Importance听
Emergency general surgery (EGS) patients have a disproportionate burden of death and complications. Chronic liver disease (CLD) increases the risk of complications following elective surgery. For EGS patients with CLD, long-term outcomes are unknown and risk stratification models do not reflect severity of CLD.
Objective听
To determine whether the Model for End-Stage Liver Disease (MELD) score is associated with increased risk of 90-day mortality following intensive care unit (ICU) admission in EGS patients.
Design, Setting, and Participants听
We performed a retrospective cohort study of patients with CLD who underwent an EGS procedure based on International Classification of Diseases, Ninth Revision (ICD-9) procedure codes and were admitted to a medical or surgical ICU within 48 hours of surgery between January 1, 1998, and September 20, 2012, at 2 academic medical centers. Chronic liver disease was identified using ICD-9 codes. Multivariable logistic regression was performed. The analysis was conducted from July 1, 2015, to January 1, 2016.
Main Outcomes and Measures听
The primary outcome was all-cause 90-day mortality.
Results听
A total of 13鈥�552 EGS patients received critical care; of these, 707 (5%) (mean [SD] age at hospital admission, 56.6 [14.2] years; 64% male; 79% white) had CLD and data to determine MELD score at ICU admission. The median MELD score was 14 (interquartile range, 10-20). Overall 90-day mortality was 30.1%. The adjusted odds ratio of 90-day mortality for each 10-point increase in MELD score was 1.63 (95% CI, 1.34-1.98). A decrease in MELD score of more than 3 in the 48 hours following ICU admission was associated with a 2.2-fold decrease in 90-day mortality (odds ratio鈥�=鈥�0.46; 95% CI, 0.22-0.98).
Conclusions and Relevance听
In this study, MELD score was associated with 90-day mortality following EGS in patients with CLD. The MELD score can be used as a prognostic factor in this patient population and should be used in preoperative risk prediction models and when counseling EGS patients on the risks and benefits of operative intervention.
Cirrhosis and chronic liver disease (CLD) are significant causes of morbidity and mortality in the United States, with CLD accounting for 36鈥�427 deaths among hospitalized patients in 2013.1 Patients with CLD undergoing surgery have comparatively higher rates of surgical complications and death.2 In addition, among patients with CLD, admission to the intensive care unit (ICU), mechanical ventilation, and renal replacement therapy have been independently shown to increase hospital mortality.3,4 Scoring tools such as the Model for End-Stage Liver Disease (MELD) score are used to predict outcomes in patients with CLD.2-6 However, most of these studies are restricted to patients who underwent liver transplantation, and most studies among patients not receiving a transplant have not focused on care delivered in emergent settings. Because patients with CLD also experience acute surgical events,7 it is important to examine and accurately predict the outcomes of acute surgical care among these patients.
Emergency general surgery (EGS) is associated with increased rates of morbidity and mortality compared with nonemergent general surgery cases.8 Patients undergoing EGS are approximately 2.5 times more likely to experience a significant complication and have a 6-fold increase in mortality relative to non-EGS patients.9 The underlying causes of this increased morbidity and mortality are not fully understood, but medical comorbidities and physiological derangements are likely to be contributing factors.10-13 Although surgical risk calculation tools such as the American College of Surgeons National Surgical Quality Improvement Project Surgical Risk Calculator are used to gain an objective sense of surgical risk stratification, such tools have yet to be comprehensively studied in this patient population and do not include the use of liver disease鈥搒pecific assessment tools such as the MELD score in the prediction of outcomes among patients with CLD undergoing EGS.14 We hypothesized that among patients with CLD who underwent EGS and were treated in the ICU, the MELD score would independently be associated with mortality. Therefore, the aim of this study was to determine whether the MELD score is associated with increased risk of 90-day mortality following ICU admission among EGS patients with CLD.
Box Section Ref IDKey Points
Question What is the relationship between Model for End-Stage Liver Disease (MELD) score and mortality in emergency general surgery patients?
Findings In this cohort study that included 707 emergency general surgery patients with chronic liver disease admitted to a surgical intensive care unit, MELD score was independently associated with 90-day mortality. A decrease in MELD score after 48 hours was associated with improved survival.
Meaning The MELD score is an important prognostic tool for emergency general surgery patients with chronic liver disease.
This is a retrospective cohort study of patients admitted between January 1, 1998, and September 20, 2012, to 2 large academic medical centers that provide primary and tertiary care to an ethnically and socioeconomically diverse population within eastern Massachusetts and the surrounding region. Patients aged 18 years or older who required ICU admission and had emergency surgery (previously defined by the American Association for the Surgery of Trauma15) within 48 hours of ICU admission were included. Patients without CLD16 or without information available for MELD score calculation at ICU admission were excluded. Data were obtained through the Research Patient Data Registry (RPDR), a computerized registry that serves as a central data warehouse for all inpatient and outpatient records at Partners HealthCare sites.17,18 The RPDR has been previously used in clinical research studies in critically ill patients.7,19,20 Approval for the study was granted by the Partners Human Research Committee. Requirement for informed consent was waived as the data were analyzed anonymously.
The exposure of interest was the MELD score at ICU admission.21 The MELD score was calculated at ICU admission using United Network for Organ Sharing modifications as follows: MELD Score鈥�=鈥塠0.957鈥壝椻�塴n(Serum Creatinine)鈥�+鈥�0.378鈥壝椻�塴n(Serum Bilirubin)鈥�+鈥�1.120鈥壝椻�塴n(INR)鈥�+鈥�0.643)鈥壝椻��10], where serum creatinine and serum bilibrubin are in milligrams per deciliter and INR indicates international normalized ratio; if the patient is undergoing hemodialysis, the value for serum creatinine is automatically set to 4.0 mg/dL. Because the MELD score uses a log scale calculation, any creatinine, bilirubin, or INR value less than 1 is given the lower limit value of 1 to prevent a negative score.22 The MELD score was categorized as 6.0 to 9.9; 10.0 to 19.9; 20.0 to 29.9; and 30.0 or higher. Chronic liver disease was identified by the presence of International Classification of Diseases, Ninth Revision (ICD-9) codes for CLD (571.x), chronic hepatitis (70.54), or chronic hepatitis B (70.32) prior to or during hospital admission. This approach had been previously validated using the RPDR data set.16 The Deyo-Charlson index was used to assess the burden of chronic illness with higher scores indicating more comorbidity via ICD-9 coding algorithms, a well-studied and validated approach.23 Race was designated by the patient or by a patient representative.
Sepsis was defined by ICD-9 codes 038, 995.91, 995.92, or 785.52, 3 days prior to critical care initiation to 7 days after critical care initiation.24 Using electronic pharmacy records, exposure to inotropes and vasopressors was determined for dopamine hydrochloride, dobutamine hydrochloride, epinephrine, norepinephrine bitartrate, phenylephrine hydrochloride, milrinone, and vasopressin. Inotropes or vasopressors were considered to be present if prescribed 3 days prior to critical care initiation to 7 days after critical care initiation.25 Acute organ failure was adapted from the study by Martin et al26 and defined by a combination of ICD-9 and Current Procedural Terminology (CPT) codes relating to acute organ dysfunction (respiratory failure, cardiovascular failure, renal, hepatic, hematologic, metabolic, and/or neurologic) assigned from 3 days prior to critical care initiation to 30 days after critical care initiation.
Acute kidney injury was defined as Risk, Injury, Failure, Loss, and End-Stage Kidney Disease (RIFLE) class injury or failure occurring between 3 days prior to critical care initiation and 7 days after critical care initiation.27 Noncardiogenic acute respiratory failure was identified by the presence of ICD-9 codes for respiratory failure or pulmonary edema (518.4, 518.5, 518.81, and 518.82) and mechanical ventilation (96.7x), excluding congestive heart failure (428.0-428.9), following hospital admission.28 For severity of illness risk adjustment, we used the acute organ failure score, an ICU risk prediction score derived and validated from demographic characteristics (age, race) as well as ICD-9, Clinical Modification code鈥揵ased comorbidity, sepsis, and acute organ failure covariates that has similar discrimination for 30-day mortality as the Acute Physiologic and Chronic Health Evaluation II score.29 All CPT or ICD-9 codes were derived from daily billing charges from individual physicians.
The primary end point was 90-day all-cause mortality following critical care initiation. We used the Social Security Administration Death Master File to determine vital status, which has high sensitivity and specificity for mortality.30 We have validated the accuracy of the Social Security Administration Death Master File for in-hospital and out-of-hospital mortality in the RPDR database.19 Among the cohort, 100% had at least 90-day follow-up after ICU admission. The censoring date was December 31, 2012. The secondary end point was 30-day hospital readmission, which was determined from RPDR hospital admission data as previously described31 and was defined as a subsequent or unscheduled admission to Brigham and Women鈥檚 Hospital or Massachusetts General Hospital within 30 days of discharge following the hospitalization associated with the critical care exposure.31-33 We excluded readmissions with diagnosis related group codes that are commonly associated with planned readmissions in addition to diagnosis related group codes for transplantation, procedures related to pregnancy, and psychiatric issues.31,34
Based on prior studies,7,35 we assumed that 90-day mortality would increase an absolute 17.7% in patients with a MELD score of 20 to 29 (25%) compared with those with a MELD score lower than 9.9 (12.5%). With an 伪 error level of 5% and a power of 80%, the minimum sample size thus required for our primary end point was 336 patients.
Categorical covariates were described by frequency distribution and compared across MELD score groups using contingency tables and 蠂2 testing. Continuous covariates were examined graphically and in terms of summary statistics and compared across MELD groups using 1-way analysis of variance. Unadjusted associations between MELD groups and outcomes were estimated by bivariable logistic regression analysis. Adjusted odds ratios (ORs) were estimated by multivariable logistic regression models with inclusion of covariate terms thought to plausibly interact with both MELD and mortality.
Overall model fit was assessed using the Hosmer-Lemeshow test. Analyses based on fully adjusted models were performed to evaluate the MELD-mortality association, and P for interaction was determined to explore for any evidence of effect modification. To evaluate for multicollinearity, we calculated the variance inflation factors and tolerances for each of the independent variables. Locally weighted scatterplot smoothing was used to graphically represent the relationship between absolute MELD count and rate of 90-day mortality. All P values presented are 2-tailed; P鈥�<鈥�.05 was considered nominally significant. All analyses were performed using Stata version 13.1 MP statistical software (StataCorp LP).
Between 1998 and 2012, there were 13鈥�552 unique EGS patients treated in the ICU. A total of 12鈥�623 patients without CLD and 222 patients without information available for MELD score calculation at ICU admission were excluded, leaving 707 patients in the final cohort.
In the study cohort, the mean (SD) age at hospital admission was 56.6 (14.2) years. Most patients were male (64%) and white (79%). The in-hospital, 90-day, and 365-day mortality rates were 25.3%, 30.1%, and 41.1%, respectively. The rate of unplanned 30-day postdischarge hospital readmissions was 16.1%. The median MELD score was 14 (interquartile range, 10-20). Patient characteristics of the study cohort were stratified according to 90-day mortality (Table 1). Age, Deyo-Charlson index score, acute organ failure, sepsis, intubation, use of vasopressors and inotropes, metastatic malignant neoplasm, renal replacement therapy, and MELD score were significantly associated with 90-day mortality. Factors that were associated with stratified MELD category included age, race, Deyo-Charlson index score, acute organ failure, sepsis, use of vasopressors and inotropes, renal replacement therapy, glomerular filtration rate, days from hospital admission to ICU admission, and mortality (Table 2).
Mortality risk in the 90 days after ICU admission was higher in patients with higher MELD scores (Figure). Compared with patients with a MELD score of 6.0 to 9.9, the odds of 90-day mortality were 2.0-fold higher in those with a MELD score of 10.0 to 19.9 (OR鈥�=鈥�1.99; 95% CI, 1.25-3.16), 3.5-fold higher in those with a MELD score of 20.0 to 29.9 (OR鈥�=鈥�3.49; 95% CI, 2.04-5.97), and 5.2-fold higher in those with a MELD score of 30.0 or higher (OR鈥�=鈥�5.25; 95% CI, 2.79-9.88) (Table 3). The MELD score level remained a significant predictor of the odds of 90-day mortality after adjustment for age, sex, race, Deyo-Charlson index score, sepsis, and acute organ failure. Again compared with patients with a MELD score of 6.0 to 9.9, the adjusted odds of 90-day mortality were 1.4-fold higher (but not statistically significant) in those with a MELD score of 10.0 to 19.9 (OR鈥�=鈥�1.45; 95% CI, 0.89-2.38), 2.1-fold higher in those with a MELD score of 20.0 to 29.9 (OR鈥�=鈥�2.12; 95% CI, 1.17-3.85), and 3.6-fold higher in those with a MELD score of 30.0 or higher (OR鈥�=鈥�3.58; 95% CI, 1.76-7.28) (Table 3).
The adjusted 90-day mortality model showed good calibration (Hosmer-Lemeshow 蠂28鈥�=鈥�12.22; P鈥�=鈥�.14), good discrimination (C statistic鈥�=鈥�0.73; 95% CI, 0.69-0.77), and an absence of multicollinearity as determined by variance inflation factor. When MELD score was analyzed as continuous, the adjusted OR of 90-day mortality for each 10-point increase in MELD score was 1.63 (95% CI, 1.34-1.98). Further, the hazard ratios of mortality adjusted for age, sex, race, Deyo-Charlson index score, sepsis, and acute organ failure were 1.25 (95% CI, 0.93-1.68) for patients with a MELD score of 10.0 to 19.9, 1.63 (95% CI, 1.14-2.34) for those with a MELD score of 20.0 to 29.9, and 1.81 (95% CI, 1.16-2.81) for those with a MELD score of 30.0 or higher compared with patients with a MELD score of 6.0 to 9.9.
There was no significant effect modification of the association between MELD score and 90-day mortality on the basis of acute kidney injury (P for interaction鈥�=鈥�.55), hospital (P for interaction鈥�=鈥�.26), or year of ICU admission (P for interaction鈥�=鈥�.08). Effect modification is present regarding the presence of sepsis (P for interaction鈥�=鈥�.007). Individually running the final model with and without a sepsis term to the final model does not alter the effect size or significance of the change in the association between MELD score and 90-day mortality (data not shown).
Elevated MELD score was a predictor of 30-day hospital readmission in ICU survivors with CLD who underwent EGS (n鈥�=鈥�528). The odds of 30-day hospital readmission in patients with a MELD score higher than 19.9 was 1.7-fold that of patients with a MELD score of 19.9 or lower (OR鈥�=鈥�1.73; 95% CI, 1.07-2.77; P鈥�=鈥�.02). The presence of a MELD score higher than 19.9 remained a significant predictor of the odds of 30-day hospital readmission after adjustment for age, sex, race, Deyo-Charlson index score, acute organ failure, and sepsis. The adjusted odds of 30-day hospital readmission in patients with a MELD score higher than 19.9 was 1.7-fold that of patients with a MELD score of 19.9 or lower (OR鈥�=鈥�1.71; 95% CI, 1.01-2.92; P鈥�=鈥�.047).
In a subset of patients with MELD score determined at ICU admission and 48 hours after (n鈥�=鈥�318), we determined the association between a change in MELD score and 90-day mortality. A decrease in MELD score of more than 3 in the 48 hours following ICU admission was associated with a 2.2-fold decrease in 90-day mortality (OR鈥�=鈥�0.46; 95% CI, 0.22-0.98; P鈥�=鈥�.045) relative to patients with a change in MELD score of 卤3 adjusted for age, sex, race, sepsis, Deyo-Charlson index score, and acute organ failure. An increase in MELD score of more than 3 in the 48 hours following ICU admission was associated with a nonsignificant increase in 90-day mortality (OR鈥�=鈥�1.40; 95% CI, 0.77-2.54; P鈥�=鈥�.27), fully adjusted, compared with patients with a change of 卤3 in MELD score.
In this study of patients with CLD who underwent EGS, we demonstrated that increasing severity of CLD was associated with increased odds of 90-day mortality and that a decrease in MELD score after 48 hours was associated with significant reductions in the odds of 90-day mortality. This study also showed that 1 in 6 EGS patients with CLD will be readmitted. This is the first study, to our knowledge, to demonstrate this relationship in the EGS patient.
The MELD score, first used to predict survival in patients undergoing transjugular intrahepatic portosystemic shunts,21 has been used extensively in the United States and Europe as a tool for the prioritization of liver transplants.22 More recently, researchers have identified an association between MELD score and mortality in trauma patients7 and major complications in cirrhotic patients undergoing emergency hernia repair.36 This growing body of evidence suggests that MELD scores may be important in predicting outcomes among a wide variety of patients not receiving a transplant.
Emergency general surgery patients have a very high burden of complications and death.8-11 Appropriate methods are therefore necessary to accurately risk stratify these patients and help predict their outcomes and to identify ICU survivors at high risk for adverse outcomes following hospital discharge. Understanding the true risk of death following surgery is critical to the shared decision-making process for both surgeons and patients.37 Current surgical risk calculators either do not include liver disease or do not include severity of liver disease as measured by laboratory data in mortality calculations.38 One common surgical risk assessment tool, the American College of Surgeons National Surgical Quality Improvement Project Surgical Risk Calculator, includes the presence of ascites within 30 days preoperatively as a universal risk factor that may relate to liver disease or to other factors.14 However, this tool has been shown to underestimate mortality in EGS patients.39 Inclusion of the MELD score in surgical risk calculators may improve accuracy and aid in patient counseling along with operative decision making.
In this study, we identified an association between improvement in MELD score at 48 hours and improved survival. This association has also been described in trauma patients.7 As there are 3 laboratory components to the MELD score (serum bilirubin level, serum creatinine level, and INR), the association between a change in MELD score and 90-day mortality may reflect changes in liver or renal function in the first 48 hours of critical care. Among the 3 variables of the MELD score, INR has the highest multiplicative value, followed by creatinine level. Variations in INR or creatinine level may translate to up to 20% differences in MELD score.40 As this was an observational study, we cannot conclude that survival can be improved by focused attention to improving laboratory values. However, we do believe that a patient鈥檚 clinical change as reflected in changes in laboratory values such as INR or serum creatinine level is reflected in both MELD score and outcomes such as survival. While these variables likely relate to clinical changes that may be seen in all patients, not just those with CLD, it is unknown whether MELD score would be associated with mortality in patients without CLD.
This study is in agreement with prior studies showing high risk for 30-day readmission following EGS.15 Emergency general surgery patients identified as having elevated MELD scores at ICU admission may benefit from increased follow-up after hospital discharge to prevent unplanned readmission. Efforts to improve continuity between inpatient and outpatient care in the form of improved communication between physicians and trained home nursing visits have been shown to reduce readmission rates in surgical patient populations.41 Further investigation into the role of such services in EGS patients, along with other surgical cohorts, is warranted.
One of the strengths of this study is that unlike prior evaluations of MELD scores that include few patients at a single center, we used a large, prospectively designed intensive care database that comprised patients treated at different hospitals and contained both administrative and clinical data. This offered us a large sample size and better ability to control for clinical conditions than purely administrative databases. Also, in addition to the previously validated approaches to define EGS,15,42 CLD,16 comorbidites,23 acute organ failure,29 and sepsis,24 we determined all-cause mortality using a validated method based on the Social Security Administration Death Master File.19 Very few studies are able to determine patients鈥� survival status once they are discharged from the hospital. Last, we examined 30-day readmission rates, which have become a quality metric used in evaluating the care of Medicare patients in the United States.43,44
However, the limitations of this study must also be considered. Ascertainment bias may be present as more than one-fifth of EGS patients with CLD had missing data on the MELD components, potentially limiting generalizability of our study findings. Residual confounding may also be present despite multivariable adjustment. We were unable to adjust for physiologically based severity of illness scores, which are strong predictors of critical illness outcome.45 It is conceivable that inclusion of a physiological score in the analysis may alter the associations between MELD score and outcomes.
These data demonstrate that in critically ill patients with CLD, increased MELD score is associated with increased mortality and hospital readmission following EGS. Concurrently, MELD score reduction exerts a protective effect in this patient population. The identification of exposures that are predictive of outcomes in EGS patients may be useful for preoperative planning and may be useful inclusions for future risk stratification models. Furthermore, EGS patients with CLD might benefit from enhanced longitudinal care following hospital discharge to reduce unplanned readmissions. Further investigation into the relationship between MELD score and outcomes in patients without CLD is warranted.
Corresponding Author: Joaquim M. Havens, MD, Division of Trauma, Burns, and Surgical Critical Care, Brigham and Women鈥檚 Hospital, 75 Francis St, Boston, MA 02115 (jhavens@partners.org).
Accepted for Publication: March 11, 2016.
Published Online: May 18, 2016. doi:10.1001/jamasurg.2016.0789.
Author Contributions: Dr Christopher had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Havens, Askari, Salim, Christopher.
Acquisition, analysis, or interpretation of data: Columbus, Olufajo, Christopher.
Drafting of the manuscript: Havens, Columbus, Olufajo, Christopher.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Christopher.
Administrative, technical, or material support: Havens, Askari.
Study supervision: Havens, Salim.
Conflict of Interest Disclosures: None reported.
Previous Presentation: This paper was presented at the 87th Annual Meeting of the Pacific Coast Surgical Association; February 16, 2016; Kohala Coast, Hawaii.
Additional Contributions: This article is dedicated to the memory of our dear friend and colleague Nathan Edward Hellman, MD, PhD.
1.Centers for Disease Control and Prevention. Chronic liver disease and cirrhosis. . Accessed December 21, 2015.
2.Suman
听A, Carey
听WD. 听Assessing the risk of surgery in patients with liver disease.听听Cleve Clin J Med. 2006;73(4):398-404.
3.Levesque
听E, Hoti
听E, Azoulay
听D,
听et al. 听Prospective evaluation of the prognostic scores for cirrhotic patients admitted to an intensive care unit.听听J Hepatol. 2012;56(1):95-102.
4.Bosetti
听C, Levi
听F, Lucchini
听F, Zatonski
听WA, Negri
听E, La Vecchia
听C. 听Worldwide mortality from cirrhosis: an update to 2002.听听J Hepatol. 2007;46(5):827-839.
5.Bittermann
听T, Makar
听G, Goldberg
听DS. 听Early post-transplant survival: interaction of MELD score and hospitalization status.听听J Hepatol. 2015;63(3):601-608.
6.Klein
听KB, Stafinski
听TD, Menon
听D. 听Predicting survival after liver transplantation based on pre-transplant MELD score: a systematic review of the literature.听听PLoS One. 2013;8(12):e80661.
7.Peetz
听A, Salim
听A, Askari
听R,
听et al. 听Association of Model for End-Stage Liver Disease score and mortality in trauma patients with chronic liver disease.听听闯础惭础 Surg. 2016;151(1):41-48.
8.Ghaferi
听AA, Birkmeyer
听JD, Dimick
听JB. 听Variation in hospital mortality associated with inpatient surgery.听听N Engl J Med. 2009;361(14):1368-1375.
9.Havens
听JM, Peetz
听AB, Do
听WS,
听et al. 听The excess morbidity and mortality of emergency general surgery.听听J Trauma Acute Care Surg. 2015;78(2):306-311.
10.Akinbami
听F, Askari
听R, Steinberg
听J, Panizales
听M, Rogers
听SO
听Jr. 听Factors affecting morbidity in emergency general surgery.听听Am J Surg. 2011;201(4):456-462.
11.Li
听LT, Jafrani
听RJ, Becker
听NS,
听et al. 听Outcomes of acute versus elective primary ventral hernia repair.听听J Trauma Acute Care Surg. 2014;76(2):523-528.
12.Matsuyama
听T, Iranami
听H, Fujii
听K, Inoue
听M, Nakagawa
听R, Kawashima
听K. 听Risk factors for postoperative mortality and morbidities in emergency surgeries.听听J Anesth. 2013;27(6):838-843.
13.To
听KB, Cherry-Bukowiec
听JR, Englesbe
听MJ,
听et al. 听Emergent versus elective cholecystectomy: conversion rates and outcomes.听听Surg Infect (Larchmt). 2013;14(6):512-519.
14.Bilimoria
听KY, Liu
听Y, Paruch
听JL,
听et al. 听Development and evaluation of the universal ACS NSQIP surgical risk calculator: a decision aid and informed consent tool for patients and surgeons.听听J Am Coll Surg. 2013;217(5):833-842, e1, e3.
15.Havens
听JM, Olufajo
听OA, Cooper
听ZR, Haider
听AH, Shah
听AA, Salim
听A. 听Defining rates and risk factors for readmissions following emergency general surgery听 [published online November 11, 2015]. 听闯础惭础 Surg. doi:.
16.Hug
听BL, Lipsitz
听SR, Seger
听DL, Karson
听AS, Wright
听SC, Bates
听DW. 听Mortality and drug exposure in a 5-year cohort of patients with chronic liver disease.听听Swiss Med Wkly. 2009;139(51-52):737-746.
17.Murphy
听SN, Chueh
听HC. 听A security architecture for query tools used to access large biomedical databases.听听Proc AMIA Symp. 2002:552-556.
18.Nalichowski
听R, Keogh
听D, Chueh
听HC, Murphy
听SN. 听Calculating the benefits of a Research Patient Data Repository.听听AMIA Annu Symp Proc. 2006:1044.
19.Zager
听S, Mendu
听ML, Chang
听D,
听et al. 听Neighborhood poverty rate and mortality in patients receiving critical care in the academic medical center setting.听听颁丑别蝉迟. 2011;139(6):1368-1379.
20.Mogensen
听KM, Robinson
听MK, Casey
听JD,
听et al. 听Nutritional status and mortality in the critically ill.听听Crit Care Med. 2015;43(12):2605-2615.
21.Malinchoc
听M, Kamath
听PS, Gordon
听FD, Peine
听CJ, Rank
听J, ter Borg
听PC. 听A model to predict poor survival in patients undergoing transjugular intrahepatic portosystemic shunts.听听贬别辫补迟辞濒辞驳测. 2000;31(4):864-871.
22.Wiesner
听R, Edwards
听E, Freeman
听R,
听et al; United Network for Organ Sharing Liver Disease Severity Score Committee. 听Model for End-Stage Liver Disease (MELD) and allocation of donor livers.听听骋补蝉迟谤辞别苍迟别谤辞濒辞驳测. 2003;124(1):91-96.
23.Deyo
听RA, Cherkin
听DC, Ciol
听MA. 听Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases.听听J Clin Epidemiol. 1992;45(6):613-619.
24.Liu
听V, Escobar
听GJ, Greene
听JD,
听et al. 听Hospital deaths in patients with sepsis from 2 independent cohorts.听听闯础惭础. 2014;312(1):90-92.
25.Purtle
听SW, Moromizato
听T, McKane
听CK, Gibbons
听FK, Christopher
听KB. 听The association of red cell distribution width at hospital discharge and out-of-hospital mortality following critical illness*.听听Crit Care Med. 2014;42(4):918-929.
26.Martin
听GS, Mannino
听DM, Eaton
听S, Moss
听M. 听The epidemiology of sepsis in the United States from 1979 through 2000.听听N Engl J Med. 2003;348(16):1546-1554.
27.Braun
听AB, Litonjua
听AA, Moromizato
听T, Gibbons
听FK, Giovannucci
听E, Christopher
听KB. 听Association of low serum 25-hydroxyvitamin D levels and acute kidney injury in the critically ill.听听Crit Care Med. 2012;40(12):3170-3179.
28.Thickett
听DR, Moromizato
听T, Litonjua
听AA,
听et al. 听Association between prehospital vitamin D status and incident acute respiratory failure in critically ill patients: a retrospective cohort study.听听BMJ Open Respir Res. 2015;2(1):e000074.
29.Elias
听KM, Moromizato
听T, Gibbons
听FK, Christopher
听KB. 听Derivation and validation of the acute organ failure score to predict outcome in critically ill patients: a cohort study.听听Crit Care Med. 2015;43(4):856-864.
30.Sohn
听MW, Arnold
听N, Maynard
听C, Hynes
听DM. 听Accuracy and completeness of mortality data in the Department of Veterans Affairs.听听Popul Health Metr. 2006;4:2.
31.Horkan
听CM, Purtle
听SW, Mendu
听ML, Moromizato
听T, Gibbons
听FK, Christopher
听KB. 听The association of acute kidney injury in the critically ill and postdischarge outcomes: a cohort study*.听听Crit Care Med. 2015;43(2):354-364.
32.Koehler
听BE, Richter
听KM, Youngblood
听L,
听et al. 听Reduction of 30-day postdischarge hospital readmission or emergency department (ED) visit rates in high-risk elderly medical patients through delivery of a targeted care bundle.听听J Hosp Med. 2009;4(4):211-218.
33.Landrum
听L, Weinrich
听S. 听Readmission data for outcomes measurement: identifying and strengthening the empirical base.听听Qual Manag Health Care. 2006;15(2):83-95.
34.Jencks
听SF, Williams
听MV, Coleman
听EA. 听Rehospitalizations among patients in the Medicare fee-for-service program.听听N Engl J Med. 2009;360(14):1418-1428.
35.Kamath
听PS, Kim
听WR; Advanced Liver Disease Study Group. 听The Model for End-Stage Liver Disease (MELD).听听贬别辫补迟辞濒辞驳测. 2007;45(3):797-805.
36.Odom
听SR, Gupta
听A, Talmor
听D, Novack
听V, Sagy
听I, Evenson
听AR. 听Emergency hernia repair in cirrhotic patients with ascites.听听J Trauma Acute Care Surg. 2013;75(3):404-409.
37.Schenker
听Y, Fernandez
听A, Sudore
听R, Schillinger
听D. 听Interventions to improve patient comprehension in informed consent for medical and surgical procedures: a systematic review.听听Med Decis Making. 2011;31(1):151-173.
38.Sherman
听SK, Hrabe
听JE, Charlton
听ME, Cromwell
听JW, Byrn
听JC. 听Development of an improved risk calculator for complications in proctectomy.听听J Gastrointest Surg. 2014;18(5):986-994.
39.Hyder
听JA, Reznor
听G, Wakeam
听E, Nguyen
听LL, Lipsitz
听SR, Havens
听JM. 听Risk prediction accuracy differs for emergency versus elective cases in the ACS-NSQIP听 [published online December 31, 2015]. 听Ann Surg. doi:.
40.Trotter
听JF, Olson
听J, Lefkowitz
听J, Smith
听AD, Arjal
听R, Kenison
听J. 听Changes in international normalized ratio (INR) and Model for Endstage Liver Disease (MELD) based on selection of clinical laboratory.听听Am J Transplant. 2007;7(6):1624-1628.
41.Hall
听MH, Esposito
听RA, Pekmezaris
听R,
听et al. 听Cardiac surgery nurse practitioner home visits prevent coronary artery bypass graft readmissions.听听Ann Thorac Surg. 2014;97(5):1488-1493.
42.Shafi
听S, Aboutanos
听MB, Agarwal
听S
听Jr,
听et al; AAST Committee on Severity Assessment and Patient Outcomes. 听Emergency general surgery: definition and estimated burden of disease.听听J Trauma Acute Care Surg. 2013;74(4):1092-1097.
43.Joynt
听KE, Jha
听AK. 听Thirty-day readmissions: truth and consequences.听听N Engl J Med. 2012;366(15):1366-1369.
44.Kocher
听RP, Adashi
听EY. 听Hospital readmissions and the Affordable Care Act: paying for coordinated quality care.听听闯础惭础. 2011;306(16):1794-1795.
45.Knaus
听WA, Draper
听EA, Wagner
听DP, Zimmerman
听JE. 听APACHE II: a severity of disease classification system.听听Crit Care Med. 1985;13(10):818-829.