REVIEWS

The Model for End-Stage Liver Disease (MELD)

Patrick S. Kamath and W. Ray Kim

The Model for End-stage Liver Disease (MELD) was initially created to predict survival in
patients with complications of portal hypertension undergoing elective placement of tran-
sjugular intrahepatic portosystemic shunts. The MELD which uses only objective variables
was validated subsequently as an accurate predictor of survival among different populations
of patients with advanced liver disease. The major use of the MELD score has been in
allocation of organs for liver transplantation. However, the MELD score has also been shown
to predict survival in patients with cirrhosis who have infections, variceal bleeding, as well as
in patients with fulminant hepatic failure and alcoholic hepatitis. MELD may be used in
selection of patients for surgery other than liver transplantation and in determining optimal
treatment for patients with hepatocellular carcinoma who are not candidates for liver trans-
plantation. Despite the many advantages of the MELD score, there are approximately 15%-
20% of patients whose survival cannot be accurately predicted by the MELD score. It is
possible that the addition of variables that are better determinants of liver and renal function
may improve the predictive accuracy of the model. Efforts at further refinement and valida-
tion of the MELD score will continue. (HEPATOLOGY 2007;45:797-805.)

F
ebruary 27, 2007 marked the fifth anniversary of
the Model for End-Stage Liver Disease (MELD)
becoming the standard by which priorities in do-
nor liver allocation were determined. Since the score was
first derived in a relatively small number of patients un-
dergoing the transjugular intrahepatic portosystemic
shunts (TIPS) procedure, it has been validated in many
different populations of patients with liver disease.
Within a relatively short period of time, MELD became a
common metric by which the severity of liver disease
could be accurately described.
In this paper, we review the initial development and
validation of the MELD score, its application in organ
allocation and management of patients with a variety of
Abbreviations: ASA, American Society of Anesthesiologists; CTP, hild-Turcotte-
Pugh; HCC, hepatocellular carcinoma; MELD, Model for End-Stage Liver Dis-
ease; SBP, spontaneous bacterial peritonitis; SOFA, Sequential Organ Failure;
TIPS, transjugular intrahepatic portosystemic shunts; UNOS, United Network of
Organ Sharing.
From the Advanced Liver Disease Study Group, Miles and Shirley Fiterman
Center for Digestive Diseases, Mayo Clinic College of Medicine, Rochester, MN.
Supported by grants from the National Institute of Diabetes, Digestive and
Kidney Diseases (DK-34238 and DK-61617).
Address correspondence to: Patrick S. Kamath, M.D., Professor of Medicine,
Mayo Clinic College of Medicine, Consultant in Gastroenterology and Hepatology,
Miles and Shirley Fiterman Center for Digestive Diseases, Mayo Clinic, 200 First
Street SW, Rochester, MN 55905. E-mail: Kamath.Patrick@mayo.edu; fax:
507-284-0538.
Copyright © 2007 by the American Association for the Study of Liver Diseases.
Published online in Wiley InterScience (www.interscience.wiley.com).
DOI 10.1002/hep.21563
Potential conflict of interest: Nothing to report.
liver conditions, its strengths and limitation, and current
and future efforts to refine and improve it further.
Creation and Validation of MELD
MELD was initially created to predict survival follow-
ing elective placement of TIPS.1 The model was subse-
quently validated as a predictor of survival in several
cohorts of patients with varying levels of liver disease se-
verity (e.g., hospitalized and ambulatory patients), as well
as patients of geographically and temporally diverse ori-
gin.2 The survival model was initially termed the “Mayo
End-Stage Liver Disease” or “MELD” model to acknowl-
edge the affiliation of the investigators who created the
model. During discussions leading to the establishment of
MELD as the basis for prioritization of organs for liver
transplantation,3,4 we changed the name to “Model for
End-Stage Liver Disease” which maintained the acronym
“MELD”, but removed the association with a particular
institution, a process that was thought would lead to
wider acceptance of the model.
MELD incorporates 3 widely available laboratory vari-
ables including the international normalized ratio (INR),
serum creatinine, and serum bilirubin. The original
mathematical formula for MELD is: MELD ⫽ 9.57 ⫻
loge(creatinine) ⫹ 3.78 ⫻ Loge(total bilirubin) ⫹ 11.2 ⫻
Loge(INR) ⫹ 6.43.
The score can be calculated on handheld computing
devices, and is available at www.mayoclinic.org/gi-rst/
mayomodel5.html. When the model was initially created
the etiology of cirrhosis was also included. In the TIPS
797
798 KAMATH AND KIM
population, the etiology of liver disease was important in
that patients with alcoholic liver disease and cholestatic
liver disease undergoing TIPS procedures had a better
survival than those patients with viral hepatitis-related or
other causes of cirrhosis.
In subsequent studies, we confirmed that the etiology
of cirrhosis was a less important variable in determining
survival in other patient cohorts with end-stage liver dis-
ease. Therefore, etiology of liver disease was removed as a
variable from the model while still preserving its accura-
cy.5 The advantage of dropping etiology of cirrhosis as a
variable was that the subjective element in determining
etiology was removed, and the model could be based
purely on objective laboratory variables.
MELD has been validated as a predictor of survival in
independent groups of patients with a wide variety of liver
diseases.5,6 In these studies, the accuracy of MELD was
evaluated by its ability to rank patients according to risk
for mortality determined by the “concordance statistic” or
the “c” statistic. For example, a c-statistic of 0.7 indicates
that patients with a higher MELD score will die earlier
than patients with a lower MELD score 7 of 10 times. A
c-statistic of 0.7 is thought to have reasonable clinical
utility, while a c-statistic of ⱖ0.8 in a prediction model
lends strong support to its accuracy. Most studies that
evaluated MELD to rank patients according to their risk
of mortality have yielded “c”-statistics upwards of 0.8,
and usually superior to the Child-Turcotte-Pugh (CTP)
class. Addition of complications such as ascites, encepha-
lopathy, variceal bleed, and SPB do not improve MELD
significantly; quantitative tests of liver function are also
not superior to MELD in predicting survival.7
Application of MELD
MELD in Liver Transplantation
Prior to February 27, 2002, patients were prioritized
for receiving organs for liver transplantation based on
their United Network of Organ Sharing (UNOS) status,
a reflection of their CTP score. Given that the waiting list
for liver transplantation approached 20,000 patients, and
there were only 3 categories on the waiting list for patients
with cirrhosis, namely Status 2A, Status 2B, and Status 3,
time spent on the waiting list became the major determi-
nant of who would receive a liver transplant. Therefore,
this policy placed at a disadvantage patients who were at a
high risk for mortality but who were listed late in their
disease course and had not accrued enough waiting time.
In 1998, the Institute of Medicine decreed that a new
allocation policy be put in place based on objective vari-
ables and which de-emphasized waiting time, that is, the
“sickest first” policy. This led to MELD which is based on
objective variables and could accurately rank patients with
HEPATOLOGY, March 2007
cirrhosis according to risk of mortality, replacing the then
current CTP-based organ allocation system.3,4,8 In apply-
ing MELD to organ allocation, UNOS made several
changes to how the MELD score was to be calculated. The
lower limit for serum creatinine, serum bilirubin, and
INR was fixed at 1 so that there would be no negative
scores; the upper limit of serum creatinine was capped at
4 mg/dl.
Implementation of MELD led to an immediate reduc-
tion in liver transplant waiting list registrations for the
first time in history of liver transplantation (12% decrease
in 2002)4 because accrual of waiting time was no longer
necessary. Accurate prediction of short-term mortality in
the vast majority (83%-87%) of wait-listed candidates9
led to a reduction of almost 15% in the mortality on the
waiting list.10 The number of deaths of patients on the
wait list increased up to 2001 (Fig. 1); since implementa-
tion of MELD in 2002, the number of deaths showed a
substantial decrease from 2046 in 2001 to 1364 in 2005.
Although this reduction in mortality is in part attribut-
able to a modest increase in available organs (4,671 in
2001 versus 5,160 in 2005), there is a wide consensus that
MELD has made a significant contribution to reducing
the mortality on the waiting list.11 A recent analysis
showed that the reduction in mortality occurred only
among patients with chronic liver disease (in whom
MELD is used to allocate organs), but not among patients
with fulminant liver disease in status 1 (in whom MELD
is not used), suggesting that at least part of the decrease in
waitlist mortality may be attributed to MELD-based or-
gan allocation.12 Moreover, the median waiting time to
liver transplantation decreased from 656 days to 416 days
in the MELD era.13 Several countries have replaced the
CTP score with MELD to rank patients according to
Fig. 1. The number and ratio of LTx (liver transplant) candidates
removed from the waiting list due to transplantation (gray) and death
(black) in the United States.
HEPATOLOGY, Vol. 45, No. 3, 2007
mortality risk.14 In a study from Australia, clinical judg-
ment, which is often used in centers to determine which
patient should rank higher on a waiting list for mortality,
has been proven to be inferior to the MELD score in
determining survival.15
In contrast to the clear benefit of accurately estimating
mortality on the waiting list, MELD has not been found
to be as useful in predicting mortality following liver
transplantation.16-19 Mortality in the post transplantation
period is related not only to the degree of liver dysfunction
prior to transplantation, but to other factors, such as do-
nor characteristics, experience of the transplantation
team, and random postoperative complications which
cannot be predicted. Moreover, patient selection by phy-
sicians will tend to negate the effect of pre-transplant
MELD on post transplant survival. Consequently, almost
all models which attempt to determine post transplant
survival are not clinically useful. Thus, MELD score be-
fore liver transplant is not predictive of post liver trans-
plant outcome because of relatively poor correlation
between pre-transplant disease severity and post trans-
plant outcome.
Likewise, in small studies on living donor liver trans-
plantation, pre-transplant MELD scores had little impact
on post transplant survival.20 There have been proposals
that organ allocation take into account donor and recipi-
ent factors simultaneously such that marginal grafts not
be used for patients with high MELD scores in the hopes
to improve post liver transplantation outcome.21,22 Simi-
larly, in patients with MELD scores ⱕ14, mortality with
transplantation was found to be higher than that of pa-
tients with the same MELD score not transplanted.23 One
must exercise caution, however, in interpreting these re-
sults, because they are simply observational (as opposed to
randomized) data.
To the degree that disease severity before liver trans-
plantation affects post transplantation morbidity and
complications, healthcare resources used correlate with
pre-transplantation MELD score. At the level of individ-
ual patient, resource utilization, as judged by days in in-
tensive care, red cell transfusions, and duration of
hospitalization was higher with higher pre-transplant
MELD scores.24,25 On the other hand, on an aggregate
level, MELD-based organ allocation has not increased
healthcare resource utilization. Based on a nationally rep-
resentative hospital utilization database, we have shown
that resources used did not increase since MELD was
implemented.26 Thus, MELD-based allocation consis-
tently directs organs to the sickest patients so that those
patients in the previous era who had high MELD but
insufficient waiting time are undergoing transplantation
KAMATH AND KIM 799
earlier, reducing the number of “outliers” that incurred
astronomical costs to the health care system.
Patients undergoing retransplantation are at a higher
risk of mortality after transplantation than those under-
going primary liver transplantation. The 2-year survival
after retransplantation is lower than that after primary
liver transplantation, with the difference in survival being
greatest in patients with MELD ⬎25. Moreover, retrans-
plantation more than 2 years after the primary transplan-
tation is associated with poor survival independent of the
MELD score.27 These data notwithstanding, the most
immediate question about MELD in allocation for re-
transplantation is whether the risk of death on waiting list
is different between primary and repeated liver transplant
candidates. Retransplantation candidates are registered
on the waiting list with significantly higher MELD than
primary liver transplant candidates (22 versus 14) and
thus retransplantation candidates as a whole experienced
higher mortality rate on the waiting list.
It is important to note that although the modifications
created by UNOS in calculating the MELD score have
some rationale, the modifications have been empirical
rather than based on validated studies. For instance, the
upper value of serum creatinine has been capped at as high
as 4 mg/dl, allowing inclusion of patients with intrinsic
renal disease. Thus, there has been a higher rate of com-
bined kidney and liver transplants than previously, but
without significant change in survival.28 The need for
renal replacement therapy is higher if MELD scores are
greater than 24,29 even though there is a suggestion that
the prevalence of chronic renal disease up to 2 years after
transplantation has not increased.30
Another area which requires more validation is the as-
signment of MELD scores to patients with hepatocellular
carcinoma. The initial recommendation for allocation of
24 MELD points for patients with Stage 2 HCC was
found to be too high, and this was later reduced to 22.31 It
remains to be seen whether this reduction in score is more
accurately reflective of the patients with hepatocellular
carcinoma removed from the waiting list because they
exceeded the Milan criteria.
Prediction of Long-term Survival in Patients with
Cirrhosis
In a recent systematic review of 118 studies outlining
the natural history and prognostic indicators of survival in
cirrhosis,32 MELD and CTP score were recognized as
predictors of long-term survival in patients with decom-
pensated cirrhosis. Recently, both MELD and HVPG
were shown to be independent predictors of survival. The
“c” statistic for predicting survival was 0.71 for the
MELD score, and the addition of HVPG and age in-
800 KAMATH AND KIM
creased the predictive mortality to 0.76, which was not
statistically significant. MELD is an accurate predictor of
long-term mortality, even in patients with chronic hepa-
titis B, with or without cirrhosis.33 Whereas patients with
NASH-related cirrhosis had a lower risk of mortality than
patients with HCV-related cirrhosis, even when “decom-
pensated”; MELD was an accurate predictor of survival in
both groups of patients.34
Application of MELD Beyond ESLD
Variceal Bleeding. Retrospective studies have shown
that both MELD and CTP are accurate predictors of
survival in patients with variceal bleeding.35 MELD and
hepatocellular carcinoma were better predictors of sur-
vival than CTP score in a study of 172 patients with
cirrhosis and first variceal bleed, with MELD score dis-
criminating between patients at high risk (ⱖ15) and low
risk (⬍15) for mortality.36 Our own unpublished data
confirms the superiority of MELD over the CTP score in
determining mortality following a variceal bleed.
Infections in Patients with Cirrhosis. Both studies
addressing the predictability of MELD in determining
mortality in patients with infections leading to renal fail-
ure have demonstrated that the only predictors of survival
in these patients were the MELD score, and the type of
hepatorenal syndrome; CTP score was not independently
predictive of mortality.37,38 These data would suggest that
patients with a high MELD score are either more likely to
get infected, or that patients with a high MELD score are
more likely to die as a result of spontaneous bacterial
peritonitis (SBP) because outpatients with SBP typically
have lower MELD scores than inpatients with SBP, and
better survival.39
Fulminant Hepatic Failure. Among 312 patients
with non-acetaminophen-induced FHF in the UNOS
database, MELD was a highly significant predictor of 30
day mortality (P ⬍ 0.0001).40 Of note, patients with
primary graft nonfunction and patients with hepatic ar-
tery thrombosis had a lower risk of 30-day mortality than
patients with FHF and did not show a significant associ-
ation between transplantation and survival. On the other
hand, patients with FHF experienced the greater benefit
with liver transplantation, survival improving from 58%
at 30 days without transplantation to 91% with liver
transplantation (P ⬍ 0.0001). This raises the question as
to whether MELD should also be used to prioritize can-
didates with FHF, that is, UNOS Status 1, for liver trans-
plantation.
In a large prospective study from Denmark, the utility
of MELD in determining the onset of FHF in patients
with acetaminophen-induced liver failure was demon-
strated (“c” statistic: 0.92). However, once patients devel-
HEPATOLOGY, March 2007
oped FHF, MELD was less accurate in predicting
mortality.41 In a U.S. multicenter study in patients with
FHF secondary to hepatitis A, MELD had a “c” statistic of
only 0.7 in ranking patients according to mortality risk.42
These studies would suggest that, in patients with FHF in
addition to the MELD score, there are other variables
such as the degree of intracranial pressure important in
predicting mortality. The U.S. study demonstrated that a
model using serum creatinine, serum bilirubin, need for
endotracheal intubation, and vasopressor support had a
“c” statistic of 0.89. However, there were only 4 deaths,
and only 9 of the 29 patients underwent liver transplan-
tation. Therefore, the limited number of endpoints for a 4
variable model (typically, 40 events would be necessary
for internal validation of a 4 variable model) is likely to be
associated with “model over-fitting”, and needs validation
in an independent data set before being widely used.
Alcoholic Hepatitis. In all 3 studies that compared
the MELD score to the Maddrey score,43 MELD was a
more accurate predictor of mortality.44-46 This is not sur-
prising because the MELD score has both the variables
included in the Maddrey score, namely the prothrombin
time and bilirubin but, in addition, has an index of renal
dysfunction which is the serum creatinine. The cut-off of
the MELD score for determining severe alcoholic hepati-
tis is ⬎21 which is associated with 3-month mortality of
20%, whereas patients with MELD score ⱕ11 have ex-
cellent survival.45
Other Chronic Liver Diseases. MELD has been
shown to be a useful predictor even in patients n whom
cirrhosis was not clearly documented. In patients with
chronic hepatitis B, MELD and antiviral treatments
(lamivudine) were independent predictors of survival.33
In patients admitted to the Intensive Care Unit, the
“Royal Free Model” had similar discrimination ability as
the MELD and Sequential Organ Failure (SOFA) score47
in predicting mortality. All three scores were superior to
APACHE II or CTP scores.48 It must be pointed out that
both the Royal Free Model and the SOFA score include
variables reflecting multiple system organ failure, and are
more likely to “reflect” the dying process rather than be-
ing predictors of mortality. Similarly, in patients with
acute on chronic liver failure, MELD and hepatic enceph-
alopathy predicted survival,49 especially when MELD was
ⱖ30.17
MELD in the Management of Patients with ESLD
Transjugular Intrahepatic Portosystemic Shunts.
MELD was originally developed in patients undergoing
TIPS and may be used most appropriately to predict
probability of survival after the procedure. The indication
for TIPS, whether refractory ascites or variceal bleeding, is
HEPATOLOGY, Vol. 45, No. 3, 2007
not an independent variable determining survival. Emer-
gency TIPS, that is TIPS carried out in a patient actively
bleeding in spite of 2 sessions of endoscopic therapy car-
ried out within 24 hours, is a predictor of mortality, but
these patients also usually have a higher MELD score.1 It
is not clear whether TIPS adds additional mortality to
that predicted by the MELD score in patients with com-
plications of portal hypertension. That is, it is not clear,
for example, whether a patient with a MELD score of 24
undergoing TIPS has a higher mortality than a patient
with similar complications and a MELD score of 24 not
undergoing TIPS. Nonetheless, TIPS is associated with a
higher risk of mortality than seen in patients on the wait-
ing list for liver transplantation with identical MELD
scores, at least a small additional risk of procedure-related
mortality,50 and a risk of morbidity as well as additional
expense.
The importance of MELD as an independent predic-
tor of death in patients undergoing TIPS has been con-
firmed by other studies.38,51 The MELD score within the
UNOS region at which patients are likely to receive a
transplant may also be used in determining whether a
TIPS should be carried out before liver transplantation.
For instance, if a patient with a MELD score of 28 and
refractory ascites is in a UNOS region where organs are
available to patients whose MELD score is 28-30, contin-
ued conservative measures rather than TIPS might be
recommended. In general, a MELD score of ⬎24 is asso-
ciated with an increased risk of 3-month post TIPS mor-
tality,52,53 and consequently, TIPS should be avoided in
such patients unless they are candidates for liver trans-
plantation.
The MELD score has been compared to both the
Emory score,35 as well as the CTP score for prediction of
long-term survival in patients undergoing TIPS. The “c”
statistic for the MELD score was superior to the Emory
score but only slightly superior to or no better than the
CTP classification in predicting post procedure mortali-
ty.38
Hepatocellular Carcinoma. In our experience, he-
patic resection for HCC can be carried out safely in pa-
tients with cirrhosis and MELD score ⱕ8. Neither minor
hepatic resections (ⱕ3 segment resection) nor major re-
section (ⱖ4 segment) were associated with any mortality
30 days postoperatively if MELD score was ⱕ8. More-
over, patients with HCC smaller than 5 cm in diameter
and MELD score ⱕ8, had a 5-year survival of 80%. In
patients undergoing ablation therapies for unresectable
HCC, patients with MELD score ⱕ10 and CLIP score
ⱕ2 had the best outcome. In patients with MELD score
ⱖ10 and CLIP ⱖ2 the outcome was poor; therefore, local
KAMATH AND KIM 801
therapies for HCC should probably be considered only in
patients with MELD score ⱕ10 and CLIP score ⱕ2.54
Selection of Patients for Surgery Other Than Liver
Transplantation. Traditionally, the CTP score has
been used to determine risk of postoperative mortality.
Additional risk factors for mortality have included serum
creatinine concentration, the American Society of Anes-
thesiologists (ASA) physical status class, and cardiopul-
monary comorbidity.55 However, these variables have not
been put together to create a model to quantitate the risk
of postoperative mortality.
We have demonstrated that patients with cirrhosis are
at low risk of mortality after hepatic resection for hepato-
cellular carcinoma (HCC) if their MELD score is 8 or
less.56 This MELD cutoff of 8 for carrying out hepatic
resection for HCC has been confirmed by a study from
Italy.57 The utility of MELD in determining postopera-
tive mortality has also been confirmed in patients under-
going cardiac surgery,58 as well as in abdominal
operations including cholecystectomy,59-61 However,
MELD was an inaccurate predictor of mortality in pa-
tients without cirrhosis undergoing liver resection.62 In
our experience, MELD, the ASA physical status, and age
can be used to determine mortality following surgery in-
dependent of the procedure performed. We have demon-
strated a close relationship between MELD score and
mortality, with the relationship persisting both short-
term and long-term following surgery, irrespective of the
type of surgery being performed. Emergency surgery too
was not an independent predictor of mortality indepen-
dent of the MELD score.63 The MELD score, ASA phys-
ical status, and age may be used in determining whether
elective surgical procedures should be carried out before
or following liver transplantation.
Strengths and Limitations of MELD
In many ways, MELD is an ideal survival model in
comparison to either models/scores used in patients with
liver disease (Table 1). Its strengths derive from the robust
statistical foundation in its development and the large
number and variety of samples in which it was validated.
The model is based on only objective variables that are
readily obtained. Inclusion of creatinine incorporates a
measure of renal function, a well-recognized predictor of
survival in patients with liver disease.64-66 Whereas the
usefulness of CTP has been appreciated by clinicians for
many decades, it did not have much statistical basis in its
development, nor did it undergo as rigorous validation as
MELD. It also includes subjective variables such as ascites
and encephalopathy. In our initial validation study, the
c-statistic associated with the CTP score in the prediction
of 3-month survival was 0.84 (95% CI 0.78-0.90), in
802 KAMATH AND KIM
Table 1. Liver Disease: Features of Current Definitions/Scores
Objective Subjective
Condition Parameters Parameters
An Ideal Model ✓ –
FHF ✓ ✓
SBP ✓ –
HRS ✓ –
Child-Pugh (“Decompensated Liver Disease”) ✓ ✓
MELD ✓ –
comparison to 0.87 for MELD. Thus, the MELD scale is
thought to be at least as good as the CTP score in predict-
ing short-term mortality, while it may overcome many
limitations of the CTP score, at least for the purpose of
prioritization in donor organ allocation.
Several authors pointed out that MELD has not been
proven to be superior to the CTP score in patients listed
for liver transplantation or in a wider population of pa-
tients with cirrhosis.48,67 When the score designation with
regard to ascites and encephalopathy is done consistently
from one patient to the next by an experienced observer,
the CTP score is probably as accurate and reproducible as
MELD.2 However, one of the drawbacks of the CTP
system is that it is much more subject to variability and
interpretation than MELD. This was one of the factors
that made MELD more attractive as a standard for organ
allocation in that it minimizes the possibility of “gaming”
the system. Another advantage of MELD over the CTP
score is that it has a much wider range of possible scores
and has a better precision with which to distinguish pa-
tients according to their mortality risk. Finally, even the
most vocal skeptics of MELD agree that the serum creat-
inine is an important contribution of MELD in highlight-
ing the importance of renal function in the assessment of
mortality risk in patients with ESLD. A model developed
by adding serum creatinine to the CTP score has not been
proven to be more accurate than the MELD score.68
There are some cautions to be exercised in applying
MELD in individual patients. First, one must remember
that MELD was created and validated in a cohort of pa-
tients who were screened carefully with certain criteria,
which included absence of acute, reversible complica-
tions, such as bacterial infection or azotemia associated
with dehydration. In deriving the TIPS model, we used
the prothrombin time and serum creatinine and bilirubin
data recorded at the time when reversible factors had been
excluded. This approach was taken because we were pri-
marily interested in a measure most accurately reflective
of the underlying liver function. Therefore, in patients on
the waiting list for liver transplantation, the MELD score
should, in principle, be calculated only after acute revers-
ible processes are adequately treated.
HEPATOLOGY, March 2007
Validated
Parameters Prospectively
Readily Available Designed Internal External
✓ ✓ ✓ ✓
✓ – – –
✓ – – –
✓ – – –
✓ – – ✓
✓ (?)✓ ✓ ✓
Second, the primary role that was asked of MELD was
to rank patients according to mortality risk in a relatively
homogenous population of registrants on the liver trans-
plant waiting list. Thus, depending on the population to
which it is applied, mortality seen in patients with a given
MELD score may not necessarily be the same. Similarly,
hospitalized patients with cirrhosis who were not candi-
dates for liver transplantation may have a higher mortality
than candidates for liver transplantation who are younger
and devoid of comorbidity. Thus, it is not possible to
provide a universally applicable survival prediction by
MELD.
Third, although the objectivity of the variables in-
cluded in MELD is far superior to previous models, the
variables used in the MELD score may be subject to some
variability depending on how they are measured. The se-
rum creatinine typically measured by a colorimetric alka-
line picric Jaffe method may be less accurate than when
the enzymatic method for measuring serum creatinine is
used. When the serum bilirubin is above 25 mg/dl, the
colorimetric method overestimates the serum creatinine.
Accordingly, in patients with serum bilirubin ⬎25 mg/dl
the enzymatic method for measuring serum creatinine is
recommended. A somewhat related question regarding
bilirubin is whether the direct fraction of the serum bili-
rubin is a more accurate predictor of survival than the
total serum bilirubin. In the absence of studies clarifying
this issue, the total serum bilirubin remains the preferred
index for expressing overall liver function.
The prothrombin time is also subject to variability.
The thromboplastins available worldwide have an Inter-
national Sensitivity Index range from 1-3, the lower num-
bers indicating a more sensitive thromboplastin. The
prothrombin time is more prolonged if a sensitive throm-
boplastin is used as compared with a less sensitive throm-
boplastin. The INR for prothrombin time was introduced
as a means of decreasing this variability when measuring
prothrombin times in patients on warfarin anticoagula-
tion. The accuracy of INR may be decreased as a measure
of the coagulation status in patients with liver disease,
since there are other abnormalities in the coagulation
pathway in those patients. However, its role in predicting
HEPATOLOGY, Vol. 45, No. 3, 2007
survival in patients with liver disease has been demon-
strated repeatedly, especially when used to calculate the
MELD score. It is possible that other methods of express-
ing prothrombin time such as the prothrombin index are
more accurate reflectors of liver function, but these meth-
ods have yet to be validated as being more accurate than
the INR, both as indices of liver function and of coagula-
tion status. In our opinion, of measures of the prothrom-
bin time, the INR is superior to others because of it wide
availability as well as the track record as a survival indica-
tor in patients with ESLD.
Finally, some debate continues with regard to patients
with intractable complications of portal hypertension
such as ascites and hepatic encephalopathy. In our initial
evaluation of these complications in conjunction with
MELD, it was clear that when the c-statistic was used as
the criterion to determine the degree of improvement in
the model associated with the addition of these complica-
tions, they had only minimal benefit to MELD. However,
as shown in the case with hyponatremia (see next section),
if these complications occur in a small proportion of pa-
tients, addition of them would not increase the model
c-statistic materially, because it changes the ranking of
only a few patients, even when they may have substantial
impact in those few patients. To date, however, we are not
aware of data to clearly demonstrate that such is the case,
except in the case of hyponatremia. If data strongly sup-
portive of these and other complications add to the prog-
nostic evaluation in patients with ESLD, the transplant
community may be faced with a difficult decision whether
to re-incorporate these potentially subjective elements
back to the organ allocation policy.
Further Refinement and Improvement of
MELD
Patients with an increasing MELD score have been
thought to have an increased risk of mortality, whereas
those with a decreasing MELD score have a lower risk of
mortality, even if their MELD scores are identical.69
Thus, it has been proposed that the change in MELD
score, that is ⌬MELD, may add prognostic information
to the MELD score.69 Intuitively, a patient whose MELD
is increasing rapidly is more likely to have a worse out-
come than those with stable MELD. We conducted a
study using the time-dependent analysis of the effect of
the current MELD score and ⌬MELD (defined as the
difference between current MELD and the lowest MELD
score measured within 30 days prior to current MELD).
Although all of these variables were significant in the uni-
variate phase, ⌬MELD was no longer significant in the
multivariable analysis, especially when acute increases in
MELD in the last few days of life were excluded. This
KAMATH AND KIM 803
analysis highlighted that the current MELD is the most
important predictor of survival, regardless how that
MELD was reached.12
MELD could potentially be improved with more ac-
curate indices of liver function and perhaps better ways of
assessing renal function. Recently, several studies have
shown that the addition of serum sodium can improve the
predictive accuracy of the MELD score.70-73 Our study,
which is based on a multicenter database, shows that there
is a linear relationship between serum sodium and mor-
tality after adjusting for MELD. In addition, serum so-
dium may be particularly relevant in patients with a low
MELD score, e.g., MELD ⬍ 20.74 As was alluded to
earlier, because of the small proportion affected by hypo-
natremia, the c-statistics of the model did not change
substantially. However, with severe hyponatremia, the
risk in mortality increased as much as what would be
equivalent to an increase of more than 20 points in
MELD. Whereas the impact of serum sodium is quite
large, it remains uncertain whether the addition of serum
sodium to the MELD score can be used to determine
allocation of organs for liver transplantation, especially
because of the possible poor post liver transplantation
outcome of patients with low serum sodium.75
In conclusion, based on its ability to rank patients with
cirrhosis according to their short term mortality, MELD
has been recognized as a major contribution to the daily
practice of hepatology. Successful implementation of
MELD-based liver allocation in the United States has
been followed by widespread adoption of the system glo-
bally, attesting to its validity. In addition to organ alloca-
tion, emerging data support MELD as a useful clinical
tool in a wide spectrum of disease severity and variety.
These achievements notwithstanding, MELD is by no
means a perfect system. Users of MELD must be aware of
several features and limitations in its application. In the
meantime, efforts for further refinement and validation
must continue.
References
1. Malinchoc M, Kamath PS, Gordon FD, Peine CJ, Rank J, Borg PC. A
model to predict poor survival in patients undergoing transjugular intra-
hepatic portosystemic shunts. HEPATOLOGY 2000;31:864-871.
2. Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM,
Kosberg CL, et al. HEPATOLOGY 2001; 33:464-470.
3. Wiesner RH, Kamath PS, Malinchoc M, Kremers WK, Krom RAF, Kim
WR. MELD and PELD: Application of survival models to liver allocation.
Liver Transpl 2001;7:567-580.
4. Wiesner R, Edwards E, Freeman R, Harper A, Kim R, Kamath P, et al.
Model for end-stage liver disease (MELD) and allocation of donor livers.
Gastroenterology 2003;124:91-96.
5. Brandsater B, Broome U, Isoniemi H, Friman S, Hansen B, Schrumpf E,
et al. Liver transplantation for primary sclerosing cholangitis in the Nordic
countries: Outcome after acceptance to the waiting list. Liver Transpl
2003;9:961-969.
804 KAMATH AND KIM
6. Said A, Williams J, Holden J, Remington P, Gangnon R, Musat A, et al.
Model for end-stage liver disease score predicts mortality across a broad
spectrum of liver disease. J Hepatol 2004;40:897-903.
7. Botta F, Giannini E, Romagnoli P, Fasoli A, Malfatti F, Chiarbonello B, et
al. MELD scoring system is useful for predicting prognosis in patients with
liver cirrhosis and is correlated with residual liver function: a European
study. Gut 2003;52:134-139.
8. Freeman R. MELD: the holy grail of organ allocation? J Hepatol 2005;42:
16-20.
9. Biggins SW, Bambha K. MELD-based liver allocation: Who is under-
served? Semin Liver Dis 2006;26:211-220.
10. Wiesner RH, Edwards E, Freeman R, Harper A, Kim WR, Kamath PS, et
al. Model for End-Stage Liver Disease (MELD) and Allocation of Donor
Livers. Gastroenterology 2003;124:91-96.
11. Brown RS Jr, Lake JR. The survival impact of liver transplantation in the
MELD era, and the future for organ allocation and distribution [Com-
ment]. Am J Transplant 2005;5:203-204.
12. Bambha K, Kim WR, Kremers WK, Therneau TM, Kamath PS, Wiesner
R, et al. Predicting survival among patients listed for liver transplantation:
an assessment of serial MELD measurements. Am J Transplant 2004;4:
1798-1804.
13. Wiesner R, Lake J, Freeman R, Gish RG. Model for end-stage liver disease
(MELD) exception guidelines. Liver Transpl 2006;12(Suppl):S85-S87.
14. Ravaioli M, Grazi GL, Ballardini G, Cavrini G, Ercolani G, Cascon M,
Zanello M. Liver transplantation with the Meld system: a prospective
study from a single European center.Am J Transplant 2006;6:1572-1577.
15. Fink MA, Angus PW, Gow PJ, Berry SR, Wang BZ, Muralidharan V, et al.
Liver transplant recipient selection: MELD vs. clinical judgment. Liver
Transpl 2005;11:621-626.
16. Habib S, Berk B, Chang CCH, Demetris AJ, Fontes P, Dvorchik I,
Eghtesad B, et al. MELD and predictionof post-liver transplantation sur-
vival. Liver Transpl 2006;12:440-447.
17. Yoo HY, Thuluvath PJ. Short-term post-liver transplant survival after the
introduction of MELD scores for organ allocation in the United States.
Liver Int 2005;25:536-541.
18. Onaca N, Levy MF, Netto GJ, Thomas MJ, Sanchez EQ, Chinnakotla S,
et al. Pretransplant MELD score as a predictor of outcome after liver
transplantation for chronic hepatitis C. Am J Transplant 2003;3:626-630.
19. Saab S, Ly D, Nieto J, Kanwal F, Lu D, Raman S, et al. Hepatocellular
carcinoma screening in patients waiting for liver transplantation: a decision
analytic model. Liver Transpl 2003;9:672-681.
20. Akyildiz M, Karasu Z, Arikan C, Kilic M, Zeytunlu M, Gunsar F, et al.
Impact of pretransplant MELD score on post-transplant outcome in living
donor liver transplantation. Transplant Proc 2004;36:1442-1444.
21. Briceno J, Padillo J, Rufian S, Solorzano G, Pera C. Assignment of steatotic
livers by the Mayo model for End-stage liver disease. Transpl Int 2005;18:
577-583.
22. Burra P, Porte RJ. Should donors and recipients be matched in liver trans-
plantation? J Hepatol 2006;45:488-494.
23. Merion RM, Schaubel DE, Dykstra DM, Freeman R, Port FK, Wolfe RA.
The survival benefit of liver transplantation. Am J Transplant 2005;5:307-
313.
24. Nourmand H, Levin P, Enriquez R, Hiatt JR, Ghobrial RM, Former DG,
et al. Preoperative characteristics and intra-operative transfusion and vaso-
pressor requirements in patients with low vs. high MELD scores. Liver
Transpl 2006;12:614-620.
25. Washburn WK, Pollock BH, Nichols L, Speeg KV, Halff G. Impact of
recipient MELD score on resource utilization. Am J Transplant 2006;6:
2449-2454.
26. Bambha K, Kim WR, Benson J, Therneau T, Wiesner R, Kremers W, et al.
Economic impact of MELD-based allocation of liver transplants [Ab-
stract]. Am J Transplant 2005;5(Suppl 11):425.
27. Onaca N, Levy MF, Ueno T, Martin AP, Sanchez EQ, Chinnakotla S, et
al. An outcome comparison between primary liver transplantation and
retransplantation based on the pretransplant MELD score. Transplant Int
2006;19:282-287.
HEPATOLOGY, March 2007
28. Gonwa TA, McBride MA, Anderson K, Mai ML, Wadei H, Ahsan N.
Continued influence of preoperative renal function on outcome of ortho-
topic liver transplant in the US: Where will MELD lead us ? Am J Trans-
plant 2006;6:2651-2659.
29. Sanchez EQ, Gonwa TA, Levy MF, Goldstein RM, Mai ML, Hays SR, et
al. Preoperative and perioperative predictors of the need for renal replace-
ment therapy after orthotopic liver transplantation. Transplantation 2004;
78:1048-1054.
30. Machicao VI, Srinivas TR, Hemming AW, Soldevilla-Pico C, Firpi RJ,
Reed AI, et al. Impact of implementation of the MELD scoring system on
the prevalence and incidence of chronic renal disease following liver trans-
plantation. Liver Transpl 2006;12:754-761.
31. Wiesner RH, Freeman RB, Mulligan DC. Liver transplantation for hepa-
tocellular cancer: the impact of the MELD allocation policy. Gastroenter-
ology 2004;127(Suppl 1):S261-S267.
32. D’Amico G, Garcia-Tsao G, Pagliaro L. Natural history and prognostic
indicators of survival in cirrhosis: A systematic review of 118 studies.
J Hepatol 2006;44:217-231.
33. Chan HLY, Chim AML, Lau JTF, Hui AY, Wong VWS, Sung JJY. Eval-
uation of model for end-stage liver disease for prediction of mortality in
decompensated chronic hepatitis B. Am J Gastroenterol 2006;101:1516-
1523.
34. Sanyal AJ, Banas C, Sargeant C, Luketic VA, Sterling RK, Stravitz RT, et
al. Similarities and differences in outcomes of cirrhosis due to nonalcoholic
steatohepatitis and hepatitis C. HEPATOLOGY 2006;43:682-689.
35. Chalasani N, Kahi C, Francois F, Pinto AO, Marathe A, Bini EJ, et al.
Model for End-stage Liver Disease (MELD for predicting mortality in
patients with acute variceal bleeding. HEPATOLOGY 2002;35:1282-1284.
36. Amitrano L, Guardascione MA, Bennato R, Manguso F, Balzano A.
MELD score and hepatocellular carcinoma identify patients at different
risk of short-term mortality among cirrhotics bleeding from esophageal
varices [See comment]. J Hepatol 2005;42:820-825.
37. Alessandria C, Ozdogan O, Guevara M, Restuccia T, Jimenez W, Arroyo
V, et al. MELD score and clinical type predict prognosis in hepatorenal
syndrome: relevance to liver transplantation. HEPATOLOGY 2005;41:
1282-1289.
38. Schepke M, Appenrodt B, Heller J, Zielinski J, Sauerbruch T. Prognostic
factors for patients with cirrhosis and kidney dysfunction in the era of
MELD: results of a prospective study. Liver Int 2006;26:834-839.
39. Evans LT, Kim WR, Poterucha JJ, Kamath PS. Spontaneous bacterial
peritonitis in asymptomatic outpatients with cirrhotic ascites. HEPATOL-
OGY 2003;37:897-901.
40. Kremers W, van IJperen M, Kim WR, Freeman R, Harper AM, Kamath
PS, et al. MELD score as a predictor of pretransplant and post-transplant
survival in OPTN/UNOS Status 1 patients. HEPATOLOGY 2004;39:764-
769.
41. Schmidt LE, Larsen FS. MELD score as a predictor of liver failure and
death in patients with acetaminophen-induced liver failure. HEPATOLOGY.
In press.
42. Taylor RM, Davern T, Munoz S, Han S-H, McGuire B, Larson AM, et al.
Acute Liver Failure study Group. Fulminant hepatitis A virus infection in
the United States: incidence, prognosis an outcomes. HEPATOLOGY 2006;
44:1589-1597.
43. Maddrey W, Boitnott J, Bedine M, Weber FJ, Mezey E, White RJ. Cor-
ticosteroid therapy of alcoholic hepatitis. Gastroenterology 1978;75:193-
199.
44. Srikureja W, Kyulo NL, Runyon BA, Hu KQ. MELD score is a better
prognostic model than Child-Turcotte-Pugh score or Discriminant Func-
tion score in patients with alcoholic hepatitis. J Hepatol 2005;42:700-706.
45. Dunn W, Jamil LH, Brown LS, Wiesner RH, Kim WR, Menon KVN, et
al. MELD accurately predicts mortality in patients with alcoholic hepatitis.
HEPATOLOGY 2005;41:353-358.
46. Sheth M, Riggs M, Patel T. Utility of the Mayo End-Stage Liver Disease
(MELD) score in assessing prognosis of patients with alcoholic hepatitis.
BMC Gastroenterol 2001;2:2. Epub 2002 Jan 22.
HEPATOLOGY, Vol. 45, No. 3, 2007
47. Wehler MK, J., Reulbach U, Hahn EG, Strauss R. Short-term prognosis in
critically ill patients with cirrhosis assessed by prognostic scoring systems.
HEPATOLOGY 2001;34:255-261.
48. Cholongitas E, Senzolo M, Patch D, Kwong K, Nikolopoulou V, Leandro
G, et al. Risk factors, sequential organ failure assessment and model for
end-stage liver disease scores for predicting short-term mortality in cir-
rhotic patients admitted to intensive care unit. Aliment Pharmacol Ther
2006;23:883-893.
49. Du WB, Li LJ, Huang JR, Yang Q, Liu XL, Li J, et al. Effects of artificial
liver support system on patients with acute or chronic liver failure. Trans-
plant Proc 2005;37:4359-4364.
50. Boyer TD, Haskal ZJ. American Association for the Study of Liver Dis-
eases. The role of transjugular intrahepatic portosystemic shunt in the
management of portal hypertension. HEPATOLOGY 2005;41:386-400.
51. Salerno F, Merli M, Riggio O, Cazzaniga M, Valeriano V, Pozzi M, et al.
Randomized controlled study of TIPS versus paracentesis plus albumin in
cirrhosis with severe ascites. HEPATOLOGY 2004;40:629-635.
52. Montgomery A, Ferral H, Vasan R, Postoak DW. MELD score as a pre-
dictor of early death in patients undergoing elective transjugular intrahe-
patic portosystemic shunt (TIPS) procedures. Cardiovascular and
Interventional Radiology 2005;28:307-312.
53. Ferral H, Vasan R, Speeg KV, Serna S, Young CR, Postoak DW, et al.
Evaluation of a model to predict poor survival in patients undergoing
elective TIPS procedures. J Vascular and Interventional Radiol 2002;13:
1103-1108.
54. Testa R, Testa E, Giannini EG, Botta F, Malfatti F, Chiarbonello B, et al.
Trans-catheter arterial chemoembolization for hepatocellular carcinoma in
patients with viral cirrhosis: role of combined staging systems, Cancer Liver
Italian Program (CLIP) and Model for End-stage Liver Disease (MELD),
in predicting outcome after treatment. Aliment Pharmacol Ther 2003;17:
1563-1569.
55. Ziser A, Plevak DJ, Wiesner RH, Rakela J, Offord KP, Brown DL. Mor-
bidity and mortality in cirrhotic patients undergoing anesthesia and sur-
gery. Anesthesiology 1999;90:40-53.
56. Teh SH, Christein J, Donohue J, Que F, Kendrick M, Farnell M, et al.
Hepatic resection of hepatocellular carcinoma in patients with cirrhosis:
Model of end-stage liver disease (MELD) score predicts perioperative mor-
tality. J Gastrointest Surg 2005;9:1207-1215.
57. Cucchetti A, Ercolani G, Vivarelli M, Cescon M, Ravaioli M, La Barba G,
et al. Impact of model for end-stage liver disease (MELD) score on prog-
nosis after hepatectomy for hepatocellular carcinoma on cirrhosis. Liver
Transpl 2006;12:966-971.
58. Seaman A, Barnes DS, Zein NN, Levinthal GN, Connor JT, Carey WD.
Predicting Outcome After Cardiac Surgery in Patients with Cirrhosis: A
Comparison of Child-Pugh and MELD Score. Clin Gastroenterol Hepatol
2004;2:719-723.
59. Befeler AS, Palmer DE, Hoffman M, Longo W, Solomon H, Di Bisceglie
AM. The safety of intra-abdominal surgery in patients with cirrhosis. Arch
Surg 2005;140:650-654.
60. Northup PG, Wanamaker RC, Lee VD, Adams RB, Berg CL. Model for
end-stage liver disease (MELD) predicts non-transplant surgical mortality
in patients with cirrhosis. Ann Surg 2005;242:244-251.
KAMATH AND KIM 805
61. Perkins L, Jeffries M, Patel T. Utility of preoperative scores for predicting
morbidity after cholecystectomy in patients with cirrhosis. Clin Gastroen-
terol Hepatol 2004;2:1123-1128.
62. Schroeder RA, Marroquin CE, Bute BP, Khuri S, Henderson WJ, Kwo
PC. Predictive indices of morbidity and mortality after liver resection. Ann
Surg 2006;243:373-379.
63. Teh SH, Nagorney DM, Stevens SR, Offord KP, Therneau TM, Plevak
DJ, et al. Risk factors for mortality after surgery in patients with cirrhosis.
Gastroenterology 2007. In press.
64. Gines A, Escorsell A, Gines P, Salo J, Jimenez W, Inglada L, et al. Inci-
dence, predictive factors, and prognosis of the hepatorenal syndrome in
cirrhosis with ascites. Gastroenterology 1993;105:229-236.
65. Arroyo V, Gines P, Gerbes A, Dudley F, Gentilini P, Laffi G, et al. Defi-
nition and diagnostic criteria of refractory ascites and hepatorenal syn-
drome in cirrhosis. HEPATOLOGY 1996;23:164-176.
66. del Olmo J, Andrez P, Serra M, Wassel A, Benages A, Rodrigo J. Predictors
of morbidity and mortality after the first episode of upper gastrointestinal
bleeding in liver cirrhosis. J Hepatol 2000;32:19-24.
67. Cholongitas E, Papatheodoridis GV, Vangeli M, Terreni N, Patch D,
Burroughs AK. Systematic review: the model for end-stage liver disease —
should it replace Child-Pugh’s classification for assessing prognosis in cir-
rhosis? Aliment Pharmacol Ther 2005;22:1079-1089.
68. Giannini EG, Botta F, Fumagalli A, Malfatti F, Testa E, Chiarbonello B, et
al. Can inclusion of serum creatinine values improve the Child-Turcotte-
Pugh score and challenge the prognostic yield of the model for end-stage
liver disease score in the short-term prognostic assessment of cirrhotic
patients? Liver Int 2004;24:465-470.
69. Merion RM, Wolfe RA, Dykstra DM, Leichtman AB, Gillespie B, Held
PJ. Longitudinal assessment of mortality risk among candidates for liver
transplantation [See comment]. Liver Transpl 2003;9:12-18.
70. Biggins SW, Rodriguez HJ, Bacchetti P, Bass NM, Roberts JP, Terrault
NA. Serum sodium predicts mortality in patients listed for liver transplan-
tation. HEPATOLOGY 2005;41:32-39.
71. Heuman D, Abou-assi SG, Habib A, Williams LM, Stravitz RT, Sanyal AJ,
et al. Persistent ascites and low serum sodium identify patients with cir-
rhosis and low MELD scores who are at high risk for early death. HEPA-
TOLOGY 2004;40:802-810.
72. Biggins SW, Kim WR, Terrault NA, Saab S, Balan V, Schiano T, et al.
Evidence-based incorporation of serum sodium concentration into
MELD. Gastroenterology 2006; 130:1652-1660.
73. Ruf AE, Kremers W, Chavez LL, Descalzi VI, Podesta LG, Villamil FG.
Addition of serum sodium into the MELD score predicts waiting list
mortality better than MELD alone. Liver Transpl 2005;11:336-343.
74. Heuman DM, Abou-Assi SG, Habib A, Williams LM, Stravitz RT, Sanyal
AJ, et al. Persistent ascites and low serum sodium identify patients with
cirrhosis and low MELD scores who are at high risk for early death. HEPA-
TOLOGY 2004;40:802-810.
75. Londono M, Guevara M, Rimola A, Navasa M, Taura P, Mas A, et al.
Hyponatremia impairs early post transplantation outcome in patients with
cirrhosis undergoing liver transplantation. Gastroenterology 2006;130:
1135-1143.

a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Kartoun U, Corey KE, Simon TG, Zheng H,
Aggarwal R, Ng K, et al. (2017) The MELD-Plus: A
generalizable prediction risk score in cirrhosis.
PLoS ONE 12(10): e0186301. https://doi.org/
10.1371/journal.pone.0186301
Editor: Sheng-Nan Lu, Chang Gung Memorial
Hospital Kaohsiung Branch, TAIWAN
Received: July 13, 2017
Accepted: September 28, 2017
Published: October 25, 2017
Copyright: © 2017 Kartoun et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The institutional
review board of Partners HealthCare and IBM
approved this study and all its methods, including
the EMR cohort assembly, data extraction, and
analyses. Data contain potentially identifying
information and may not be shared publicly.
Deidentified data may be requested from The
Partners Human Research Committee, the
Institutional Review Board of Partners HealthCare
(Address: 399 Revolution Drive, Suite # 710,
Somerville MA, 02145, USA, Telephone: 857-282-
1900; ecor@partners.org).
PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017
RESEARCH ARTICLE
The MELD-Plus: A generalizable prediction risk
score in cirrhosis
Uri Kartoun1,2¤, Kathleen E. Corey1,3, Tracey G. Simon1,3, Hui Zheng4, Rahul Aggarwal1,2,
Kenney Ng5, Stanley Y. Shaw1,2*
1 Harvard Medical School, Boston, Massachusetts, United States of America, 2 Center for Systems Biology;
Center for Assessment Technology & Continuous Health (CATCH), Massachusetts General Hospital, Boston,
Massachusetts, United States of America, 3 Gastrointestinal Unit, Massachusetts General Hospital, Boston,
Massachusetts, United States of America, 4 Center for Biostatistics, Massachusetts General Hospital,
Boston, Massachusetts, United States of America, 5 IBM Research, Cambridge, Massachusetts, United
States of America
¤ Current address: IBM Research, Cambridge, Massachusetts, United States of America
* stanley_shaw@hms.harvard.edu
Abstract
Background and aims
Accurate assessment of the risk of mortality following a cirrhosis-related admission can
enable health-care providers to identify high-risk patients and modify treatment plans to
decrease the risk of mortality.
Methods
We developed a post-discharge mortality prediction model for patients with a cirrhosis-
related admission using a population of 314,292 patients who received care either at Massa-
chusetts General Hospital (MGH) or Brigham and Women’s Hospital (BWH) between 1992
and 2010. We extracted 68 variables from the electronic medical records (EMRs), including
demographics, laboratory values, diagnosis codes, and medications. We then used a regu-
larized logistic regression to select the most informative variables and created a risk score
that comprises the selected variables. To evaluate the potential for generalizability of our
score, we applied it on all cirrhosis-related admissions between 2010 and 2015 at an inde-
pendent EMR data source of more than 18 million patients, pooled from different health-
care systems with EMRs. We calculated the areas under the receiver operating characteris-
tic curves (AUROCs) to assess prediction performance.
Results
We identified 4,781 cirrhosis-related admissions at MGH/BWH hospitals, of which 778
resulted in death within 90 days of discharge. Nine variables were the most effective predic-
tors for 90-day mortality, and these included all MELD-Na’s components, as well as albumin,
total cholesterol, white blood cell count, age, and length of stay. Applying our nine-variable
risk score (denoted as “MELD-Plus”) resulted in an improvement over MELD and MELD-Na
scores in several prediction models. On the MGH/BWH 90-day model, MELD-Plus
1 / 15

Funding: The study was funded in part by grants
from the NIH K23 DK099422 (KEC), and NIH U54
LM008748 (SYS). NIH did not play a role in the
study design, data collection and analysis, decision
to publish, or preparation of the manuscript and
only provided financial support in the form of
authors’ salaries and/or research materials. NIH
provided support in the form of salaries for authors
UK, KEC, and SYS. IBM neither provided authors
[UK, KN] salaries related to the study nor played
any role in the study design, data collection and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist. UK and KN
confirm that the commercial affiliation with IBM
does not alter their adherence to all PLOS ONE
policies on sharing data and materials.
Abbreviations: AUROC, Area under the receiver
operating characteristic curve; CI, Confidence
interval; COPD, Chronic obstructive pulmonary
disease; CPT, Current procedural terminology;
EMR, Electronic medical record; GGT, Gamma
glutamyl transpeptidase; HE, Hepatic
encephalopathy; HIPAA, Health Insurance
Portability and Accountability Act; OR, Odds ratio;
ICD-9-CM, International Classification of Diseases,
Ninth Revision, Clinical Modification; LASSO, Least
absolute shrinkage and selection operator; MELD,
Model for end-stage liver disease; NAFLD,
Nonalcoholic fatty liver disease; Na, Sodium’s
chemical element symbo; NS, Not significant; SBP,
Spontaneous bacterial peritonitis; SE, Standard
error; STD, Standard deviation; WBC, White blood
cell count; eGFR, Estimated glomerular filtration
rate.
PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017
The MELD-Plus
improved the performance of MELD-Na by 11.4% (0.78 [95% CI, 0.75–0.81] versus 0.70
[95% CI, 0.66–0.73]). In the MGH/BWH approximate 1-year model, MELD-Plus improved
the performance of MELD-Na by 8.3% (0.78 [95% CI, 0.76–0.79] versus 0.72 [95% CI,
0.71–0.73]). Performance improvement was similar when the novel MELD-Plus risk score
was applied to an independent database; when considering 24,042 cirrhosis-related admis-
sions, MELD-Plus improved the performance of MELD-Na by 16.9% (0.69 [95% CI, 0.69–
0.70] versus 0.59 [95% CI, 0.58–0.60]).
Conclusions
We developed a new risk score, MELD-Plus that accurately stratifies the short-term mortal-
ity of patients with established cirrhosis, following a hospital admission. Our findings demon-
strate that using a small set of easily accessible structured variables can help identify novel
predictors of outcomes in cirrhosis patients and improve the performance of widely used tra-
ditional risk scores.
Introduction
Cirrhosis-related complications account for 1.1% and 1.8% of all deaths in the United States
and Europe, respectively [1, 2]. In addition to increased mortality, individuals with cirrhosis
suffer from significantly worse health issues and greater disability compared to those without
cirrhosis [3].
Although risk-stratification tools for the prediction of cirrhosis-related mortality are avail-
able [4–9], these models are based on small populations and use a limited number of prese-
lected traditional predictors. Improved mortality prediction scores may highlight the clinical
variables that contribute to mortality risk, including modifiable factors, and guide the alloca-
tion of resources to improve cirrhosis care for high-risk patients.
The recent availability of large cohorts of data from electronic medical records (EMRs)
allows for the development of improved mortality prediction scores through inclusion of a
broader set of clinically applicable, unbiased variables. Not only that, but developing such pre-
diction models allows clinicians to identify the clinical variables that contribute to mortality
risk, including modifiable factors. Improving current standard models like the model for end-
stage liver disease (MELD) and MELD-Na can guide clinicians in better targeting treatment to
improve cirrhosis care and outcomes for high-risk patients.
Cohorts assembled from EMRs represent a powerful resource to study disease complica-
tions at the population level. Recent studies have demonstrated the usefulness of EMR analysis
to discover or confirm outcome correlations, sub-categories of disease, and adverse drug
events [10–14]. The MELD score is based on three commonly used laboratory tests available in
the EMRs, and it is the most widely used tool to predict outcomes in patients with cirrhosis
[15, 16]. An extended version of MELD, one that incorporates serum sodium levels, the
MELD-Na score, has been recently adopted by The Liver and Intestine Transplantation Com-
mittee for liver transplant allocation [17]. Although the two scores are simple to calculate and
apply in a practical sense, the improved accessibility of a wide variety of variables from EMRs
raises the possibility that prediction models could benefit from the inclusion of a broader,
unbiased set of clinical variables. Identifying a combination of the most informative variables
may improve the prognostic utility beyond that of current risk scores.
2 / 15
 The MELD-Plus

The aim of the present study was to develop a risk score to predict mortality following a cir-
rhosis-related admission. We demonstrated that a score composed of a small set of easily
accessible clinical variables improves the prediction performance of both the MELD and
MELD-Na scores. We further demonstrated the generalizability of our model through inde-
pendent validation in a large EMR-based data source.

Methods
Study population
We analyzed a previously defined cohort of 314,292 patients at increased risk for metabolic
disease who were admitted to Massachusetts General Hospital (MGH) or Brigham and Wom-
en’s Hospital (BWH) between 1992 and 2010 [13]. We identified an admission as cirrhosis-
related when the keyword “cirrhosis” was present in the discharge summary of the admission
and we observed at least one ICD-9 code (571.2, 571.5, or 571.6 as in [18]) within the 30 days
preceding the discharge date, including during the admission. This identification method was
validated by a physician (Dr. Kathleen Corey) chart review.
We excluded elective admissions if they included at least one diagnosis or procedure code
for liver biopsy, radiofrequency ablation, transarterial chemoembolization, hepatic resection,
or liver transplant. We included only patients 18 years of age or older at the time of the admis-
sion, and we tracked the records of all patients for 90 days after their discharge. We deter-
mined mortality through linkage to the social security master death index.

Prediction modeling
To predict mortality within 90 days, we developed a model that included a large set of struc-
tured variables extracted from the EMRs. In addition to variables available during the period
of admission, we considered variables available for the period of 12 months preceding the dis-
charge date (see Table 1).
The variables included demographics (e.g., gender, ethnicity, marital status), laboratory
measurements (e.g., albumin, sodium), and medications (e.g., anticoagulants, lipid lowering
agents). For laboratory variables, we used the most recent values found during admission
(when no value was found during admission, we considered the preceding 12 months). Typi-
cally, common laboratory measurements were available during the admission (as seen in
Table 1). We determined comorbidities from the number of diagnosis codes within the 12
months prior to the discharge date, and we determined medication count by recording the
number of prescriptions within the 12 months preceding the discharge date.
Additional variables included body mass index, NAFLD fibrosis score (Eq 1), and the
MELD score (Eq 2). Missing values were imputed with the mean of the available data for each
variable. We randomly selected two thirds of the admissions to serve as a derivation set,
whereas the remaining one third served as a validation set. A complete list of the variables we
used is available in S1 Table, and all diagnoses and procedure definitions used in this study are
available in S2 Table.
NAFLD Fibrosis Score ¼ 1:675 þ 0:037  Age þ 0:094  BMI
þ1:13  IFG=Diabetes ðyes ¼ 1; no ¼ 0Þ ð1Þ
þ0:99  AST=ALT ratio 0:013  Platelet 0:66  Albumin

MELD Score ¼ þ6:43 þ 9:57  lnðCreatinineÞ

ð2Þ
þ3:78  lnðTotal BilirubinÞ þ 11:2  lnðINRÞ

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 3 / 15

 The MELD-Plus

Table 1. Baseline characteristics. All values extracted during the 12 months preceding discharge date. For
laboratory variables, values are the most recent. Comorbidity calculations count the number of diagnosis
codes. Prevalence calculations consider admissions with at least one measurement for laboratories and at
least one diagnosis code for comorbidities.
Variable and category Cirrhosis-related admissions (n = 4,781)
Age (years); Mean (SD) 60.0 (13.7)
Gender (%)
Male 64.4
Female 35.6
Ethnicity (%)
Caucasian 77.4
African American 6.8
Other 2.3
Unknown 13.5
Insurance Type (could be  1 types per patient) (%)
Medicaid 5.6
Medicare 60.0
Other 98.6
BMI (kg/m2); Mean (SD) 28.7 (8.2)
Laboratory values; Mean (SD) / Prevalence (%)
Sodium (mmol/l) 136.5 (5.8) / 99.7
eGFR (ml/min/1.73m2) 59.6 (33.7) / 31.9
WBC (th/cumm) 6.7 (3.8) / 99.8
Platelets (th/cumm) 141.3 (100.0) / 99.8
Prothrombin time (INR) 1.5 (0.5) / 91.3
Albumin (g/dl) 2.9 (0.7) / 98.7
Total Bilirubin (mg/dl) 2.5 (4.3) / 98.7
Transaminase SGOT (u/l) 60.2 (76.6) / 98.8
Transaminase SGPT (u/l) 36.1 (38.2) / 96.7
GGT (u/l) 249.4 (346.6) / 6.8
MELD score 14.21 (6.1) / 84.8
NAFLD Fibrosis score 1.70 (2.1) / 13.7
Comorbidities; Mean (SD) / Prevalence (%)
Variceal hemorrhage / Gastrointestinal bleed 0.8 (2.1) / 24.2
Spontaneous bacterial peritonitis 0.1 (0.6) / 3.4
Hepatocellular carcinoma 0.7 (5.5) / 4.3
Hepatorenal syndrome 0.1 (0.5) / 3.2
Hepatic encephalopathy 1.4 (3.6) / 28.6
Ascites 2.4 (5.7) / 37.9
Renal failure 1.8 (8.4) / 13.1
Cerebrovascular disease 0.5 (2.4) / 10.8
Congestive heart failure 2.7 (7.9) / 36.7
Hypertension 3.1 (7.2) / 58.9
Acute myocardial infarction 0.5 (2.1) / 13.4
Ischemic heart disease 0.3 (1.2) / 12.4
Peripheral vascular disease 0.5 (2.2) / 13.0
Diabetes 5.9 (10.3) / 57.0
https://doi.org/10.1371/journal.pone.0186301.t001

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 4 / 15

 The MELD-Plus

To select the most informative variables, we applied feature selection on the derivation set.
We used logistic regression with the adaptive least absolute shrinkage and selection operator
(LASSO) algorithm [19] because it is considered an efficient algorithm for parsimoniously
ranking variables in clinical predictive modeling [20, 21]. We considered all variables that
were statistically significantly different from a univariate analysis (P < 0.05) as in [22]. The
generalized linear model (GLM) equations used to calculate prediction risk at MGH/BWH
and at the independent EMR source are presented in Eqs 3 and 4.
L¼
þ11:794383
þ2:076192  Log10ð1 þ Total BilirubinÞ
þ2:494291  Log10ð1 þ CreatinineÞ
6:049540  Log10ð1 þ AlbuminÞ
þ2:525904  Log10ð1 þ INRÞ
ð3Þ
þ1:911856  Log10ð1 þ WBCÞ
þ0:015411  Length of stay
ðnumber of nights the patient spent in the hospital during the cirrhosis related admissionÞ
þ0:041047  Age ðyearsÞ
6:625270  Log10ð1 þ SodiumÞ
1:445666  Log10ð1 þ Total CholesterolÞ

expðLÞ
MELD Plus ¼ pð90 day mortalityÞ ¼ ð4Þ
1 þ expðLÞ

To calculate 95% confidence intervals, we applied the bootstrap procedure with 1,000
replicates. We calculated the area under the receiver operating characteristic curves
(AUROC) to measure the model’s accuracy in the validation set. Additionally, we evaluated
for overfitting by comparing the AUROC in the validation set to an average AUROC value
for 100 permutations of randomly selected derivation and validation sets (each including
two thirds and one third of the derivation set’s cirrhosis-related admissions, respectively).
We compared categorical variables using a chi-squared test, and we compared the differ-
ences in the means of continuous variables using a t-test or Wilcoxon rank sum test, as
appropriate. We further compared the differences in standard deviations by using an F-test.
All statistical tests were two-sided, with Bonferroni corrections for the 68 comparisons, and
the adjusted P value was 7.4
10−4 for each comparison. We performed all programming
using the R statistical language [23].

Independent validation
We were granted access to a data source of 18,345,793 individuals, pooled from multiple differ-
ent health-care systems with EMRs (“The IBM Explorys Network”) [24]. The data were stan-
dardized and normalized using common ontologies, searchable through a HIPAA-enabled,
de-identified cloud-computing platform. Patients were seen in multiple health-care systems
between 2010 and 2015, with a combination of data from clinical EMRs, outgoing health-care
system bills, and adjudicated payor claims.
We first identified all cirrhosis-related admissions in this database, and then we extracted
values for the selected variables of our MGH/BWH 90-day model. We further deployed the

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 5 / 15

 The MELD-Plus

GLM equations on the independent source (Eqs 3 and 4). Missing values were imputed based
on the mean values of the MGH/BWH 4,781 cirrhosis-related admissions: total bilirubin
(2.486604493), creatinine (1.375274633), albumin (2.888940902), INR (1.499619615), WBC
(6.673836966), sodium (136.5401552), and total cholesterol (133.6246914).
Because death dates were not available for patients in the IBM Explorys Network, we used
the year of death to determine the outcome. We were able then only to use a 1-year estimated
death for this population (e.g., for a patient discharged on October 13, 2010, we could only
determine if the patient died either in 2010 or 2011, or survived after that). To compare the
performance of the IBM Explorys approximate one-year prediction model, we used the origi-
nal 314,292-patient population at MGH/BWH and applied the same approximate one-year
mortality outcome identification method. We calculated AUROCs for MELD, MELD-Na (Eq
5), and for our risk score (Eqs 3 and 4).
MELD Na ¼ MELD þ 1:59  ð135 NaÞ ð5Þ

The institutional review board of Partners HealthCare and IBM approved this study and all
its methods, including the EMR cohort assembly, data extraction, and analyses.

Results
Univariate analysis
We identified a total of 4,781 admissions as cirrhosis-related, of which 778 resulted in death
within 90 days of the discharge date (16.3%). In a sample of 50 randomly selected patients,
64% were admitted primarily for cirrhosis, for instance, due to the presence of ascites or spon-
taneous bacterial peritonitis, and the rest had the comorbidity of cirrhosis but were admitted
primarily for different reasons such as heart failure or chronic obstructive pulmonary disease
(COPD). Individuals who died within the 90-day period after discharge were older in compari-
son with those who survived (64.1 years versus 59.2 years, P = 4.22
10−17); however, the two
populations did not differ by gender (65.0% male versus 64.0% female, P = 1.0) or ethnicity
(77.0% Caucasian for both, P = 1.0).
We calculated event ratios by dividing the values ascertained for the two populations (e.g.,
the mean MELD scores were 18.5 and 13.3 for admissions that resulted in death and survival,
respectively, yielding a ratio of 1.4, P = 4.45
10−65). Individuals who died within the 90-day
period after discharge had higher ratios of liver-related comorbidities than those who survived,
and these comorbidities included hepatorenal syndrome (ratio = 5.1, P = 4.60
10−21), hepatocel-
lular carcinoma (ratio = 5.0, P = 1.40
10−16), and ascites (ratio = 2.1, P = 3.55
10−24). Laboratory
measurements also significantly differentiated the two populations. For instance, albumin was
lower in those who died within the 90-day period (2.60 g/dl versus 2.95 g/dl, P = 5.64
10−35),
and the total bilirubin (4.87 mg/dl versus 2.02 mg/dl, P = 1.82
10−41), INR (1.70 versus 1.46,
P = 5.63
10−30), and creatinine (1.80 mg/dl versus 1.29 mg/dl, P = 1.24
10−36) were higher in
those who died within the 90-day period.
No difference was found in the prevalence of COPD, cerebrovascular disease, diabetes, cor-
onary artery disease, peripheral vascular disease, pneumonia, or sleep apnea between the popu-
lations. The complete list of variables, indicating the differences between the surviving and the
deceased populations, is presented in S3 Table.

Logistic regression model

The AUROCs of 0.78 were identical for all three models composed of multiple variables (Fig
1). With generalizability in mind and the potential ease of extraction of commonly available

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 6 / 15

 The MELD-Plus

Fig 1. AUROCs using differing variable combinations in a 90-day mortality prediction model at MGH/
BWH.
https://doi.org/10.1371/journal.pone.0186301.g001

laboratory values and other trivial variables (e.g., age, length of stay), we decided to follow the
model that comprised the 9 readily available clinical variables. To evaluate the contribution of
the MELD score to the 90-day mortality prediction, we evaluated the performance of MELD
and MELD-Na scores alone. Considering the 4,781 admissions, using the MELD score alone
to predict the 90-day mortality resulted in an AUROC value of 0.69. An additional model
using the MELD-Na score alone yielded an AUROC value of 0.70.
Each of the MELD-Na components were associated with an increased mortality, including
INR (OR, 1.58; 95% CI, 1.30–1.96), creatinine (OR, 1.25; 95% CI, 1.16–1.34), total bilirubin
(OR, 1.11; 95% CI, 1.08–1.14), and sodium (OR, 0.97; 95% CI, 0.95–0.99). Other laboratory
measurements associated with mortality included WBC (OR, 1.10; 95% CI, 1.07–1.13), total
cholesterol (OR, 0.996; 95% CI, 0.993–0.999), and albumin (OR, 0.45; 95% CI, 0.37–0.52).
Additional predictors included age at time of the admission (OR, 1.04; 95% CI, 1.03–1.05) and
length of stay (OR, 1.02; 95% CI, 1.005–1.03).
Because total cholesterol and hospital length of stay are typically not uniform factors
across different hospitals and may vary in different countries, we evaluated an additional
model that included only 7 of the 9 variables. This yielded an AUROC of 0.77 and
resulted in the following associations with increased mortality: INR (OR, 1.66; 95% CI,
1.38–2.05), creatinine (OR, 1.25; 95% CI, 1.17–1.35), total bilirubin (OR, 1.11; 95% CI,
1.08–1.14), sodium (OR, 0.97; 95% CI, 0.95–0.98), WBC (OR, 1.10; 95% CI, 1.07–1.13),
albumin (OR, 0.43; 95% CI, 0.36–0.51), and age (OR, 1.04; 95% CI, 1.03–1.05). We pres-
ent the GLM equations used to calculate prediction performance at MGH/BWH in Eqs 6

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 7 / 15

 The MELD-Plus

and 7.
L¼
þ8:53499496
þ2:06503238  Log10ð1 þ Total BilirubinÞ
þ2:59679650  Log10ð1 þ CreatinineÞ
6:34990436  Log10ð1 þ AlbuminÞ ð6Þ
þ2:99724802  Log10ð1 þ INRÞ
þ1:92811726  Log10ð1 þ WBCÞ
þ0:04070442  Age ðyearsÞ
6:47834101  Log10ð1 þ SodiumÞ

MELD Plus ðexcluding length of stay and total cholesterolÞ ¼

expðLÞ ð7Þ
pð90 day mortalityÞ ¼
1 þ expðLÞ

Prediction of 90-day mortality after a cirrhosis-related admission

Using our 9-variable risk score, we divided our population into quintiles and compared the
average predicted 90-day mortality with the observed mortality within each quintile. The pre-
dicted 90-day mortality derived from a logistic regression model for each admission and indi-
cated the probability that a patient who survived the admission would die within 90 days post
discharge. As shown in Fig 2A–2C, the predicted 90-day mortality was strongly correlated
with the observed mortality rate throughout the range of risk in both derivation and validation
sets (Kendall’s τ = 1.0; P = 0.027; Pearson correlation r = 0.995 for the correlation between the
average calculated and observed mortality). We provide the logistic regression equations used
to calculate the predicted 90-day mortality probabilities in Eqs 3 and 4. The complete list of
variables that indicate the differences between the highest-risk quantile and the lowest-risk
quantile populations are presented in S4 Table.

Generalization evaluation
Applying our 9-variable risk score (the MELD-Plus score) demonstrated an improvement
over MELD and MELD-Na scores in all prediction models, as shown in Fig 3. On the MGH/
BWH 90-day model, MELD-Plus improved the performance of MELD-Na by 11.4% (0.78
[95% CI, 0.75–0.81] versus 0.70 [95% CI, 0.66–0.73]). On the MGH/BWH approximate 1-year
model, MELD-Plus improved the performance of MELD-Na by 8.3% (0.78 [95% CI, 0.76–
0.79] versus 0.72 [95% CI, 0.71–0.73]). On the IBM Explorys Network model used for external
validation, MELD-Plus improved the performance of MELD-Na by 16.9% (0.69 [95% CI,
0.69–0.70] versus 0.59 [95% CI, 0.58–0.60]).
It is notable that the performance of MELD-Plus on the IBM Explorys data was lower in
comparison with both MGH/BWH models (0.69 versus 0.78). Consistent with MELD-Plus,
the performance of MELD and MELD-Na were also much lower on the IBM Explorys data in
comparison with MGH/BWH. A potential reason for this is that the IBM Explorys Network
population was relatively healthier. Patients in the IBM Explorys network had lower severity of
liver disease in comparison with the corresponding MGH/BWH 1-year prediction model

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 8 / 15

 The MELD-Plus

Fig 2. Predicted versus observed 90-day mortality within each risk quintile. (A) Entire cohort of 4,781
cirrhosis-related admissions. (B) Derivation set of 3,187 cirrhosis-related admissions. (C) Validation set of
1,594 cirrhosis-related admissions.
https://doi.org/10.1371/journal.pone.0186301.g002

(mean MELD: 9.4 versus 16.8; P < 0.0001, mean MELD-Na: 11.4 versus 18.1; P < 0.0001).
There may be other differences in the data or populations in the independent systems; the
Partners HealthCare Research Patient Data Registry collected the MGH/BWH data, whereas
dozens of distinct data aggregation mechanisms collected the data for the IBM Explorys Net-
work. Furthermore, the variability in the levels of prediction performance might be influenced
by the variability in the data; prediction performance might be higher when there is more vari-
ability in the data source (i.e., the population comprising patients with a broad spectrum of

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 9 / 15

 The MELD-Plus

Fig 3. Prediction performance across different cirrhosis populations. (A) MGH/BWH 90-day mortality
(4,781 cirrhosis-related admissions). (B) The IBM Explorys Network approximate 1-year mortality (24,042
cirrhosis-related admissions). (C) MGH/BWH approximate 1-year mortality (4,680 cirrhosis-related
admissions).
https://doi.org/10.1371/journal.pone.0186301.g003

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 10 / 15

 The MELD-Plus

levels of cirrhosis severity). In the other direction, when the data is more uniform (e.g., most
patients have just been diagnosed with cirrhosis for the first time, and only a minority suffers
from an advanced cirrhosis), then prediction accuracy is lower. This hypothesis was confirmed
because the IBM Explorys network had a statistically significant lower standard deviation of
severity of liver disease in comparison with the MGH/BWH 1-year population (STD MELD:
1.8 versus 8.2; P < 0.0001, STD MELD-Na: 3.6 versus 8.2; P < 0.0001).

Discussion
In this study, we used accessible EMR variables to develop a highly accurate, predictive model
of 90-day post-discharge mortality in individuals with cirrhosis. We identified 9 variables that
accurately predicted 90-day mortality with an AUROC of 0.78. Our risk score improved the
performance of MELD and MELD-Na scores in multiple, independent patient populations,
and this also held true in a large external validation patient cohort. Furthermore, our model’s
calculated 90-day mortality risk was highly correlated with the observed mortality rate across
all five risk quintiles. In particular, the model’s performance on the highest-risk quintile (the
calculated and observed 90-day mortality was 31.6% and 31.2%, respectively) suggests that
high-risk patients can be accurately identified. An additional model that included only 7 of the
9 variables and excluded length of stay and total cholesterol yielded an AUROC of 0.77 [95%
CI, 0.74–0.80]. Although the 7-variable model demonstrated improved identification ability
compared to MELD or MELD-Na, the improved prediction performance achieved by includ-
ing total cholesterol in MELD-Plus suggests that it may be beneficial for cholesterol labs to be
routinely collected in cirrhosis admission order sets.
The MELD score has been used extensively to predict patient outcomes, mortality, and
readmission rates in individuals with cirrhosis [25, 26, 4]. Furthermore, although MELD-Na
[17] was superior to MELD, the MELD-Plus score yielded improved levels of discrimination
consistently in all prediction models, with AUROCs that significantly outperformed the tradi-
tional scores. These findings suggest that new types of cirrhosis-related risk indexes utilizing
novel risk indicators may improve prognostication in this high-risk population.
MELD-Plus includes all MELD-Na’s components, as well as additional variables (albumin,
total cholesterol, WBC, age, and length of stay). It is logical that a predication model that has all
the MELD-Na model variables and additional ones would perform better, as was observed by
MELD-Plus. Not only that, but many of the variables have physiological plausibility for inclu-
sion in a prediction model. Decreased albumin correlated with worse outcomes in our model,
which may be the result of decreased albumin marking decreased liver function in cirrhosis
patients [25, 5]. Increasing age and length of hospital stay helped predict worse outcomes as well
as could be expected. Along with that, higher WBC was correlated with a worse prognosis,
potentially indicating poorer patient status (e.g., infection) at time of score calculation. Although
patients may have multiple WBC measurements during admission, our model is both internally
and externally valid because it uses the most recent WBC lab value. We chose the most recent
WBC during model development because the last available set of labs is more reflective of the
current health of patients than older measurements. Surprisingly, increased total cholesterol pre-
dicted a more favorable prognosis. Although unintuitive at first, this aligns with previous reports
that claim cholesterol levels become less of a risk factor or even an inverse risk factor for mortal-
ity because serious diseases may lower cholesterol soon before death occurs [27].
Although our study describes analyses of retrospective medical databases, the proposed
score could be used to identify patients that are at a high-risk of mortality in real time and thus
may inform risk-stratification and therapeutic decision-making. In a desirable scenario, our
score could be calculated automatically as an integrated component of an EMR system; the

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 11 / 15

 The MELD-Plus

clinician would see a risk score (probability) or a risk quantile (highest, lowest, or in between)
associated with the discharged patient, and this could be used to guide outpatient monitoring
strategies. With further validation, the MELD-Plus score could also be used longitudinally in
outpatients to monitor disease progression and/or responses to therapy.
Our study has limitations. First, it is a retrospective analysis limited to two academic, ter-
tiary-care hospitals. Even though we validated our model on a large external patient cohort,
subsequent studies must further assess the validity of our model in the external population and
consider different age ranges, coding systems, and data-collection methods. Second, the cir-
rhosis populations may vary at different centers—for example, alcohol use might significantly
vary between patients residing in the Boston area versus patients residing in other states [28].
Furthermore, although MGH and BWH are urban care facilities, the high prevalence of rural
populations at the IBM Explorys Network might affect prediction performance. Third,
although mortality was recorded, either through linkage to the social security master death
index as in the MGH/BWH models or through using EMR or billing/claims in the IBM Exp-
lorys model, such death indications may under-represent the true mortality rates. To minimize
this potential under-representation, we considered only patients who survived the study fol-
low-up. All patients had EMR data entries (such as laboratory measurements) after the study
follow-up, indicating survival, or had a recorded indication of death during the study follow-
up, with no EMR data entries found afterward.
Another limitation of MELD-Plus is that it did not specifically consider which procedures
patients underwent during the cirrhosis-related admissions. Furthermore, all the patients con-
sidered in our models survived the admission, but neither MGH/BWH’s nor IBM’s databases
contained information on post-discharge cause of death. To further assess MELD-Plus’s appli-
cability in clinical practice, future analyses should consider subgroups of patients to determine
linkages between invasive inpatient procedures and causes of mortality. Regardless of this limi-
tation, however, our MELD-Plus displayed validity in predicting overall mortality, which is
clinically applicable, because it provides clinicians with information on populations of patients
who need more intense or closer care.
Although we excluded elective admissions for liver biopsy, radiofrequency ablation, trans-
arterial chemoembolization, hepatic resection, or liver transplant, these criteria might exclude
patients with early and intermediate hepatocellular carcinoma (HCC), but not patients with
advanced HCC who underwent medical treatments only. Liver cancer can lead to early mortal-
ity, even in patients with mild liver cirrhosis, and, as such, our exclusion criteria may reduce
the applicability of our risk score when applying it to patients with more advanced HCC. Fur-
thermore, because we excluded admissions associated with a liver transplant, mortality risk
may decrease after a cirrhosis-related admission if patients successfully underwent a transplant
in a preceding admission.
Another limitation of our study is algorithmic. The adaptive LASSO method identified 9
predictors and left out variables that may also be correlated with predicting death. Feature
selection algorithms are known to be blind to the clinical importance of variables, and when
highly correlated predictors are identified, the algorithm randomly selects one. On the one
hand, important variables such as ascites, hepatocellular carcinoma, and diuretic medications
were not selected as predictors. On the other hand, the feature selection algorithm assures that
a minimal set of covariates produce a high level of prediction accuracy. Furthermore, we con-
ducted our model performance evaluation on a held-out data set not used for training.
Although a prediction model’s error usually decreases when more variables are included, this
is not always the case. This is true when performance is evaluated on the training set (due to
overfitting) but not the case when performance is evaluated on a held-out test dataset, as was
used across all our models.

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 12 / 15

 The MELD-Plus

In conclusion, we describe an unbiased and well-validated score to estimate 90-day mortal-

ity after a cirrhosis-related admission. This score, comprising a small set of easily available clin-
ical variables extracted from EMRs, improved the MELD and MELD-Na scores in predicting
90-day mortality and approximate 1-year mortality. In addition, we identified high-risk
patients with great accuracy. MELD-Plus’s strong performance demonstrates potential for it to
replace current standard models, allowing for greater accuracy in the identification of high-
risk cirrhosis patients.

Supporting information
S1 Table. Summary of variables.
(DOCX)
S2 Table. Billing codes used to define conditions.
(DOCX)
S3 Table. Comparison of variables in patients who died vs. survived 90-days after dis-
charge.
(DOCX)
S4 Table. Comparison between the highest-risk (1st) and the lowest-risk (5th) quintiles.
(DOCX)

Acknowledgments
We acknowledge Chin Hur MD MPH (Director, GI Health Outcomes Research, GI Unit, Mas-
sachusetts General Hospital / Harvard Medical School) for his critical review of this manuscript.

Author Contributions
Conceptualization: Uri Kartoun, Kathleen E. Corey, Stanley Y. Shaw.
Data curation: Uri Kartoun, Rahul Aggarwal.
Formal analysis: Uri Kartoun, Rahul Aggarwal, Kenney Ng, Stanley Y. Shaw.
Funding acquisition: Kathleen E. Corey, Stanley Y. Shaw.
Investigation: Stanley Y. Shaw.
Methodology: Uri Kartoun, Kathleen E. Corey, Kenney Ng, Stanley Y. Shaw.
Resources: Stanley Y. Shaw.
Software: Uri Kartoun, Kenney Ng.
Supervision: Kathleen E. Corey, Stanley Y. Shaw.
Validation: Uri Kartoun, Rahul Aggarwal, Kenney Ng.
Writing – original draft: Uri Kartoun.
Writing – review & editing: Kathleen E. Corey, Tracey G. Simon, Hui Zheng, Rahul Aggar-
wal, Kenney Ng, Stanley Y. Shaw.

References
1. Kim WR, Brown RS Jr, Terrault NA, El-Serag H. Burden of liver disease in the United States: summary
of a workshop. Hepatology 2002; 36:227–42. https://doi.org/10.1053/jhep.2002.34734 PMID:
12085369

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 13 / 15

 The MELD-Plus

2. Blachier M, Leleu H, Peck-Radosavljevic M, Valla DC, Roudot-Thoraval F. The burden of liver disease
in Europe: a review of available epidemiological data. J Hepatol 2013; 58(3):593–608. https://doi.org/
10.1016/j.jhep.2012.12.005 PMID: 23419824
3. Rakoski MO, McCammon RJ, Piette JD, Iwashyna TJ, Marrero JA, Lok AS, et al. Burden of cirrhosis on
older Americans and their families: analysis of the health and retirement study. Hepatology 2012; 55
(1):184–91. https://doi.org/10.1002/hep.24616 PMID: 21858847
4. Johnson KB, Campbell EJ, Chi H, Zheng H, King LY, Wu Y, et al. Advanced disease, diuretic use, and
marital status predict hospital admissions in an ambulatory cirrhosis cohort. Dig Dis Sci 2014; 59
(1):174–82. https://doi.org/10.1007/s10620-013-2832-5 PMID: 23990035
5. Younossi ZM, Henry L, Stepanova M. A new comorbidity model for predicting mortality in patients with
cirrhosis: does it work? Gastroenterology 2014; 146(1):19–24. https://doi.org/10.1053/j.gastro.2013.11.
026 PMID: 24287302
6. Kamath PS, Kim WR. The model for end-stage liver disease (MELD). Hepatology 2007; 45(3):797–805.
https://doi.org/10.1002/hep.21563 PMID: 17326206
7. Gotthardt D, Weiss KH, Baumgärtner M, Zahn A, Stremmel W, Schmidt J, et al. Limitations of the
MELD score in predicting mortality or need for removal from waiting list in patients awaiting liver trans-
plantation. BMC Gastroenterol 2009; 9(72):1–7.
8. Huo TI, Lin HC, Wu JC, Hou MC, Lee FY, Lee PC, et al. Limitation of the model for end-stage liver dis-
ease for outcome prediction in patients with cirrhosis-related complications. Clin Transplant 2006; 20
(2):188–94. https://doi.org/10.1111/j.1399-0012.2005.00463.x PMID: 16640525
9. Montano-Loza AJ, Duarte-Rojo A, Meza-Junco J, Baracos VE, Sawyer MB, Pang JX, et al. Inclusion of
Sarcopenia within MELD (MELD-Sarcopenia) and the prediction of mortality in patients with cirrhosis.
Clin Transl Gastroenterol 2015; 6(e102):1–8.
10. Kohane IS, Drazen JM, Campion EW. A glimpse of the next 100 years in medicine. N Engl J Med 2012;
367:2538–39. https://doi.org/10.1056/NEJMe1213371 PMID: 23268669
11. Liao KP, Kurreeman F, Li G, Duclos G, Murphy S, Guzman R, et al. Associations of autoantibodies,
autoimmune risk alleles, and clinical diagnoses from the electronic medical records in rheumatoid arthri-
tis cases and non-rheumatoid arthritis controls. Arthritis Rheum 2013; 65(3):571–81. https://doi.org/10.
1002/art.37801 PMID: 23233247
12. Doshi-Velez F, Ge Y, Kohane I. Comorbidity clusters in autism spectrum disorders: an electronic health
record time-series analysis. Pediatrics 2014; 133(1):e54–e63. https://doi.org/10.1542/peds.2013-0819
PMID: 24323995
13. Liao KP, Cai T, Savova GK, Murphy SN, Karlson EW, Ananthakrishnan AN, et al. Development of phe-
notype algorithms using electronic medical records and incorporating natural language processing.
BMJ 2015; 350:h1885. https://doi.org/10.1136/bmj.h1885 PMID: 25911572
14. Beam AL, Kartoun U, Pai JK, Chatterjee AK, Fitzgerald TP, Shaw SY, Kohane IS. Predictive modeling
of physician-patient dynamics that influence sleep medication prescriptions and clinical decision-mak-
ing. Sci Rep 2017; 7(42282); https://doi.org/10.1038/srep42282 PMID: 28181568
15. 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):864–71.
https://doi.org/10.1053/he.2000.5852 PMID: 10733541
16. Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM, Kosberg CL, et al. A model to pre-
dict survival in patients with end-stage liver disease. Hepatology 2001; 33(2):464–70. https://doi.org/10.
1053/jhep.2001.22172 PMID: 11172350
17. Mulligan DC, Hirose R. OPTN / UNOS, Liver and Intestinal Organ Transplantation Committee. Report
to the Board of Directors. June 23–24, 2014. Richmond, Virginia.
18. Myers RP, Quan H, Hubbard JN, Shaheen AA, Kaplan GG. Predicting in-hospital mortality in patients
with cirrhosis: results differ across risk adjustment methods. Hepatology 2009; 49(2):568–77. https://
doi.org/10.1002/hep.22676 PMID: 19085957
19. Zou H. The adaptive lasso and its oracle properties. J Am Stat Assoc 2006; 101:1418–29.
20. Liao KP, Cai T, Gainer V, Goryachev S, Zeng-treitler Q, Raychaudhuri S, et al. Electronic medical rec-
ords for discovery research in rheumatoid arthritis. Arthritis Care Res (Hoboken) 2010; 62(8):1120–7.
21. Ananthakrishnan AN, Cai T, Savova G, Cheng SC, Chen P, Perez RG, et al. Improving case definition
of Crohn’s disease and ulcerative colitis in electronic medical records using natural language process-
ing: a novel informatics approach. Inflamm Bowel Dis 2013; 19(7):1411–20. https://doi.org/10.1097/
MIB.0b013e31828133fd PMID: 23567779
22. Corey KE, Kartoun U, Zheng H, Shaw SY. Development and validation of an algorithm to identify nonal-
coholic fatty liver disease in the electronic medical record. Dig Dis Sci 2016; 61(3):913–9. https://doi.
org/10.1007/s10620-015-3952-x PMID: 26537487

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 14 / 15

 The MELD-Plus

23. Ihaka R. & Gentleman R. R: a language for data analysis and graphics. J Comput Graph Stat 1996;
5:299–314.
24. The IBM Explorys Platform / Solution Brief. IBM Watson Health. IBM Corporation 2016.
25. Singal AG, Rahimi RS, Clark C, Ma Y, Cuthbert JA, Rockey DC, et al. An automated model using elec-
tronic medical record data identifies patients with cirrhosis at high risk for readmission. Clin Gastroen-
terol Hepatol 2013; 11(10):1335–41. https://doi.org/10.1016/j.cgh.2013.03.022 PMID: 23591286
26. Volk ML, Tocco RS, Bazick J, Rakoski MO, Lok AS. Hospital readmissions among patients with decom-
pensated cirrhosis. Am J Gastroenterol 2012; 107(2):247–52. https://doi.org/10.1038/ajg.2011.314
PMID: 21931378
27. Ravnskov U, Diamond DM, Hama R, Hamazaki T, Hammarskjöld B, Hynes N, Kendrick M, Langsjoen
PH, Malhotra A, Mascitelli L, McCully KS, Ogushi Y, Okuyama H, Rosch PJ, Schersten T, Sultan S,
Sundberg R. Lack of an association or an inverse association between low-density-lipoprotein choles-
terol and mortality in the elderly: a systematic review. BMJ Open 2016; 6(6):e010401. https://doi.org/10.
1136/bmjopen-2015-010401 PMID: 27292972
28. Sacks JJ, Gonzales KR, Bouchery EE, Tomedi LE, Brewer RD. 2010 National and State Costs of
Excessive Alcohol Consumption. Am J Prev Med 2015; 49(5):e73–9. https://doi.org/10.1016/j.amepre.
2015.05.031 PMID: 26477807

PLOS ONE | https://doi.org/10.1371/journal.pone.0186301 October 25, 2017 15 / 15