Downloaded from [Link] on February 27, 2014 - Published by [Link].

com

Pancreas

The early prediction of mortality in acute pancreatitis:

a large population-based study
B U Wu,1 R S Johannes,1,2 X Sun,2 Y Tabak,2 D L Conwell,1 P A Banks1

See Commentary, p 1645 ABSTRACT originally developed as an intensive care instru-

1
Brigham and Women’s Background: Identification of patients at risk for ment and requires the collection of a large number
Hospital, Division of mortality early in the course of acute pancreatitis (AP) is of parameters, some of which may not be relevant
Gastroenterology, Center for an important step in improving outcome. to prognosis in AP.
Pancreatic Disease, Harvard Methods: Using Classification and Regression Tree The purpose of this study was to develop a
Medical School, Boston
Massachusetts, USA; 2 Cardinal (CART) analysis, a clinical scoring system was developed simple and accurate clinical scoring system for
Health, Marlborough, for prediction of in-hospital mortality in AP. The scoring stratifying patents according to their risk of in-
Massachusetts, USA system was derived on data collected from 17 992 cases hospital mortality. To develop a clinical tool useful
of AP from 212 hospitals in 2000–2001. The new scoring early in the disease course, we examined data
Correspondence to:
Dr B U Wu, Division of system was validated on data collected from 18 256 AP collected within the first 24 h of hospitalisation.
Gastroenterology, Endoscopy cases from 177 hospitals in 2004–2005. The accuracy of We used data collected from a large population-
Suite, Brigham & Women’s the scoring system for prediction of mortality was based cohort study in both the derivation and
Hospital, 75 Francis Street, measured by the area under the receiver operating
Boston, MA 02115, USA;
validation of the scoring system. For further
buwu@[Link] characteristic curve (AUC). The performance of the new validation, we compared the accuracy of the new
scoring system was further validated by comparing its scoring system with that of the APACHE II for
Revised 18 April 2008 predictive accuracy with that of Acute Physiology and prediction of mortality.
Accepted 13 May 2008 Chronic Health Examination (APACHE) II.
Published Online First Results: CART analysis identified five variables for
2 June 2008
prediction of in-hospital mortality. One point is assigned METHODS
for the presence of each of the following during the first
Patient population and data collection
24 h: blood urea nitrogen (BUN) .25 mg/dl; impaired
The current study was approved by the Brigham &
mental status; systemic inflammatory response syndrome
Women’s Hospital Institutional Review Board.
(SIRS); age .60 years; or the presence of a pleural
Patient data were generated from the Cardinal
effusion (BISAP). Mortality ranged from .20% in the
Health Clinical Outcomes Research Database
highest risk group to ,1% in the lowest risk group. In the
(Cardinal Health Clinical Research Services,
validation cohort, the BISAP AUC was 0.82 (95% CI 0.79
Cardinal Health, Marlborough, Massachusetts,
to 0.84) versus APACHE II AUC of 0.83 (95% CI 0.80 to
USA). This large population data set has supported
0.85).
public reporting of hospital performance in
Conclusions: A new mortality-based prognostic scoring
Pennsylvania and elsewhere for purposes of quality
system for use in AP has been derived and validated. The
improvement for .20 years. Details of the data
BISAP is a simple and accurate method for the early
collection and abstraction process for the Cardinal
identification of patients at increased risk for in-hospital
database have been described previously.10–13 The
mortality.
database contains information on patient demo-
graphics, vital signs, laboratory values, co-morbid-
Acute pancreatitis (AP) is a disease with a ities, physical exam findings as well as procedure
substantial burden on the US healthcare system. and diagnosis codes. Unlike previous versions of
Recent data indicate a rise in the absolute number the database, all laboratory and vital sign data are
as well as the rate of emergency room visits, now recorded as continuous values.
hospital admissions and direct healthcare costs for The derivation cohort consisted of all cases in
AP in the USA (210 000 admissions in 2002; 4.6 of the Cardinal Health Research Database with
every 1000 hospitalisations from 1988 to 2003, principal diagnosis (from the International
annual direct costs in excess of US$2 billion).1–3 Classification of Diseases, ninth revision, clinical
With an overall mortality rate of 2–5%,3 4 a reliable modification) ICD9-CM 577.0 (AP) from January
method of risk stratification for AP is of significant 2000 to December 2001. The validation cohort
clinical importance. included all patients with the principal diagnosis of
Current methods of risk stratification in AP have AP admitted from January 2004 to September
important limitations. The Ranson5 and modified 2005.
Glasgow score6 contain data not routinely collected
at the time of hospitalisation. In addition, both Assessment of risk factors for mortality in AP
require 48 h to complete, missing a potentially We considered the following candidate risk factors
valuable early therapeutic window.5 7 The most in model development:
commonly utilised prediction scoring system for 5
c Individual Ranson signs : age, white blood cell
clinical research studies in AP is the Acute (WBC) count, glucose, aspartate aminotransfer-
Physiology and Chronic Health Examination ase (AST), blood urea nitrogen (BUN), lactate
(APACHE) II.8 9 However, the APACHE II was dehydrogenase (LDH) and serum calcium.

1698 Gut 2008;57:1698–1703. doi:10.1136/gut.2008.152702

 Downloaded from [Link] on February 27, 2014 - Published by [Link]

Pancreas

c Pleural effusion14 15 (on chest radiography or CT). accuracy. In order to assess model calibration, we compared
c The systemic inflammatory response syndrome (SIRS)
16 17
observed mortality by point score in both the derivation and
defined by the presence of >2 of the following criteria: validation cohorts. In testing the new scoring system on the
– pulse .90 beats/min validation cohort we treated missing data as normal (reference
– respirations .20/min or PaCO2 ,32 mmHg range) values.
To validate the model further, we compared its performance
– temperature .38uC or ,36uC
with that of the APACHE II. We anticipated that a large
– WBC count .12 000 or ,4000 cells/mm3 or .10% number of patients would not have complete data for
immature neutrophils (bands)
calculation of an APACHE II score. Therefore, in generating
c Haemoconcentration (haemoglobin was included as a
APACHE II scores, we treated missing data as normal (reference
continuous variable)18
19
range) values. Comparison of model AUC with the APACHE II
c Atlanta Symposium criteria for organ failure : systolic
was performed using the method described by De Long et al.25
blood pressure, creatinine, partial pressure of arterial oxygen
(PaO2).
20
c Altered mental status : defined as any record of disorienta-
Subgroup analysis
tion, lethargy somnolence, coma or stupor in the medical The ability to identify patients at increased risk of mortality
record. from AP prior to the onset of overt organ failure is of significant
In order to develop a model with widespread applicability, we clinical importance. Patients who present with evidence of
limited potential laboratory and vital sign parameters to those organ failure within the first 24 h of hospitalisation have
with collection rates of >85%. SIRS was collected as a already declared themselves as being at increased risk for
dichotomous composite variable. Physical exam findings and experiencing persistent organ failure and death.26 Numerous
co-morbidities were recorded at admission. Data for laboratory organ failure scoring systems exist for use in critical care settings
tests and vital signs were recorded for the first 24 h admission that measure the extent of organ failure. An important
period including emergency department values. The worst application of a new scoring system in AP is to identify patients
(most extreme) value for vital signs and laboratory data within at risk for mortality prior to the onset of organ failure. We were
the first 24 h of admission were utilised for model development. interested in determining how well the new scoring system
All laboratory and vital sign parameters were included as could predict mortality among patients without evidence of
continuous variables with thresholds determined by early organ failure during the first 24 h of hospitalisation.
Classification and Regression Tree (CART) analysis. Therefore, we performed a subgroup analysis in which we
applied the new prediction rule exclusively to patients without
evidence of early organ failure by Atlanta criteria (creatinine
Model derivation .2.0 mg/dl, systolic blood pressure ,90 mmHg or PaO2
We used CART analysis to identify factors for use in the new ,60 mmHg on arterial blood gas).
clinical prediction rule. CART is a non-parametric, empiric
statistical method21 that has been increasingly utilised for
clinical applications across a number of disease groups22–24 but Statistical analysis
not as yet for clinical prediction in AP. Patients are classified CART analysis was performed using the CART statistical
into two groups at each stage of analysis based on classification software package (CART Professional Extended Edition version
variables. The optimum split point for each variable is 6.0, Salford Systems, California Statistical Software, San Diego,
determined by a statistical search algorithm. Patients are California, USA). Additional statistical analysis was performed
grouped into nodes by cut-off points for classification variables. in SAS statistical software version 9.1 (SAS Institute, Cary,
The process is reiterated for subsequent classification variables. North Carolina, USA). All reported p values are two sided. We
Tree building is carried forward until a pruning process used the Bonferonni method of adjustment for multiple testing
determines the optimum tree size without overfitting the data. when examining differences in mortality between risk groups.
In order to avoid model overtraining, we used 10-fold cross-
validation in tree development. In addition, we specified a 10- RESULTS
fold misclassification cost such that misclassifying a true death Patient characteristics
was 10 times worse than misclassification of a case that In the derivation cohort, there were 17 922 cases of AP
ultimately survived. We used the Gini index as the splitting rule identified from 212 hospitals. Median age was 53 years, and
in tree building. Missing data were incorporated into the tree 50.5% were men. In the validation cohort there were 18 256
building process through use of surrogate splits. cases of AP identified from 177 hospitals. Median age was 53
A new prediction rule was subsequently generated from the years, and 49.4% were men.
parameters identified in the CART analysis. Specifically, a scoring There were a total of 335 (1.9%) deaths in the derivation
system was created in which one point was assigned for the cohort and 234 deaths (1.28%) in the validation cohort. There
presence of each parameter identified in the CART prediction was a significant reduction in overall mortality between
algorithm. We then calculated scores for each case in the derivation 2000–2001 and 2004–2005 (x2 p,0.001). Distributions for
cohort and compared this with their observed outcome. demographic and clinical features between the two study
populations are depicted in table 1. The serum calcium, LDH,
Model validation PaO2 and AST measurements were excluded from further
To validate performance of the new scoring system on a consideration in developing the new scoring system due to their
separate group of patients, we tested its ability to predict in- failure to meet the pre-specified 85% collection rate threshold.
hospital mortality in the validation cohort. We calculated scores
for each case and compared this with observed mortality. From CART analysis
this analysis we calculated the area under the receiver operating Using CART analysis we identified five variables as most
characteristic curve (AUC) as a measure of discrimination efficient in stratifying patients according to risk of mortality:

Gut 2008;57:1698–1703. doi:10.1136/gut.2008.152702 1699

 Downloaded from [Link] on February 27, 2014 - Published by [Link]

Pancreas

Table 1 Demographic and clinical characteristics of the two study populations

Derivation cohort Validation cohort
January 2000–December 2001 January 2004–September 2005

Hospital demographics
Hospitals 212 177
Teaching hospitals 49.9% 39.7%
Bed size
,100 beds 10% 27%
100–300 beds 58% 50%
.300 beds 32% 23%
Patient demographics and co-morbidity
Acute pancreatitis cases (n) 17922 18256
Median age in years (range) 53 (18–106) 53 (18–108)
Men 50.5% 49.4%
Gallstones 25.4% 23.8%
Alcohol history (dependency or abuse) 22.4% 21.1%
Chronic pancreatitis 7.0% 8.5%
Congestive heart failure 7.04% 6.5%
Chronic kidney disease 5.1% 5.7%
Chronic obstructive pulmonary disease 12.0% 10.4%
Clinical parameters
SIRS present* 48.96% 50.3%
WBC count 6103 (median, IQR) 10.2 (7.5–13.7) 9.9 (7.3–13.3)
Temperature, uC (median IQR) 36.1 (35.6 to 36.7) 36.4 (36.1–37.0)
Pulse, beats/min (median, IQR) 70 (60–100) 85 (70–100)
Respirations per minute (median, IQR) 10 (10–20) 18 (10–20)
Systolic blood pressure, mmHg (median, IQR)* 140 (130–160) 134 (118–160)
Haemoglobin, mg/dl (median, IQR)* 13.8 (12.4–15.1) 13.8 (12.4–15.1)
Blood urea nitrogen, mg/dl (median, IQR)* 15 (10–21) 14 (10–21)
Creatinine, mg/dl (median, IQR)* 0.9 (0.8–1.2) 0.9 (0.8–1.2)
Glucose, mg/dl (median, IQR)* 128 (105–170) 123 (102–163)
Pleural effusion* 4.1% 4.4%
Altered mental status* 7.4% 6.7%
Proportions listed in this table reflect the number of patients with each finding divided by the total number of patients in each
cohort. Vital signs and laboratory test values were the most extreme within the first 24 h of admission.
*Candidate variables entered into CART analysis.
CART, Classification and Regression Tree; IQR, interquartile range; SIRS, systemic inflammatory response syndrome; WBC, white
blood cell.

BUN .25 mg/dl, impaired mental status, SIRS, age .60 years was a significant trend towards higher mortality with increas-
and pleural effusion (BISAP). The CART tree is depicted in fig 1. ing BISAP score (Cochrane–Armitage trend test p,0.001). In
BUN was identified as the most efficient first splitting variable. addition, significant differences existed between risk groups (x2
Age and SIRS further discriminated between high- and low-risk p,0.001 overall, pairwise x2 Bonferroni-adjusted p,0.001).
cases. The remaining parameters (mental status and pleural Below average mortality was observed in patients with ,2
effusion) served to differentiate intermediate risk patients points (,1.0% mortality). Patients with a score of 2 had
further. increased mortality (2%). Mortality continued to rise sharply
with BISAP scores of >3 (5–20%).
Validation of scoring system In both the derivation and validation cohort, the majority of
The five variables from the CART were incorporated into a new patients (,60%) presented with BISAP scores ,2 and were at
scoring system in which the presence of each variable very low risk for mortality (,1.0%). The new scoring system
contributes one point to an overall 5-point score. After was able to identify subgroups of patients (those with scores of
calculating scores for patients in the derivation cohort, the >3) with substantially increased risk of dying in the course of
BISAP score AUC was 0.83 for prediction of in-hospital their hospitalisation.
mortality.
In the validation cohort there were 17 350 (96.8%) cases with APACHE II
complete laboratory and vital sign data for the five parameters There were 405 (2.2%) patients with complete data for
included in the scoring system. There were 213 (1.2%) deaths APACHE II. There were 40 (9.9%) deaths among these patients.
among these patients. After calculating BISAP scores for After imputation, APACHE II scores were calculated for each
patients in the validation cohort, the AUC for prediction of patient as previously described. The median calculated APACHE
in-hospital mortality was 0.82. II score for the 2004–2005 AP population was 7. The receiver
Table 2 depicts observed mortality stratified by BISAP point operating characteristic (ROC) curve for the APACHE II score is
score in both cohorts of patients. Also depicted in table 2 is the shown in fig 2. Calculated AUC was 0.83 (95% CI 0.80 to 0.85).
frequency of patients within each score category. Using the 5- For purposes of comparison, a similar ROC curve was plotted
point scoring system, patients could be reliably classified within for the BISAP score on the 2004–2005 population. The BISAP
24 h of admission into distinct risk groups for mortality. There AUC was 0.82 (95% CI 0.79 to 0.84) in the validation cohort.

1700 Gut 2008;57:1698–1703. doi:10.1136/gut.2008.152702

 Downloaded from [Link] on February 27, 2014 - Published by [Link]

Pancreas

Figure 1 Classification and Regression

Tree analysis of risk factors for mortality
in the derivation cohort. Mortality rates
listed are crude (unadjusted) values.
AP, acute pancreatitis; BUN, blood urea
nitrogen; IMS, impaired mental status;
PI effusion, pleural effusion;
SIRS, systemic inflammatory response
syndrome.

Significance testing for differences in AUC between BISAP and DISCUSSION

APACHE II yielded a x2 p value of 0.2, indicating no significant We have derived and validated the first population-based
difference in predictive accuracy. prognostic scoring system for use in AP. Using BUN, impaired
mental status, SIRS, age and pleural effusion (BISAP), we were
Subgroup analysis (classification prior to onset of organ failure)) able to stratify patients within the first 24 h of hospitalisation
There were 1753 patients with early organ failure in the into distinct risk groups for in-hospital mortality.
validation cohort. Among these patients there were 98 (5.5%) In the subgroup analysis we examined the ability of the new
deaths. After we excluded cases with early organ failure, prediction rule to identify patients at increased risk of mortality
there were 16 503 cases remaining in the subgroup analysis. prior to the onset of organ failure. Specifically, we excluded
Among these cases there were 136 (0.8%) deaths. Fifty-eight patients with evidence of early organ failure by Atlanta criteria
percent of patients that died did not have evidence of organ (within the first 24 h). Although patients without early organ
failure by Atlanta criteria within the first 24 h of hospitalisa- failure had a low mortality (0.8%), 58% of the patients that
tion. Table 3 depicts the observed mortality by BISAP score ultimately died came from this subgroup. Among these patients,
among patients without evidence of early organ failure. The the prediction rule was still able to identify patients with
scoring system was once again able reliably to identify substantially increased mortality.
patients at increased risk of mortality in this subgroup The ability to risk-stratify patients early in their disease
analysis. The model’s AUC for prediction of in-hospital course has several important implications. First, early identifica-
mortality in the subgroup analysis was 0.79 versus APACHE tion of high-risk patients may alert doctors to institute
II AUC of 0.78 (x2 = 0.45, p = 0.5). aggressive resuscitation efforts and to consider specialty care

Table 2 Observed mortality by BISAP point score in the derivation and validation cohorts
Derivation cohort, n = 17 922 Validation cohort, n = 18 256
BISAP score Number of cases Observed mortality Number of cases Observed mortality

0 5121 0.2% 4912 0.1%

I 7206 0.7% 7722 0.5%
2 3829 2.1% 3941 1.9%
3 1390 8.3% 1349 5.3%
4 331 19.3% 292 12.7%
5 45 26.7% 40 22.5%
Overall x2 p,0.001 for each study group. Within study group x2 p,0.001 for differences in mortality between individual risk
(score) categories in pairwise tests of significance.
BISAP, blood urea nitrogen, impaired mental status, systemic inflammatory response syndrome, age and pleural effusion.

Gut 2008;57:1698–1703. doi:10.1136/gut.2008.152702 1701

 Downloaded from [Link] on February 27, 2014 - Published by [Link]

Pancreas

Table 3 Subgroup analysis of the validation cohort

excluding cases with evidence of organ failure within first
24 h of hospitalisation
n = 16503
BISAP score Number of cases Observed mortality

0 4796 0.1%
1 7287 0.4%
2 3307 1.6%
3 916 3.6%
4 176 7.4%
5 21 9.5%
Overall and pairwise x2 p,0.001.
BISAP, blood urea nitrogen, impaired mental status, systemic
inflammatory response syndrome, age and pleural effusion.

mortality (1.3% validation cohort vs 1.9% derivation cohort, t

test p,0.0001). We validated the new scoring system in the
more recent population of patients who may have benefited
from improvements in critical care and the management of
Figure 2 Receiver operator characteristic curves. BISAP (blood urea severe AP, including changes in the management of sterile as
nitrogen, impaired mental status, systemic inflammatory response well as infected necrosis.4 These changes may have contributed
syndrome, age and pleural effusion.) derivation (area under the curve to the reduced mortality observed among patients with higher
(AUC) = 0.83). BISAP validation (AUC = 0.82). Acute Physiology and BISAP scores in the validation cohort.
Chronic Health Examination (APACHE) II (AUC = 0.83). The early identification of patients at risk for adverse
outcome from AP has been an area of active investigation for
referral. Second, a severity index provides standardised criteria many years.16 28–39 Previous studies have attempted either to
for enrolment of subjects into future clinical studies. In develop prognostic scoring systems or to identify individual risk
addition, a population-based system of risk stratification factors for severe disease. Some of these studies have included
provides an instrument for additional outcomes research. For mortality as an end point.16 28 30–32 35–37 Among recently proposed
example, identification of factors associated with death among prognostic scoring systems, three have used data collected
patients with low BISAP scores may help to lead to improve- within the first 24 h of hospitalisation.28 31 40 Because all of these
ments in future management strategies in AP. scoring systems were based on data from high-risk patient
The primary advantage of BISAP is simplicity. The presence populations, their ability to predict in-hospital mortality among
of each variable contributes one point to a total 5-point score. patients with varying disease severity is unknown.
There is no need for additional computation. In addition, each Among studies of individual or combinations of risk factors in
of the parameters can be easily obtained early in the course of a AP, several have focused on the first 24 h admission per-
general hospital admission. The only subjective parameter in the iod.32 35 37 39 41 42 These smaller cohort studies identified age,35
new scoring system is the assessment of mental status. obesity,30 32 glucose,37 42 serum creatinine,37 BUN42 and organ
Although an uncommon finding, the presence of an altered failure41 43 as admission parameters associated with increased
mental state was a significant predictor of mortality in both mortality. Of these parameters, we were able to evaluate age,
populations. Although the Glasgow Coma Score is used as part glucose, BUN and organ failure (in terms of hypotension,
of the calculation of an APACHE II score as well as the Multiple elevated creatinine and hypoxia) within the first 24 h.
Organ Failure Score,20 26 we simplified determination of this To evaluate the performance of the new prediction rule
parameter by developing the model in such a way that any further, we compared its predictive accuracy with that of
evidence of disorientation or further disturbance in mental APACHE II. Although more recent versions of the APACHE
status qualifies as a positive finding. Although SIRS is a system have been developed, the APACHE II remains the most
composite parameter that involves the use of four criteria, widely accepted method for risk stratification in AP4 8 9 44 45 and
evaluation of the systemic inflammatory response has become was therefore chosen as the reference standard. Its major
increasingly widespread in clinical practice and has also been limitations are complexity and reliance upon parameters not
demonstrated to have prognostic value in AP.16 17 routinely collected during a general hospital admission. For
The use of population-based data in this study has several example, in our validation cohort, only 2.2% of cases had
advantages. First, the large number of cases provided sufficient complete APACHE II data versus 96.8% for the laboratory and
power to focus on mortality as an end point. Second, the vital sign parameters included in the BISAP score. Moreover,
derivation and validation of this new scoring system utilised cases with a complete APACHE II score had a mortality rate of
data collected from a large number of hospitals. Patient 9.9%, which was eight times greater than that of the general
information was collected from community hospitals, tertiary population. This increased mortality most probably reflects a
care centres, teaching and non-teaching institutions. As a result, form of selection bias wherein more severely ill patients are the
the performance of BISAP in our study reflects the combined ones most likely to have the exhaustive data collection required
experience from a variety of treatment settings. for completion of an APACHE II score. Nevertheless, the BISAP
While patient characteristics were similar between the two score was able to achieve a similar level of predictive accuracy to
study populations, the validation cohort differed from the the more complex APACHE II score, with far fewer variables.
derivation cohort in several important aspects. These differences There were several potential limitations to the present study.
included a shift in hospital demographics and a decreased overall We relied upon ICD-9 data for diagnosis rather than the more

1702 Gut 2008;57:1698–1703. doi:10.1136/gut.2008.152702

 Downloaded from [Link] on February 27, 2014 - Published by [Link]

Pancreas

strict Atlanta Symposium criteria.19 As a result, milder causes of 20. Tran DD, Cuesta MA. Evaluation of severity in patients with acute pancreatitis.
Am J Gastroenterol 1992;87:604–8.
abdominal pain may have been misclassified as AP. 21. Breiman L. Classification and regression trees. Belmont, California: Wadsworth
Nevertheless, observed mortality in both cohorts of patients International Group, 1984.
was consistent with data from recent studies examining trends 22. Fonarow GC, Adams KF Jr, Abraham WT, et al. Risk stratification for in-hospital
in AP.3 4 46 47 A second limitation was limited information mortality in acutely decompensated heart failure: classification and regression tree
analysis. JAMA 2005;293:572–80.
regarding aetiology, obesity or initial versus recurrent episode 23. Miller PD, Barlas S, Brenneman SK, et al. An approach to identifying osteopenic
of AP, each of which may have prognostic value in AP.30 32 48 49 women at increased short-term risk of fracture. Arch Intern Med 2004;164:1113–20.
In summary, we have derived and validated a prognostic 24. Takahashi O, Cook EF, Nakamura T, et al. Risk stratification for in-hospital mortality
scoring system for use in AP. The BISAP score stratifies patients in spontaneous intracerebral haemorrhage: a Classification and Regression Tree
analysis. QJM 2006;99:743–50.
within the first 24 h of admission according to their risk of in- 25. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or
hospital mortality and was able to identify patients at increased more correlated receiver operating characteristic curves: a nonparametric approach.
risk of mortality prior to the onset of organ failure. The ability Biometrics 1988;44:837–45.
26. Johnson CD, Abu-Hilal M. Persistent organ failure during the first week as a marker
to risk-stratify patients early in their course is a major step to of fatal outcome in acute pancreatitis. Gut 2004;53:1340–4.
improving future management strategies in acute pancreatitis. 27. Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a severity of disease
classification system. Crit Care Med 1985;13:818–29.
Acknowledgements: The authors would like to thank Linda Hyde, Karen Derby and 28. Ueda T, Takeyama Y, Yasuda T, et al. Simple scoring system for the prediction of the
Stephen Kurtz of Cardinal Health for their assistance with data management. In prognosis of severe acute pancreatitis. Surgery 2007;141:51–8.
addition, we would like to thank Dr Earl F Cook of the Harvard School of Public Health 29. Rau BM, Bothe A, Kron M, et al. Role of early multisystem organ failure as major risk
for assistance with study design and data analysis. factor for pancreatic infections and death in severe acute pancreatitis. Clin
Gastroenterol Hepatol 2006;4:1053–61.
Competing interests: None. 30. Papachristou GI, Papachristou DJ, Avula H, et al. Obesity increases the severity of
Ethics approval: The study was approved by the Brigham & Women’s Hospital acute pancreatitis: performance of APACHE-O score and correlation with the
Institutional Review Board. inflammatory response. Pancreatology 2006;6:279–85.
31. Spitzer AL, Barcia AM, Schell MT, et al. Applying Ockham’s razor to pancreatitis
prognostication: a four-variable predictive model. Ann Surg 2006;243:380–8.
REFERENCES 32. Martinez J, Johnson CD, Sanchez-Paya J, et al. Obesity is a definitive risk factor of
1. Fagenholz PJ, Fernandez-del Castillo C, Harris NS, et al. Direct medical costs of severity and mortality in acute pancreatitis: an updated meta-analysis. Pancreatology
acute pancreatitis hospitalizations in the United States. Pancreas 2007;35:302–7. 2006;6:206–9.
2. Fagenholz PJ, Fernandez-del Castillo C, Harris NS, et al. National study of United 33. Mentula P, Kylanpaa ML, Kemppainen E, et al. Early prediction of organ failure by
States emergency department visits for acute pancreatitis, 1993–2003. BMC Emerg combined markers in patients with acute pancreatitis. Br J Surg 2005;92:68–75.
Med 2007;7:1. 34. Gan SI, Romagnuolo J. Admission hematocrit: a simple, useful and early predictor of
3. Fagenholz PJ, Castillo CF, Harris NS, et al. Increasing United States hospital severe pancreatitis. Dig Dis Sci 2004;49:1946–52.
admissions for acute pancreatitis, 1988–2003. Ann Epidemiol 2007;17:491–7. 35. Company L, Saez J, Martinez J, et al. Factors predicting mortality in severe acute
4. Banks PA, Freeman ML. Practice guidelines in acute pancreatitis. Am J Gastroenterol pancreatitis. Pancreatology 2003;3:144–8.
2006;101:2379–400. 36. Halonen KI, Leppaniemi AK, Lundin JE, et al. Predicting fatal outcome in the early
5. Ranson JH, Rifkind KM, Roses DF, et al. Objective early identification of severe acute phase of severe acute pancreatitis by using novel prognostic models. Pancreatology
pancreatitis. Am J Gastroenterol 1974;61:443–51. 2003;3:309–15.
6. Blamey SL, Imrie CW, O’Neill J, et al. Prognostic factors in acute pancreatitis. Gut 37. Blum T, Maisonneuve P, Lowenfels AB, et al. Fatal outcome in acute pancreatitis: its
1984;25:1340–6. occurrence and early prediction. Pancreatology 2001;1:237–41.
7. Ranson JH, Pasternack BS. Statistical methods for quantifying the severity of clinical 38. Losanoff JE, Asparouhov OK, Jones JW. Multiple factor scoring system for risk
acute pancreatitis. J Surg Res 1977;22:79–91. assessment of acute pancreatitis. J Surg Res 2001;101:73–8.
8. Yeung YP, Lam BY, Yip AW. APACHE system is better than Ranson system in the
39. Halonen KI, Leppaniemi AK, Puolakkainen PA, et al. Severe acute pancreatitis:
prediction of severity of acute pancreatitis. Hepatobiliary Pancreat Dis Int 2006;5:294–9.
prognostic factors in 270 consecutive patients. Pancreas 2000;21:266–71.
9. Larvin M, McMahon MJ. APACHE-II score for assessment and monitoring of acute
40. Harrison DA, D’Amico G, Singer M. The Pancreatitis Outcome Prediction (POP)
pancreatitis. Lancet 1989;2:201–5.
Score: a new prognostic index for patients with severe acute pancreatitis. Crit Care
10. Iezzoni LI, Hotchkin EK, Ash AS, et al. MedisGroups data bases. The impact of data
Med 2007;35:1703–8.
collection guidelines on predicting in-hospital mortality. Med Care 1993;31:277–83.
11. Tabak YP, Johannes RS, Silber JH. Using automated clinical data for risk 41. Isenmann R, Rau B, Beger HG. Early severe acute pancreatitis: characteristics of a
adjustment: development and validation of six disease-specific mortality predictive new subgroup. Pancreas 2001;22:274–8.
models for pay-for-performance. Med Care 2007;45:789–805. 42. Fan ST, Choi TK, Lai EC, et al. Prediction of severity of acute pancreatitis: an
12. Iezzoni LI, Moskowitz MA. A clinical assessment of MedisGroups. JAMA alternative approach. Gut 1989;30:1591–5.
1988;260:3159–63. 43. Perez A, Whang EE, Brooks DC, et al. Is severity of necrotizing pancreatitis increased
13. Fine MJ, Auble TE, Yealy DM, et al. A prediction rule to identify low-risk patients in extended necrosis and infected necrosis? Pancreas 2002;25:229–33.
with community-acquired pneumonia. N Engl J Med 1997;336:243–50. 44. Taylor SL, Morgan DL, Denson KD, et al. A comparison of the Ranson, Glasgow, and
14. Heller SJ, Noordhoek E, Tenner SM, et al. Pleural effusion as a predictor of severity APACHE II scoring systems to a multiple organ system score in predicting patient
in acute pancreatitis. Pancreas 1997;15:222–5. outcome in pancreatitis. Am J Surg 2005;189:219–22.
15. Belfar HL, Radecki PD, Friedman AC, et al. Pancreatitis presenting as pleural 45. Robert JH, Frossard JL, Mermillod B, et al. Early prediction of acute pancreatitis:
effusions: computed tomography demonstration of pleural space extension of prospective study comparing computed tomography scans, Ranson, Glascow, Acute
pancreatitis exudate. J Comput Tomogr 1987;11:184–7. Physiology and Chronic Health Evaluation II scores, and various serum markers.
16. Mofidi R, Duff MD, Wigmore SJ, et al. Association between early systemic World J Surg 2002;26:612–9.
inflammatory response, severity of multiorgan dysfunction and death in acute 46. Yadav D, Lowenfels AB. Trends in the epidemiology of the first attack of acute
pancreatitis. Br J Surg 2006;93:738–44. pancreatitis: a systematic review. Pancreas 2006;33:323–30.
17. Buter A, Imrie CW, Carter CR, et al. Dynamic nature of early organ dysfunction 47. Frey CF, Zhou H, Harvey DJ, et al. The incidence and case-fatality rates of acute
determines outcome in acute pancreatitis. Br J Surg 2002;89:298–302. biliary, alcoholic, and idiopathic pancreatitis in California, 1994–2001. Pancreas
18. Brown A, Orav J, Banks PA. Hemoconcentration is an early marker for organ failure 2006;33:336–44.
and necrotizing pancreatitis. Pancreas 2000;20:367–72. 48. Pezzilli R, Billi P, Morselli-Labate AM. Severity of acute pancreatitis: relationship
19. Bradley EL 3rd. A clinically based classification system for acute pancreatitis. with etiology, sex and age. Hepato-gastroenterology 1998;45:1859–64.
Summary of the International Symposium on Acute Pancreatitis, Atlanta, Ga, 49. Lankisch PG, Assmus C, Pflichthofer D, et al. Which etiology causes the most
September 11 through 13, 1992. Arch Surg 1993;128:586–90. severe acute pancreatitis? Int J Pancreatol 1999;26:55–7.

Gut 2008;57:1698–1703. doi:10.1136/gut.2008.152702 1703

 Downloaded from [Link] on February 27, 2014 - Published by [Link]

The early prediction of mortality in acute

pancreatitis: a large population-based study
B U Wu, R S Johannes, X Sun, et al.

Gut 2008 57: 1698-1703 originally published online June 2, 2008

doi: 10.1136/gut.2008.152702

Updated information and services can be found at:

[Link]

These include:
References This article cites 48 articles, 4 of which can be accessed free at:
[Link]

Article cited in:

[Link]

Email alerting Receive free email alerts when new articles cite this article. Sign up in
service the box at the top right corner of the online article.

Topic Articles on similar topics can be found in the following collections

Collections
Pancreas and biliary tract (1829 articles)
Pancreatitis (493 articles)

Notes

To request permissions go to:

[Link]

To order reprints go to:

[Link]

To subscribe to BMJ go to:

[Link]