[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Download PDF
Figure 1.
Flow Diagram
Flow Diagram

CPT indicates Current Procedural Terminology; ICD-9, International Classification of Diseases, Ninth Revision; ICU, intensive care unit; INR, international normalized ratio.

Figure 2.
Survival Analysis by Change in Model for End-Stage Liver Disease (MELD)
Survival Analysis by Change in Model for End-Stage Liver Disease (MELD)

Unadjusted event rates were calculated using Kaplan-Meier methods and compared with the use of the log-rank test. Categorization of the change in MELD scores is per the primary analyses. The global comparison log-rank, P = .003.

Table 1.  
Clinical and Demographic Characteristics of the Cohorta
Clinical and Demographic Characteristics of the Cohorta
Table 2.  
Patient Characteristics of the Cohort by Change in MELD Scorea
Patient Characteristics of the Cohort by Change in MELD Scorea
Table 3.  
Patient Characteristics of the Cohort by Nature and Body Region of Injury
Patient Characteristics of the Cohort by Nature and Body Region of Injury
Table 4.  
Unadjusted and Adjusted Associations Between Change in MELD Score and Mortality in Trauma Patients With Chronic Liver Disease
Unadjusted and Adjusted Associations Between Change in MELD Score and Mortality in Trauma Patients With Chronic Liver Disease
1.
Schuppan  D, Afdhal  NH.  Liver cirrhosis. Lancet. 2008;371(9615):838-851.PubMedArticle
2.
Levesque  E, Hoti  E, Azoulay  D,  et al.  Prospective evaluation of the prognostic scores for cirrhotic patients admitted to an intensive care unit. J Hepatol. 2012;56(1):95-102.PubMedArticle
3.
Tinkoff  G, Rhodes  M, Diamond  D, Lucke  J.  Cirrhosis in the trauma victim; effect on mortality rates. Ann Surg. 1990;211(2):172-177.PubMedArticle
4.
Christmas  AB, Wilson  AK, Franklin  GA, Miller  FB, Richardson  JD, Rodriguez  JL.  Cirrhosis and trauma: a deadly duo. Am Surg. 2005;71(12):996-1000.PubMed
5.
Hug  BL, Lipsitz  SR, Seger  DL, Karson  AS, Wright  SC, Bates  DW.  Mortality and drug exposure in a 5-year cohort of patients with chronic liver disease. Swiss Med Wkly. 2009;139(51-52):737-746.PubMed
6.
Zager  S, Mendu  ML, Chang  D,  et al.  Neighborhood poverty rate and mortality in patients receiving critical care in the academic medical center setting. Chest. 2011;139(6):1368-1379.PubMedArticle
7.
Moromizato  T, Litonjua  AA, Braun  AB, Gibbons  FK, Giovannucci  E, Christopher  KB.  Association of low serum 25-hydroxyvitamin D levels and sepsis in the critically ill. Crit Care Med. 2014;42(1):97-107.PubMedArticle
8.
Purtle  SW, Moromizato  T, McKane  CK, Gibbons  FK, Christopher  KB.  The association of red cell distribution width at hospital discharge and out-of-hospital mortality following critical illness. Crit Care Med. 2014;42(4):918-929.PubMedArticle
9.
McKane  CK, Marmarelis  M, Mendu  ML, Moromizato  T, Gibbons  FK, Christopher  KB.  Diabetes mellitus and community-acquired bloodstream infections in the critically ill. J Crit Care. 2014;29(1):70-76.PubMedArticle
10.
Braun  AB, Litonjua  AA, Moromizato  T, Gibbons  FK, Giovannucci  E, Christopher  KB.  Association of low serum 25-hydroxyvitamin D levels and acute kidney injury in the critically ill. Crit Care Med. 2012;40(12):3170-3179.PubMedArticle
11.
Davie  G, Cryer  C, Langley  J.  Improving the predictive ability of the ICD-based Injury Severity Score. Inj Prev. 2008;14(4):250-255.PubMedArticle
12.
Malinchoc  M, Kamath  PS, Gordon  FD, Peine  CJ, Rank  J, ter Borg  PC.  A model to predict poor survival in patients undergoing transjugular intrahepatic portosystemic shunts. Hepatology. 2000;31(4):864-871.PubMedArticle
13.
Kim  JD, Choi  JY, Kwon  JH,  et al.  Performance of posttransplant Model for End-Stage Liver Disease (MELD) and delta-MELD scores on short-term outcome after living donor liver transplantation. Transplant Proc. 2009;41(9):3766-3768.PubMedArticle
14.
Martin  GS, Mannino  DM, Eaton  S, Moss  M.  The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med. 2003;348(16):1546-1554.PubMedArticle
15.
Cooke  CR, Erickson  SE, Eisner  MD, Martin  GS.  Trends in the incidence of noncardiogenic acute respiratory failure: the role of race. Crit Care Med. 2012;40(5):1532-1538.PubMedArticle
16.
Shafi  S, Aboutanos  MB, Agarwal  S  Jr,  et al; AAST Committee on Severity Assessment and Patient Outcomes.  Emergency general surgery: definition and estimated burden of disease. J Trauma Acute Care Surg. 2013;74(4):1092-1097.PubMedArticle
17.
Elias  KM, Moromizato  T, Gibbons  FK, Christopher  KB.  Derivation and validation of the acute organ failure score to predict outcome in critically ill patients: a cohort study. Crit Care Med. 2015;43(4):856-864.PubMedArticle
18.
Osler  T, Rutledge  R, Deis  J, Bedrick  E.  ICISS: an International Classification of Disease-9 based injury severity score. J Trauma. 1996;41(3):380-386.PubMedArticle
19.
Ciesla  DJ, Tepas  JJ  III, Pracht  EE, Langland-Orban  B, Cha  JY, Flint  LM.  Fifteen-year trauma system performance analysis demonstrates optimal coverage for most severely injured patients and identifies a vulnerable population. J Am Coll Surg. 2013;216(4):687-695.PubMedArticle
20.
Tepas  JJ  III, Pracht  EE, Orban  BL, Flint  LM.  High-volume trauma centers have better outcomes treating traumatic brain injury. J Trauma Acute Care Surg. 2013;74(1):143-147.PubMedArticle
21.
Stephenson  S, Henley  G, Harrison  JE, Langley  JD.  Diagnosis based injury severity scaling: investigation of a method using Australian and New Zealand hospitalisations. Inj Prev. 2004;10(6):379-383.PubMedArticle
22.
Barell  V, Aharonson-Daniel  L, Fingerhut  LA,  et al.  An introduction to the Barell body region by nature of injury diagnosis matrix. Inj Prev. 2002;8(2):91-96.PubMedArticle
23.
Schisterman  EF, Whitcomb  BW.  Use of the Social Security Administration Death Master File for ascertainment of mortality status. Popul Health Metr. 2004;2(1):2.PubMedArticle
24.
Pencina  MJ, D’Agostino  RB  Sr, D’Agostino  RB  Jr, Vasan  RS.  Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med. 2008;27(2):157-172.PubMedArticle
25.
Moreau  R, Jalan  R, Gines  P,  et al; CANONIC Study Investigators of the EASL–CLIF Consortium.  Acute-on-chronic liver failure is a distinct syndrome that develops in patients with acute decompensation of cirrhosis. Gastroenterology. 2013;144(7):1426-1437, 1437.e1-1437.e9.PubMedArticle
26.
McPhail  MJ, Shawcross  DL, Abeles  RD,  et al.  Increased survival for patients with cirrhosis and organ failure in liver intensive care and validation of the Chronic Liver Failure-Sequential Organ Failure Scoring System. Clin Gastroenterol Hepatol. 2015;13(7):1353-1360.e8.PubMedArticle
27.
Trotter  JF, Olson  J, Lefkowitz  J, Smith  AD, Arjal  R, Kenison  J.  Changes in international normalized ratio (INR) and Model for Endstage Liver Disease (MELD) based on selection of clinical laboratory. Am J Transplant. 2007;7(6):1624-1628.PubMedArticle
28.
Linde-Zwirble  WT, Angus  DC.  Severe sepsis epidemiology: sampling, selection, and society. Crit Care. 2004;8(4):222-226.PubMedArticle
29.
Henley  G, Harrison  JE. Injury Severity Scaling: A Comparison of Methods for Measurement of Injury Severity. Canberra: Australian Institute of Health and Welfare; 2009. Injury technical paper series No. 10. Catalog No. INJCAT 126.
30.
Quan  H, Sundararajan  V, Halfon  P,  et al.  Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):1130-1139.PubMedArticle
Original Investigation
January 2016

Association of Model for End-Stage Liver Disease Score and Mortality in Trauma Patients With Chronic Liver Disease

Author Affiliations
  • 1Division of Trauma, Burns, and Surgical Critical Care, Department of Surgery, Brigham and Women's Hospital, Boston, Massachusetts
  • 2Division of Trauma, Emergency Surgery, and Surgical Critical Care, Department of Surgery, Massachusetts General Hospital, Boston
  • 3Department of Medicine, Brigham and Women's Hospital, Boston, Massachusetts
  • 4Division of Pulmonary and Critical Care Medicine, Department of Medicine, Massachusetts General Hospital, Boston
  • 5The Nathan E. Hellman Memorial Laboratory, Renal Division, Department of Medicine, Brigham and Women's Hospital, Boston, Massachusetts
JAMA Surg. 2016;151(1):41-48. doi:10.1001/jamasurg.2015.3114
Abstract

Importance  The Model for End-Stage Liver Disease (MELD) score is predictive of trauma outcomes.

Objective  To determine whether a decrease in MELD score is associated with improved mortality in critically ill trauma patients.

Design, Setting, and Participants  We performed a retrospective registry study of critically ill trauma patients 18 years or older with chronic liver disease treated between August 3, 1998, and January 5, 2012, at 2 level I trauma centers in Boston, Massachusetts. The consecutive sample included 525 patients (male, 373 [71.0%]; white, 399 [76.0%]; mean [SD] age, 55.0 [12.4] years).

Exposures  Change in MELD score from intensive care unit (ICU) admission to 48 to 72 hours later.

Main Outcomes and Measures  Thirty-day all-cause mortality.

Results  The mean (SD) MELD score at ICU admission was 19.3 (9.7). The 30-day mortality was 21.9%. The odds of 30-day mortality with a change in MELD score of less than −2, −2 to −1, +1 to +4, and greater than +4 were 0.23 (95% CI, 0.10-0.51), 0.30 (95% CI, 0.10-0.85), 0.57 (95% CI, 0.27-1.20), and 1.31 (95% CI, 0.58-2.96), respectively, relative to a change in MELD score of 0 and adjusted for age, sex, race, Charlson/Deyo Index, sepsis, number of acute organ failures, International Classification of Diseases, Ninth Revision–based injury severity score, and ICU admission MELD score.

Conclusions and Relevance  A decrease in MELD score within 72 hours of ICU admission is associated with improved mortality.

Introduction

Cirrhosis and chronic liver disease (CLD) are important causes of morbidity and mortality in the United States and account for approximately 25 000 deaths per year.1Quiz Ref ID Acute deterioration of CLD, termed acute-on-chronic liver failure, often requires critical care and has high mortality rates of 36% to 85%.2 Cirrhosis is present in approximately 1% of admitted trauma patients.3 In the trauma and postoperative settings, cirrhosis is a well-established risk factor for death and complications.4 Even after minor injuries, cirrhosis results in a substantial increase in the risk of mortality.4

An observational study4 found an association between severity of cirrhosis in trauma patients and worse outcomes. The Child-Turcotte-Pugh classification and Model for End-Stage Liver Disease (MELD) scores have been used to predict outcomes in trauma patients with CLD. In the trauma population, the MELD has high utility because of its ease of use and objectivity. However, no data currently indicate whether any modification of liver function is associated with an improvement in outcomes after trauma.

Because MELD scores appear to be predictive of outcomes in critically ill trauma patients with CLD, we performed this study to determine the association between a change in MELD score and all-cause 30-day mortality after critical care initiation. We hypothesized that a decrease in MELD score would be associated with improvement in critical care outcomes among injured patients.

Methods
Source Population

We used administrative and laboratory data from individuals admitted to 2 teaching hospitals in Boston, Massachusetts: Brigham and Women’s Hospital (BWH), with 793 beds, and Massachusetts General Hospital (MGH), with 902 beds. The BWH and MGH are level I trauma centers with 45 000 to 47 000 hospital admissions per year. The BWH and MGH are members of Partners HealthCare, which is the largest health care organization in Massachusetts.

Data Sources

Data on all patients admitted to the BWH or MGH between August 3, 1998, and January 5, 2012, were obtained through the Research Patient Data Registry, a computerized registry that serves as a central data warehouse for all inpatient and outpatient records at Partners HealthCare sites.510 Approval for the study was granted by the Partners Human Research Committee (Institutional Review Board). Requirement for consent was waived because the data were analyzed anonymously.

Study Population

The inclusion criteria were age of 18 years or older, the presence of injuries identified by International Classification of Diseases, Ninth Revision (ICD-9) codes,11 CLD,5 assignment of the Current Procedural Terminology (CPT) code 99291 (critical care, first 30-74 minutes),6 and the presence of total bilirubin level, creatinine level, and international normalized ratio (INR) measured within the first 72 hours of intensive care unit (ICU) admission. During the study period, 23 721 unique patients met the inclusion criteria. Exclusions included 22 029 patients without CLD and 1167 patients with missing laboratory data (total bilirubin level, creatinine level, and INR within the first 72 hours of ICU admission). Thus, 525 patients constituted the total study cohort (Figure 1).

Exposure of Interest and Comorbidities

The exposure of interest was a change in MELD score12 from ICU admission to the maximum MELD score between 48 and 72 hours after ICU admission. The MELD score was calculated with the United Network for Organ Sharing modification as follows: [0.957 × ln(Serum Creatinine) + 0.378 × ln(Serum Bilirubin) + 1.120 × ln(INR) + 0.643] × 10 (if the patient is undergoing hemodialysis, the value for serum creatinine is automatically set to 4.0). Because the MELD uses log scale calculations, any creatinine, bilirubin, or INR value less than 1 is given a lower limit value of 1 to prevent generating a negative score. Changes in MELD cut points were modified from the work of others13 and defined a priori as less than −2, −2 to −1, 0, +1 to +4, and greater than +4. The −2 and +4 cut points were selected because it was believed that small decreases in creatinine during the 72 hours after ICU admission would more greatly affect outcome than increases in creatinine of the same magnitude.

In administrative data from Partners HealthCare, we identified patients with CLD using a validated assignment of ICD-9 codes for CLD (571.x), chronic hepatitis C (70.54), and chronic hepatitis B (70.32) before or during hospitalization for critical care.5 In the Research Patient Data Registry data set, it was found that 8.4% of patients had a CLD diagnosis associated with alcohol use.5 Definition and determination of the following covariates are outlined in the eMethods in the Supplement: Charlson/Deyo Index, ICU admission,6 race, sepsis,7,14 exposure to inotropes and vasopressors,8,9 acute kidney injury,7,10 noncardiogenic acute respiratory failure,15 emergency general surgery,16 packed red blood cells (PRBCs) transfused,8 number of acute organ failures,6,7,14,17 and ICD-9–derived Injury Severity Score (ICISS).1821 The trauma-related ICD-9 diagnosis codes were grouped into nature and body region of injury categories based on the Barell Injury Diagnosis Matrix.22

Assessment of Mortality

Information on vital status for the study cohort was obtained from the Social Security Administration Death Master File, which has high sensitivity and specificity for mortality,23 and which we validated for in-hospital and out-of-hospital mortality in our administrative database.6 The censoring date was March 15, 2012. The entire cohort had at least 90 days of follow-up.

End Points

The primary end point was 30-day all-cause mortality after critical care initiation. The prespecified secondary end point was 90-day all-cause mortality.

Power Calculations and Statistical Analysis

On the basis of our prior data, the 30-day mortality rate in critically ill trauma patients with CLD is 15.3%.6 In the current study cohort with 247 patients with an increase in MELD score, 55 patients with no change in MELD score, and 223 patients with a decrease in MELD score, an α error level of 5%, and a power of 80%, the minimum absolute increase in percent mortality that the study is powered to detect between patients with a decrease in MELD score and patients with an increase in MELD score is 11.0%.

Categorical covariates were described by frequency distribution and compared across MELD groups using contingency tables and χ2 testing. Continuous covariates were examined graphically and in terms of summary statistics and compared across exposure groups using 1-way analysis of variance. Unadjusted associations between MELD groups and outcomes were estimated by bivariable logistic regression models. Adjusted odds ratios were estimated by multivariable logistic regression models with inclusion of covariate terms thought to plausibly be associated with change in MELD score and mortality. Logistic regression was chosen to predict the likelihood of mortality occurring at 30 days given our exposure and covariates, which is most aligned with the hypothesis under study. For the primary model (30-day mortality), specification of each continuous covariate (as a linear vs categorical term) was adjudicated by the empirical association with the primary outcome using the Akaike Information Criterion. The discriminatory ability for 30-day mortality was quantified using the C statistic. Model fit was assessed using the Hosmer-Lemeshow χ2 goodness-of-fit test. We individually tested for effect modification by MELD scores at day 0 of ICU admission, hospital, emergency general surgery, vasopressors or inotropes, year of hospitalization, and number of PRBC transfusions by adding an interaction term to the multivariate models. Improvement in model performance by the inclusion of change in MELD score to other covariates was assessed by evaluating the significant net reclassification improvement or integrated discrimination improvement.24 A multivariable Cox proportional hazards regression model was used to illustrate survival among patients as MELD score changes.

Because the allocation of patients into different MELD score change groups was not randomized, it might have produced a resulting imbalance in the baseline risk profile for mortality. For the time to mortality, we estimated the survival curves according to MELD score group with the use of the Kaplan-Meier method and compared the results by means of the log-rank test. All P values presented are 2-tailed; P < .05 was considered statistically significant. All analyses are performed using STATA software, version 13.1MP (StataCorp).

Results

Table 1 gives the characteristics of the study population. Most patients were men (373 [71.0%]) and white (399 [76.0%]). The mean (SD) age at hospital admission was 55.0 (12.4) years. Quiz Ref IDA total of 194 individuals (37.0%) in the cohort had sepsis, 125 (23.8%) had acute kidney injury, and 146 (27.8%) had noncardiac acute respiratory failure. The mean (SD) MELD score at ICU admission was 19.3 (9.7). The mean (SD) change in MELD score was +0.38 (4.9). The median (interquartile range) of days between hospital and ICU admission was 0 (0-2) days. The crude 30- and 90-day mortality rates were 21.9% and 32.0%, respectively. There were 291 patients who subsequently died with 1140 person-years of follow-up, yielding a mortality rate of 255 per 1000 person-years. Table 1 indicates that age, acute kidney injury, noncardiogenic acute respiratory failure, renal replacement, operative status, and ICISS are significantly associated with 30-day mortality. Patient characteristics of the study cohort were stratified according to change in MELD score (Table 2). Factors that significantly differed between the change in MELD score groups included acute kidney injury, sepsis, and noncardiogenic acute respiratory failure. Table 3 indicates that the most common nature of the injury was classified as foreign body related.

Quiz Ref IDThe ICU admission MELD score was associated with mortality. The odds of 30-day mortality for each 1-point increase in ICU admission MELD score was 1.03 (95% CI, 1.01-1.06), adjusted for age, race, sex, Charlson/Deyo Index, sepsis, number of acute organ failures, and ICISS. The adjusted ICU admission MELD model had good calibration (Hosmer-Lemeshow χ2 = 6.33, P = .61) and moderate discrimination for 30-day mortality (C statistic, 0.68; 95% CI, 0.62-0.79).

Subsequent decreases in MELD score were associated with 30-day mortality and remained so after multivariable adjustment (Table 4). Quiz Ref IDThe odds of 30-day mortality in patients with a change in MELD score less than −2, −2 to −1, +1 to +4, and greater than +4 were 0.23 (95% CI, 0.10-0.51), 0.30 (95% CI, 0.10-0.85), 0.57 (95% CI, 0.27-1.20), and 1.31 (95% CI, 0.58-2.96), respectively, relative to patients with no change in MELD score and adjusted for ICU admission MELD score, age, race, sex, Charlson/Deyo Index, sepsis, number of acute organ failures, and ICISS (Table 4). This finding indicates that patients who have a change in MELD scores less than −2 have a 77% reduced odds of 30-day mortality. In patients with a change in MELD scores less than −2 (n = 174), the mean (SD) change in MELD score components was −0.39 (0.83) for INR, −0.12 (2.44) for total bilirubin level, and −0.59 (0.86) for creatinine level. Figure 2 illustrates the different survival curves for the change in MELD score groups. The log-rank test indicates that there is a significant difference in the overall survival distributions among the patient groups.

The adjusted change in the MELD score revealed good calibration (Hosmer-Lemeshow χ2 = 6.90, P = .55) and discrimination (C statistic = 0.74; 95% CI, 0.69-0.79). Differences in discrimination between the ICU admission MELD score and the change in MELD score are significant (χ2 = 6.80, P = .009). Furthermore, the net reclassification improvement was estimated at 8.2% (P = .02), and the integrated discrimination improvement was estimated at 5.0% (P < .001). The net reclassification improvement and integrated discrimination improvement suggest that the addition of the change in MELD score to the ICU admission MELD model results in a significant improvement in model performance. When the analysis includes only patients with at least 2 sequential days of critical care–associated CPT codes (codes 99291 or 99292, n = 435), although decreased in power, the decrease in MELD-improved, 30-day mortality association is preserved after adjustment for age, sex, race, Charlson/Deyo Index, sepsis, number of acute organ failures, ICISS, and MELD score at ICU admission. Furthermore, the hazard ratio of mortality adjusted for age, sex, race, sepsis, Charlson/Deyo Index, number of acute organ failures, MELD score at ICU admission, and ICISS in patients with a change in MELD score less than −2, −2 to −1, +1 to +4, and greater than +4 were 0.53 (95% CI, 0.35-0.81), 0.59 (95% CI, 0.35-0.99), 0.84 (95% CI, 0.56-1.27), and 1.07 (95% CI, 0.67-1.71), respectively, relative to patients without a change in MELD score.

There is no significant effect modification of the change in MELD score and 30-day mortality association on the basis of MELD values at day 0 of ICU admission (P for interaction = .42), hospital (P for interaction = .15), emergency general surgery (P for interaction = .13), or vasopressors or inotropes (P for interaction = .21). Effect modification is present with year of hospitalization (P for interaction = .005) and PRBC transfusions (P for interaction = .01). The 30-day mortality in the cohort before 1998-2007 (n = 305) was 25% compared with 18% after 2007 (n = 220). The 30-day mortality was 18% for those with no PRBC transfusions and 36% for those with more than 4 units of PRBCs. Individually adding a hospital, year, or a PRBC transfusion term to the final model does not alter the effect size or significance of the change in MELD score and 30-day mortality association. When the cohort is constrained by year of hospitalization (before or after 2007), although limited by power, the estimates maintain significance and directionality with a MELD score less than −2. With addition of a vasopressor or inotrope term, the directionality of the association between a decrease in MELD score and decreased 30-day mortality remains, but the association between an increase in MELD score and increased 30-day mortality is attenuated. For the primary model, there was no multicollinearity as determined by variance inflation factor using a cut point of 10. Furthermore, no multicollinearity exists between the change in MELD score and acute kidney injury.

Discussion

Our 2-center study aimed to determine whether a decrease in MELD score was associated with decreased mortality in critically ill trauma patients with CLD. In both unadjusted and adjusted analyses, we found a decrease in the odds and risk of 30-day mortality relative to a decrease in MELD score within 48 to 72 hours after ICU admission. Quiz Ref IDThe change in MELD score improves model performance and discrimination when added to a model that includes MELD score at ICU admission; however, because our study is observational and not interventional, a causal relationship between change in MELD score and outcomes after trauma cannot be inferred from these data alone.

The management of critically ill patients with cirrhosis is resource intensive; in this patient population, more than 35% of the total cost of ICU care is spent on those who do not survive. The need to optimize hospital resources has resulted in calls for an improved definition of this subpopulation and more reliable, validated measures for predicting patient outcomes. In patients with cirrhosis, the Sequential Organ Failure Assessment (SOFA) score and chronic liver failure–specific modification of the SOFA score (CLIF-SOFA)25 have higher accuracy for predicting mortality in patients with cirrhosis than MELD or Acute Physiology and Chronic Health Evaluation (APACHE) II.26

The number of extrahepatic acute organ failures is associated with mortality in ICU patients with CLD. In our cohort, patients with acute kidney injury (RIFLE [risk, injury, failure, loss, and end-stage kidney disease] class injury or failure) have a 30-day mortality rate of 32%. Because the MELD score has 3 components (serum bilirubin level, serum creatinine level, and INR), the association between changes in the MELD and mortality may reflect improvements in renal function in the 48 to 72 hours after ICU admission. The INR and creatinine level have higher multiplicative values and larger contributions to mortality differences than the total bilirubin level in the MELD score. Changes in INR or creatinine level alone may result in up to 20% differences in MELD score.27

The present study has all the inherent limitations of a retrospective study. Because our study is observational and lacks a clear dose-response relationship, causality is limited. Selection bias may exist because the patient cohort under study had MELD score components investigated for a particular reason that may be absent in other patients. Ascertainment bias may be present because a large number of patients did not have all MELD components measured twice in the first 72 hours after ICU admission and, thus, were not included in our study. These issues may decrease the generalizability of our results to all trauma patients. Reliance on ICD-9 codes to determine the Charlson/Deyo Index, ICISS, sepsis diagnosis, and CLD does not measure the true incidence, which is likely higher.28 Use of the CPT code 99291 to identify ICU admissions will not capture patients admitted to an ICU who are not assigned the CPT code 99291; furthermore, a small percentage of those with a CPT code 99291 assignment in our data set received critical care only in the emergency department.6 Although we used a validated approach that has a high positive predictive value,5 it is possible that misclassification of the CLD diagnosis occurred because the diagnosis was made with an ICD-9 code combination. We cannot exclude the possibility that other unmeasured variables influence mortality independently of the change in MELD score, which may have biased estimates. Furthermore, because of a lack of physiologic and hemodynamic data present in our data set, we are unable to compare the discrimination for mortality of the change in MELD score to the SOFA, CLIF-SOFA, or APACHE II. Because of limitations of the data set, we cannot determine with precision what treatment modality may have led to improvement in the MELD score. Although we adjusted for multiple potential confounders, there might be residual confounding of unmeasured variables, leading to observed differences in outcomes. The relatively wide CIs reflect lower precision and more statistical instability than what might be expected in a larger study.

The present study has several strengths and is unique in that it investigates the association of a change in MELD score on 30-day all-cause mortality. The current study has ample statistical power to detect a clinically relevant difference in 30-day all-cause mortality if one exists. In determining the ICISS, we included deaths that occurred outside the hospital (up to 30 days after ICU admission), which likely improves the accuracy of the resulting severity estimates.29 Our inclusion of the Charlson/Deyo Index likely improves the predictive ability of ICISS.11 We used previously validated approaches to define CLD, Charlson/Deyo Index, sepsis, and acute kidney injury.5,7,30 Although we do not have cause of death data, determination of all-cause mortality was based on the Social Security Administration Death Master File and validated in the administrative data set.6

Conclusions

Our results reveal an association between a decrease in MELD score within 48 to 72 hours of ICU admission and decreased mortality after trauma in critically ill patients with CLD. Additional studies are needed to confirm our findings and establish mechanisms underlying the apparent mortality benefit from decreased MELD score after trauma in patients with CLD.

Back to top
Article Information

Accepted for Publication: June 7, 2015.

Corresponding Author: Kenneth B. Christopher, MD, The Nathan E. Hellman Memorial Laboratory, Renal Division, Department of Medicine, Brigham and Women’s Hospital, 75 Francis St, Medical Research Bldg 418, Boston, MA 02115 (kbchristopher@partners.org).

Published Online: September 30, 2015. doi:10.1001/jamasurg.2015.3114.

Author Contributions: Dr Christopher had full access to all 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: Salim, Askari, Simon, Christopher.

Acquisition, analysis, or interpretation of data: Peetz, Salim, De Moya, Olufajo, Simon, Gibbons, Christopher.

Drafting of the manuscript: Peetz, Askari, Christopher.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Peetz, Olufajo, Christopher.

Administrative, technical, or material support: Salim, Askari, Christopher.

Study supervision: Peetz, Salim, De Moya.

Conflict of Interest Disclosures: None reported.

Additional Contributions: Shawn Murphy, MD, PhD, Henry Chueh, MD, and the Partners HealthCare Research Patient Data Registry group facilitated use of the database.

Additional Information: This article is dedicated to the memory of our dear friend and colleague Nathan Edward Hellman, MD, PhD.

References
1.
Schuppan  D, Afdhal  NH.  Liver cirrhosis. Lancet. 2008;371(9615):838-851.PubMedArticle
2.
Levesque  E, Hoti  E, Azoulay  D,  et al.  Prospective evaluation of the prognostic scores for cirrhotic patients admitted to an intensive care unit. J Hepatol. 2012;56(1):95-102.PubMedArticle
3.
Tinkoff  G, Rhodes  M, Diamond  D, Lucke  J.  Cirrhosis in the trauma victim; effect on mortality rates. Ann Surg. 1990;211(2):172-177.PubMedArticle
4.
Christmas  AB, Wilson  AK, Franklin  GA, Miller  FB, Richardson  JD, Rodriguez  JL.  Cirrhosis and trauma: a deadly duo. Am Surg. 2005;71(12):996-1000.PubMed
5.
Hug  BL, Lipsitz  SR, Seger  DL, Karson  AS, Wright  SC, Bates  DW.  Mortality and drug exposure in a 5-year cohort of patients with chronic liver disease. Swiss Med Wkly. 2009;139(51-52):737-746.PubMed
6.
Zager  S, Mendu  ML, Chang  D,  et al.  Neighborhood poverty rate and mortality in patients receiving critical care in the academic medical center setting. Chest. 2011;139(6):1368-1379.PubMedArticle
7.
Moromizato  T, Litonjua  AA, Braun  AB, Gibbons  FK, Giovannucci  E, Christopher  KB.  Association of low serum 25-hydroxyvitamin D levels and sepsis in the critically ill. Crit Care Med. 2014;42(1):97-107.PubMedArticle
8.
Purtle  SW, Moromizato  T, McKane  CK, Gibbons  FK, Christopher  KB.  The association of red cell distribution width at hospital discharge and out-of-hospital mortality following critical illness. Crit Care Med. 2014;42(4):918-929.PubMedArticle
9.
McKane  CK, Marmarelis  M, Mendu  ML, Moromizato  T, Gibbons  FK, Christopher  KB.  Diabetes mellitus and community-acquired bloodstream infections in the critically ill. J Crit Care. 2014;29(1):70-76.PubMedArticle
10.
Braun  AB, Litonjua  AA, Moromizato  T, Gibbons  FK, Giovannucci  E, Christopher  KB.  Association of low serum 25-hydroxyvitamin D levels and acute kidney injury in the critically ill. Crit Care Med. 2012;40(12):3170-3179.PubMedArticle
11.
Davie  G, Cryer  C, Langley  J.  Improving the predictive ability of the ICD-based Injury Severity Score. Inj Prev. 2008;14(4):250-255.PubMedArticle
12.
Malinchoc  M, Kamath  PS, Gordon  FD, Peine  CJ, Rank  J, ter Borg  PC.  A model to predict poor survival in patients undergoing transjugular intrahepatic portosystemic shunts. Hepatology. 2000;31(4):864-871.PubMedArticle
13.
Kim  JD, Choi  JY, Kwon  JH,  et al.  Performance of posttransplant Model for End-Stage Liver Disease (MELD) and delta-MELD scores on short-term outcome after living donor liver transplantation. Transplant Proc. 2009;41(9):3766-3768.PubMedArticle
14.
Martin  GS, Mannino  DM, Eaton  S, Moss  M.  The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med. 2003;348(16):1546-1554.PubMedArticle
15.
Cooke  CR, Erickson  SE, Eisner  MD, Martin  GS.  Trends in the incidence of noncardiogenic acute respiratory failure: the role of race. Crit Care Med. 2012;40(5):1532-1538.PubMedArticle
16.
Shafi  S, Aboutanos  MB, Agarwal  S  Jr,  et al; AAST Committee on Severity Assessment and Patient Outcomes.  Emergency general surgery: definition and estimated burden of disease. J Trauma Acute Care Surg. 2013;74(4):1092-1097.PubMedArticle
17.
Elias  KM, Moromizato  T, Gibbons  FK, Christopher  KB.  Derivation and validation of the acute organ failure score to predict outcome in critically ill patients: a cohort study. Crit Care Med. 2015;43(4):856-864.PubMedArticle
18.
Osler  T, Rutledge  R, Deis  J, Bedrick  E.  ICISS: an International Classification of Disease-9 based injury severity score. J Trauma. 1996;41(3):380-386.PubMedArticle
19.
Ciesla  DJ, Tepas  JJ  III, Pracht  EE, Langland-Orban  B, Cha  JY, Flint  LM.  Fifteen-year trauma system performance analysis demonstrates optimal coverage for most severely injured patients and identifies a vulnerable population. J Am Coll Surg. 2013;216(4):687-695.PubMedArticle
20.
Tepas  JJ  III, Pracht  EE, Orban  BL, Flint  LM.  High-volume trauma centers have better outcomes treating traumatic brain injury. J Trauma Acute Care Surg. 2013;74(1):143-147.PubMedArticle
21.
Stephenson  S, Henley  G, Harrison  JE, Langley  JD.  Diagnosis based injury severity scaling: investigation of a method using Australian and New Zealand hospitalisations. Inj Prev. 2004;10(6):379-383.PubMedArticle
22.
Barell  V, Aharonson-Daniel  L, Fingerhut  LA,  et al.  An introduction to the Barell body region by nature of injury diagnosis matrix. Inj Prev. 2002;8(2):91-96.PubMedArticle
23.
Schisterman  EF, Whitcomb  BW.  Use of the Social Security Administration Death Master File for ascertainment of mortality status. Popul Health Metr. 2004;2(1):2.PubMedArticle
24.
Pencina  MJ, D’Agostino  RB  Sr, D’Agostino  RB  Jr, Vasan  RS.  Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med. 2008;27(2):157-172.PubMedArticle
25.
Moreau  R, Jalan  R, Gines  P,  et al; CANONIC Study Investigators of the EASL–CLIF Consortium.  Acute-on-chronic liver failure is a distinct syndrome that develops in patients with acute decompensation of cirrhosis. Gastroenterology. 2013;144(7):1426-1437, 1437.e1-1437.e9.PubMedArticle
26.
McPhail  MJ, Shawcross  DL, Abeles  RD,  et al.  Increased survival for patients with cirrhosis and organ failure in liver intensive care and validation of the Chronic Liver Failure-Sequential Organ Failure Scoring System. Clin Gastroenterol Hepatol. 2015;13(7):1353-1360.e8.PubMedArticle
27.
Trotter  JF, Olson  J, Lefkowitz  J, Smith  AD, Arjal  R, Kenison  J.  Changes in international normalized ratio (INR) and Model for Endstage Liver Disease (MELD) based on selection of clinical laboratory. Am J Transplant. 2007;7(6):1624-1628.PubMedArticle
28.
Linde-Zwirble  WT, Angus  DC.  Severe sepsis epidemiology: sampling, selection, and society. Crit Care. 2004;8(4):222-226.PubMedArticle
29.
Henley  G, Harrison  JE. Injury Severity Scaling: A Comparison of Methods for Measurement of Injury Severity. Canberra: Australian Institute of Health and Welfare; 2009. Injury technical paper series No. 10. Catalog No. INJCAT 126.
30.
Quan  H, Sundararajan  V, Halfon  P,  et al.  Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):1130-1139.PubMedArticle
×