependent variable was the number of days until mortality occurred. A p value of 0.05 was required for entry into the model and was considered significant in the final model. Because there were a significant number of missing data points in the cohort, we used the missing data analysis; expectation maximization function in SPSS 21 to complete the data set prior to analysis. Missing data were considered to be missing completely at random because there was no logical association between the occurrence of postoperative complications (independent variables) and missing data points. Beta values with 95% confidence intervals were reported for regression coefficients included in the final model and the coefficient of determination was reported for the model. In order to determine the relationships between postoperative length of ICU stay, postoperative length of hospital stay, and in-hospital mortality we performed univariate logistic regression. In the first analysis, postoperative length of ICU stay was treated as a continuous independent variable and in-hospital mortality was treated as a dichotomous dependent variable. In the second analysis, postoperative length of hospital stay was treated as a continuous independent variable and inhospital mortality was treated as a dichotomous dependent variable. For these analyses, odds ratios with 95% confidence intervals were reported which reflect the odds of death per postoperative day spent in the ICU or hospital. Finally, we created a risk table containing the daily incremental observed mortality rate for each postoperative day in the hospital. This was done by first counting the number of patients remaining in the hospital on a given hospital day (denominator) and then looking at the number of those patients who subsequently died during their hospitalization (numerator). We calculated the daily incremental observed mortality rate for every day up to postoperative day 50 and then for individual days 75, 100, 125, and 150.
Statistical Methods
Results
All statistical analyses for the study were carried out using SPSS 21 (SPSS Inc, Chicago IL). First, we performed descriptive statistics on the in-hospital mortality cohort to determine the time distribution of mortality as well as the distributions for other study variables (demographics, preoperative lab values, preoperative comorbidities, and postoperative complications). Non-normally distributed continuous data were described using median [Q1–Q3] and nominal data were described using number (%).
The overall in-hospital mortality rate for the entire cohort was 3.4% (621 of 18,348). The most common principal causes of death were cardiac death 62.1%, respiratory failure 11.8%, infectious 7.7%, and neurologic injury 6.0%. The mode day (most common day) of death was postoperative day 0 with 15.0% of all deaths occurring on that day. The median day of death was 7. Of all deaths, 51.2% occurred in the first 7 days, while only 59 deaths or 9.5% occurred after postoperative day 30; 89.0% of patients
Ann Thorac Surg 2013;-:-–-
MAZZEFFI ET AL IN-HOSPITAL DEATH AND CARDIAC SURGERY
who were discharged went home, 7.8% went to transitional care facilities, 1.6% went to other hospitals, 1.4% went to nursing homes, and 0.2% went to hospice. Table 1 shows the characteristics of patients who had an inhospital death and Table 2 shows the common postoperative complications that occurred in them. Table 3 shows the results of the correlation studies and multivariate linear regression analysis for the in-hospital mortality cohort. Four variables (cardiac arrest, deep sternal wound infection, stroke, and pneumonia) were associated with mortality timing in the final model (all had p < 0.05). The coefficient of determination for the linear regression model was 0.15. Table 4 shows the results of the univariate logistic regression analyses. Both postoperative length of ICU
Table 1. Characteristics of Patients Who Had In-Hospital Mortality Variable Demographics: Age (years) Body mass index (kg/m2) Male gender, no (%) Comorbidities: Chronic lung disease – no (%) Diabetes mellitus – no (%) Dyslipidemia – no (%) Endocarditis – no (%) Hypertension – no (%) NYHA class – no (%) I II III IV Peripheral arterial disease – no (%) Predicted mortality (%) Previous cardiac surgery – no (%) Previous myocardial infarction – no (%) Lab values: Baseline creatinine (mg/dL) Hematocrit (g/dL) Left ventricular ejection fraction Surgical variables: CPB time (minutes) Type of procedure – no (%): CABG: AVR: AVR þ CABG: Isolated MVR: Other:
Median (Q1, Q3) or no (%)
66 (56, 76) 26 (23, 32) 368/621 (59.3) 159/605 214/609 150/263 44/601 524/617
(26.3) (35.1) (57.0) (7.3) (84.9)
55/487 (11.3) 93/487 (19.1) 126/487 (25.9) 213/487 (43.7) 132/601 (22.0) 7 (3, 16) 318/608 (52.3) 110/267 (41.2) 1.3 (1.0, 1.8) 35 (31, 40) 0.50 (0.35, 0.60) 159 (112, 219) 133/605 52/605 34/605 21/605 365/605
(22.0) (8.6) (5.6) (3.5) (60.3)
Total n ¼ 621, data were incomplete for some variables. Non-normally distributed continuous variables are represented by median (Q1–Q3) and nominal variables are listed as n (%). AVR ¼ aortic valve replacement; CABG ¼ coronary artery bypass graft; CPB ¼ cardiopulmonary bypass; NYHA ¼ New York Heart Association; MVR ¼ mitral valve repair or replacement.
3
Table 2. Complications in Patients Who Had an In-Hospital Death Variable Day of death (postoperative) Total ICU days Acute renal failure – no (%) Cardiac arrest – no (%) Cardiac reoperation for bleeding – no (%) Cardiac reoperation other than bleeding – no (%) Deep sternal wound infection – no (%) Stroke – no (%) Pneumonia – no (%)
Median (Q1, Q3) or no (%) 7 (1, 18) 5 (1, 15) 234/517 (45.3) 250/517 (48.4) 101/517 (19.5) 23/517 (4.4) 9/476 (1.9) 57/517 (11.0) 57/517 (11.0)
Total n ¼ 621, data were incomplete for some variables. Non-normally distributed continuous data are listed as median (Q1–Q3) and nominal variables are listed as n (%). ICU ¼ intensive care unit.
stay and postoperative length of hospital stay were positively associated with in-hospital mortality. The odds ratios (odds of death per postoperative day) for these 2 variables were 1.071 (1.064–1.078) and 1.033 (1.027–1.039), respectively (both p < 0.001). Finally, Figure 1 shows the daily incremental observed mortality rate for every postoperative hospital day up to day 50. On postoperative hospital day 0, the incremental mortality rate was 3.4%. By day 7 it increased to 4.2% and by day 30 it was 14.9%. The mortality rates for days 75, 100, 125, and 150 are not shown in Figure 1, but were 25.8%, 40.0%, 50.0%, and 100.0% respectively. After 30 postoperative ICU days the daily incremental observed mortality rate was 25.6%. A total of 234 patients in the cohort of 18,348 patients stayed longer than 30 postoperative days in the ICU.
Comment In a cohort of 18,348 cardiac surgery patients we found an in-hospital mortality rate of 3.4%. Most mortality occurred early after surgery with few deaths occurring after a protracted hospital course. There were high incidences of postoperative complications in patients who had a mortality event, but the occurrence of these complications explained only a small part of the variability in mortality timing. Further, we demonstrated that increasing postoperative length of ICU stay and postoperative length of hospital stay are associated with an increased likelihood of in-hospital mortality. Understanding the timing of in-hospital mortality after cardiac surgery is critical for physicians who take care of patients in the ICU and their families. Decisions about end of life care are complex and often practitioners and family members are forced to make decisions with only a crude estimation of mortality probability. In addition, medicine is practiced in a setting of scare resources and it
4
MAZZEFFI ET AL IN-HOSPITAL DEATH AND CARDIAC SURGERY
Ann Thorac Surg 2013;-:-–-
Table 3. Correlations and Linear Regression: Postoperative Complications and Mortality Timing Variable Acute renal failure Cardiac arrest Cardiac reoperation for bleeding Cardiac reoperation other than bleeding Deep sternal wound infection Pneumonia Stroke
Correlation
p Value
Beta (95% CI)
p Value
0.16 0.15 0.003 0.07 0.17 0.30 0.19
