You are on page 1of 6

The Pancreatitis Outcome Prediction (POP) Score: A new prognostic index for patients with severe acute pancreatitis

*
David A. Harrison, PhD; Giovanna D’Amico, MD; Mervyn Singer, MD, FRCP

Objective: Severe acute pancreatitis, defined as pancreatitis with distant organ dysfunction, is a condition carrying a high mortality and morbidity. Current outcome prediction scores are based on small populations, usually from single specialist centers. Some scores cannot be applied until several days into hospital admission. We thus sought to develop a new and more sensitive outcome prediction score—the Pancreatitis Outcome Prediction (POP) Score—for these high-risk patients. Design: Retrospective cohort study of a large multicenter intensive care database. Setting: One hundred fifty-nine U.K. intensive care units. Patients: Participants were 2,462 patients admitted to intensive care units with severe acute pancreatitis. Interventions: None. Measurements and Main Results: Demographic, physiologic, and biochemical data collected within the first 24 hrs of intensive care unit admission were used to develop a risk prediction score using logistic regression. The six variables with the strongest

relationship to hospital outcome—arterial pH, age, serum urea, mean arterial pressure, PaO2/FIO2 ratio, and total serum calcium (in order of decreasing impact)—produced a model with a prognostic discrimination (area under the receiver operating characteristic curve ‫ ؍‬0.838) superior to other models. These six factors were used to develop an objectively weighted multivariate prognostic score ranging from 0 to 40 points. Conclusions: Prognostic stratification of patients with severe acute pancreatitis requiring intensive care offers a useful audit tool to gauge unit performance and improve delineation of subsets for prospective trials. Prospective validation of this new outcome prediction score is required, preferably in different countries. The validity of the POP Score for either hospital or intensive care admission could also be tested and assessed for superiority over existing scores. (Crit Care Med 2007; 35:1703–1708) KEY WORDS: pancreatitis; critical care; hospital mortality; intensive care units; models; statistical; severity of illness index

A

cute pancreatitis is a common disorder, affecting 10 –20 per million of the U.K. population per annum (1). The clinical picture ranges from relatively mild symptoms to multiple organ failure with intercurrent sepsis, particularly in the severe necrotizing or hemorrhagic forms of the disease. The overall mortality is reported as 10% to 15%, ranging from Ͻ5% for those presenting with mild acute pancreatitis to 20% to 25% with severe acute pancreatitis (1). Time to death is biphasic, occurring either within days of ad-

*See also p. 1787. From the Intensive Care National Audit & Research Centre, Tavistock House, Tavistock Square, London, United Kingdom (DAH); and Bloomsbury Institute of Intensive Care Medicine, University College London, Cruciform Building, Gower Street, London, United Kingdom (GD, MS). The authors have not disclosed any potential conflicts of interest. For information regarding this article, E-mail: david.harrison@icnarc.org Copyright © 2007 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: 10.1097/01.CCM.0000269031.13283.C8

mission or after a protracted hospital stay lasting weeks to months (2–5). Various prognostic scoring systems have been developed yet are generally based on either expert consensus—such as the Atlanta classification (4)— or small specialistcenter populations—such as the Ranson (6), Glasgow (7), and modified Glasgow (8) criteria— usually collected several decades ago and only containing a small proportion of patients with severe acute pancreatitis. Their applicability to the 21st century patient with severe acute pancreatitis in a more technologically advanced, albeit nonspecialist, intensive care unit (ICU) is uncertain. Indeed, a recent international consensus conference (9) redefined severe acute pancreatitis from a critical care perspective to represent pancreatitis in the context of systemic organ dysfunction, irrespective of local complications, such as peripancreatic fluid collection, necrosis, abscess, or pseudocyst. We thus decided to interrogate a large national intensive care patient database collected from U.K. ICUs to examine outcomes in a large cohort of patients admit-

ted with severe acute pancreatitis and to see whether an improved prognostic scoring system can be developed based on this sizable data set using physiologic and biochemical variables collected within the first 24 hrs of intensive care admission.

MATERIALS AND METHODS
The Intensive Care National Audit and Research Centre Case Mix Programme Database (CMPD) contains information on case mix, outcome, and activity for all admissions to adult, general critical care units participating in the Case Mix Programme, a national comparative audit in England, Wales, and Northern Ireland, active since 1995. The Case Mix Programme has received approval from the Patient Information Advisory Group to hold patient identifiable information without consent (approval number: PIAG 2-10(f)/2005). Approval of the study by an institutional review board was not required. The data are collected by trained data collectors according to precise rules and definitions and undergo extensive validation for completeness, logic, and consistency (10). The data available for analysis covered admissions to 159 ICUs during the time period December 1995 to June 2003, inclusive.

Crit Care Med 2007 Vol. 35, No. 7

1703

same hospital HDU. lowest arterial pH. gender.2 1.0) 3. lowest mean arterial pressure. activity.3 34. highest serum urea. and lowest platelet count. operative status. RESULTS Of a total of 219.468 intensive care admissions in the database. in-hospital 2.0 46. The variables included in the initial full model were age.5 48. highest serum glucose. Age.462 valid admissions included in the model. a Admissions staying Ն8 hrs in the ICU only. or infective pancreatitis. 1704 Crit Care Med 2007 Vol. 1.2 60. same hospital Radiograph.6 44. TX). same hospital HDU. All analyses were performed in Stata version 8. All variables were entered into a multiple logistic regression model.4–11. Finally.7 44.3 26. days Pre-ICU ICU ICU survivors ICU nonsurvivors Post-ICU Hospital survivors Hospital nonsurvivors Mortality ICU Post-ICU. same hospital Other hospital (not ICU/HDU) Other intermediate care area. and outcome data for the 2.0) 3.1 61.3% of all admissions.0 23. two were excluded because they were missing operative status. The least significant variable was removed from the model. This left a total of 2.9% of all ICU deaths.0 33.3 59. The performance of each model was assessed in the validation data.0 17 (13–22) 56.9 40. patients aged Ͻ16 yrs. we do not feel any distinction should be made between acute and infective pancreatitis. interquartile range. College Station.462 admissions are shown Table 1.2 2. APACHE.9 4. the resulting categorical variables were each entered into a logistic regression model and adjacent categories were combined if the resulting coefficients were not significantly different at the 10% level.0 6. the Intensive Care National Audit and Research Centre coding method (11).1 15.4 1. and the process was repeated until no variables remained in the model. and patients with missing operative status or outcome (mortality at ultimate discharge from an acute hospital). logistic regression was used to relate the score to a predicted probability of mortality. Admissions coded as chronic pancreatitis were included in the analyses as chronic pancreatitis alone was not considered sufficient reason for admission to ICU and so these patients were presumed to have severe acute exacerbations.7 (1. highest serum creatinine.3 43.7 46. and 73 were excluded because they were missing hospital outcome.6 44. yrs Gender Male Female Primary reason for admission Acute pancreatitis Infective pancreatitis Chronic pancreatitis Surgical status Nonsurgical Emergency surgery Elective surgery Source of admission Ward. Admissions were selected if their primary reason for admission was recorded as acute pancreatitis. 35.Reasons for admission to the ICU are recorded in the CMPD using a specially derived hierarchical coding system. other hospital APACHE II scorea Vasoactive drug treatment during first 24 hrsb No Yes Mechanical ventilation during first 24 hrs No Yes Length of stay.6 10. Calibration was assessed by the Hosmer-Lemeshow C statistic (13) and discrimination by the area under the receiver operating characteristic curve (AUC) (14).0 1 (1–3) 3. 7 . Demographic.2 26.3 41.3 51. Continuous variables were examined using generalized additive models (12) to plot the relationship between each variable and risk of death.8 9. Variables for the model were selected a priori based on an anticipated relationship with mortality from existing models for pancreatitis and expert clinical opinion. chronic pancreatitis.0 5. computed tomography. high-dependency unit. HDU. Acute Physiology and Chronic Health Evaluation. transfers from another ICU. and a best model was selected to balance parsimony against model performance.4 89.0 1. Finally. Likewise. activity. lowest serum albumin.2 (StataCorp LP.5 42.4 5.0–13. and 2. To simplify the model further.839 were admitted with pancreatitis given as the primary reason for ICU admission. lowest PaO2/FIO2 ratio.5 5. same hospital Theater and recovery.0 84.8 (1. intensive care unit. badmissions staying Ն24 hrs in the ICU only. same hospital Recovery only. The discrimination of this score was compared with the Acute Physiology and Chronic Health Evaluation (APACHE) II score (15) and to a score composed of seven of the eight modified Glasgow criteria (8) (excluding serum lactate dehydrogenase levels) that could be evaluated in the CMPD. No.5 18. and outcome data for admissions with severe acute pancreatitis Median (IQR) or % 63 (48–74) 55. 2.4) 15 (8–30) 15 (9–30) 14 (4–37) 31.3 29.9 42. CT. These plots were used to select a suitable categorization. representing 1. highest white blood cell count. The following exclusions were applied: readmissions of the same patient within the same hospital stay. CT scanner or similar. 161 were excluded because they were transferred from another ICU.462 1369 1093 2191 148 123 2074 258 130 1456 388 232 145 104 58 30 25 24 2367 983 763 1326 1131 2457 2453 1679 774 1637 1419 218 778 254 IQR. nine were excluded because they were aged Ͻ16 yrs. same hospital Emergency department.9 41.7 54. All physiologic measurements were assessed over the first 24 hrs following admission to ICU. Demographic. a further 132 were excluded because they were readmissions of the same patient within the same hospital stay.839 admissions with pancreatitis. % No.6–10. The results of the best model were converted to an integer score by multiplying all coefficients by a constant and rounding. ICU.2 10.1 59. This was fitted in a development data set of two thirds of critical care units and assessed for calibration and discrimination in a validation data set of the remaining one third of units.3 Hospital Mortality.6% of all ICU bed days. lowest total serum calcium. admissions with missing physiologic values were assigned to the category with the most similar mortality rate. Of the 2.0 41. highest total serum bilirubin.7 (1.

Predicted probability of mortality from a logistic GAM with five degrees of freedom. No. 7 from intensive care. The majority were nonsurgical patients.13 29. The predicted probability of mortality.21 21. The median length of stay in the hospital before admission to ICU was 1 day.651– 0.8272 0.670 (0.30 indicate a significant lack of fit at the 5% significance level. A total of 14 variables.7305 0. including the categorical variables gender and operative status.866) compared with 0. and total serum calcium (in order of decreasing impact). HL.90 14.804 (0. Bold type indicates the variables included in the final model and the fit of the final model in the validation sample.8411 0.04 20. Vertical lines show categorical variables entered into the logistic regression model. Compared with the other prognostic models. serum urea.50 22.908 ϩ (0.853 (0. Nearly 42% of patients died in the hospital.36.119). the Pancreatitis Outcome Prediction (POP) Score. p ϭ .28 20. Hosmer-Lemeshow. Stepwise model selection Variables Dropped None White blood cell count Serum creatinine Serum glucose Platelet count Serum bilirubin Surgical status Gender Serum albumin Total serum calcium PaO2/FIO2 ratio Mean arterial pressure Serum urea Age Arterial pH Variables 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 AUC 0. were entered into a stepwise model selection (Table 2). Generalized additive model (GAM) plots of risk of ultimate hospital mortality against each continuous variable.69 22. with 22% of admissions admitted to the ICU on the same day as their DISCUSSION Patients admitted to ICUs with severe acute pancreatitis constitute a significant 1705 .248 ϫ POP score) [1] p ϭ exp(R)/(1 ϩ exp[R]) [2] AUC. The selected model included the six variables with the strongest relationship to hospital outcome—arterial pH. in Table 1.820) for the APACHE II score.8413 0.8422 0. APACHE II model was 0. Table 2. values Ͼ18.688) for the seven available modified Glasgow criteria and 0.97 (0.8428 0. a quarter of whom died after discharge Crit Care Med 2007 Vol. ranging from 0 to 40 points (Table 3). Figure 2 depicts this relationship between the POP Score and predicted probability of mortality. Figure 3 shows the calibration of the final model. the AUC (95% confidence interval) of the final score in all admissions was 0. Figure 1 shows generalized additive model plots of the risk of ultimate hospital mortality against each of 12 continuous variables measured within the first 24 hrs of ICU admission.Figure 1. and there was no significant lack of calibration (Hosmer-Lemeshow ␹210 ϭ 15.8381 0.8415 0.06 15. The development sample consisted of 87 ICUs (n ϭ 1.92–1. mean arterial pressure. age.5000 HL ␹2a 30.838 – 0. PaO2/FIO2 ratio. area under the receiver operating characteristic curve. The discrimination of this model in the validation sample was only slightly worse than that of the full model (AUC ϭ 0. These six factors were used to develop an objectively weighted multivariate prognostic score.841). 35.8406 0.8154 0.53 22. Vertical lines divide each variable into the categories that were entered into the model.51 20. AUC and HL ␹2 calculated in validation sample of one third of units (n ϭ 968).66 25. The ratio of observed to expected deaths based on the U. Models fitted in development sample of two thirds of units (n ϭ 1494).36 14.494) and the validation sample 61 ICUs (n ϭ 968).787– 0.8406 0. admitted from general wards. is described by the following equations: R ϭ Ϫ4. p. a Hosmer-Lemeshow chi-square goodness-of-fit statistic on 10 df.8347 0.7911 0.90 — — admission to hospital and 14% of admissions having been in hospital for Ն1 wk.838 compared with 0.8399 0.02). after combining adjacent categories that were not significantly different.K.

Indeed.9) 7. Patients coded as having either infective or chronic pancreatitis comprised only 11% of the total population.8–47. population carrying a high mortality rate and often a prolonged length of stay relative to other ICU admissions. such patients were likely to have an acute exacerbation requiring ICU admission.20–7. dating back several decades.29 22. First.6 (Ն17) Ͻ7. Predicted probability of ultimate hospital mortality from the Pancreatitis Outcome Prediction Score.3–2.8–1.or two-center specialist units rather than a general hospital population and mainly consisted of relatively small numbers of general ward patients with few risk factors and a good eventual outcome. these were developed on outcome data from single. No. 17) and Glasgow (or Imrie) (7) criteria and modifications thereof (8.5) 40–49 6 7 60–69 8 Ն70 Ͻ40 10 7.0–2.2–9.29) 1 30–39 80 to 89 2 3 40–49 60–79 75–224 (10–29. nutrition. mm Hg (kPa) Arterial pH Serum urea.19 (2. mg/dL (mmol/L) 0 16–29 Ն90 Ն225 (Ն30) Ն7. and treatment of infection and other complications (16).6–1. Observed vs.24 30. Calibration of the Pancreatitis Outcome Prediction Score.9) Ͻ6.5 (11–16.19 7.99) 7. the Crit Care Med 2007 Vol.99 (1.79) OR 9.005 patients of whom only 25% fulfilled the criteria for severe acute pancreatitis (20). However.10–7. and their inclusion does not affect the results. Early prediction of the severity of an acute attack thus has important implications for management and timely intervention. mg/dL (mmol/L) Total serum calcium. thus giving a potentially biased reflection of nationwide mortality rates.6) OR Ն10 (Ն2. Given the high mortality and morbidity associated 1706 with severe acute pancreatitis. such patients are probably best treated in an intensive care/high-dependency environment with close attention paid to volume resuscitation. These tools fall into four broad categories. 7 .0–9. predicted mortality in ten equal-sized groups formed by deciles of predicted mortality.99 (2. 35.2–7.19 (1.4–7.30–7.49) 4 5 50–59 50–59 Ͻ75 (Ͻ10) 7. yrs Mean arterial pressure.3 (5–7. the largest cohort was based on 1. A variety of assessment tools have been used to identify patients with acute pancreatitis at risk of developing complications and/or death. there are clinico-biochemical tools specific to acute pancreatitis. As discussed earlier. mm Hg PaO2/FIO2 ratio.35 Ͻ14 (Ͻ5) 8. Finally.9) 6.25–7. Figure 3.35 14–22. and we thus feel it is reasonable to include them in the analysis.Table 3. this difference in mortality was fully explained by the POP Score.9) 7.7 (8–10. gas exchange.00 Figure 2. these outcome data are largely historical. 18 –20). While mortality was lower for admissions coded as chronic pancreatitis.00–7. such as the Ranson (6. and do not necessarily reflect contemporary patient management and outcomes.4–30.4 (Ͻ1. Pancreatitis Outcome Prediction Score Score Age. The increased expertise in dealing with the disease in a specialist center may be offset by a high number of sick tertiary referral patients.09 Ն47. In addition.

serum lactate dehydrogenase as was included in the modified Glasgow criteria).” Despite this. those with low scores (Յ13) who eventually died did so after an average ICU stay of 9 days. (39) recently reported that only one half Crit Care Med 2007 Vol. which gives a nonsurgical coefficient of 0. we previously reported ICU and hospital mortality rates for the CMPD of 20.g.658 and a surgical coefficient of 0. The median stay for the pancreatitis patients of 3. The APACHE II system shows similar sensitivity and specificity rates using data collected over the first 24 hrs (29). pancreatitis did not warrant a separate diagnostic coding in the original APACHE II system. Surg Gynecol Obstet 1973. A recent single-center study claimed superiority of the Balthazar computed tomography severity index over the Ranson and APACHE II scores in predicting outcome from acute pancreatitis (25). respectively (38). this new score needs to be validated prospectively and. which was developed by expert panel consensus and uses all-encompassing criteria (local complications. It was therefore not possible to validate the diagnosis of pancreatitis (e. lactate dehydrogenase. as the senior author recently acknowledged. systemic complications. the data set was not collected for the purpose of identifying patients with pancreatitis. The presence of three or more criteria in either the Ranson or Glasgow scoring systems or an APACHE II score Ն8 has been used to indicate severe acute pancreatitis (28. the 1992 Atlanta classification (4). 322:595–598 2. are also missing.6% and 41. may. et al: 1707 . and these patients were included within a diagnostic category of “gastrointestinal infection. This was more sensitive than existing scoring systems for which most criteria were accessible in our data set. The ideal cutoff score also remains to be determined. Our database does not enable this to be performed.K. generic intensive care scoring systems. Daggett WM. REFERENCES 1.. on first sight.442 for gastrointestinal infection. Fourth. However. the inability to offer routine interleukin-6 measurement precludes its introduction at present into clinical practice. We also lack data collected at ICU admission for these patients to test the validity of the POP Score at this earlier time point. perhaps.5 (10). or death). Third.g.6%. Yet. but a deteriorating score at 48 hrs had greater prognostic value (29). Beckingham IJ. The utility of data taken on ICU admission and. with a mean first 24-hr APACHE II score of 16. has been proposed. These high values reflect the high mortality rate of this condition relative to many others. Other pertinent details. We have applied the U. 35. although a severe episode was unlikely with scores Ͻ8 (negative predictive value of 89%). these modified scores have many of the same drawbacks of the pancreatitis-specific scores. respectively. based on blood urea nitrogen. and so these patients were weighted in the general “gastrointestinal” category with a nonsurgical coefficient of 0. ICUs with a primary diagnosis of severe acute pancreatitis enabled the development of a new scoring system based on six readily available physiologic and biochemical variables collected within the first 24 hrs of ICU admission.613.0 days (nonsurvivors) for all ICU patients in the CMPD (10). but it is more cumbersome to perform (30). such as the etiology of the pancreatitis and comorbidities such as diabetes and nonsevere heart failure.3% and 28. DuPont HV: Infectious complications of acute pancreatitis. and interleukin-6 levels taken on samples at admission to an international phase III trial of the plateletactivating factor receptor-antagonist Lexipafant (22). prognostication has been proposed using computed tomographic examination and based on a five-grade severity score (23) or the presence and degree of pancreatic necrosis (24). The ICU and hospital mortality rates of 31. BMJ 2001.9%.. Second. Venkatesan et al.Ranson and Glasgow scores require data collected at 48 hrs after admission. be removed and patients often deteriorate further despite intensive care. However.7 days (survivors) and 3. ACKNOWLEDGMENTS We thank all those intensive care units contributing data to the Intensive Care National Audit and Research Centre Case Mix Programme Database. to provide an even earlier indication of severity and prognosis. APACHE II model (40). There was a biphasic distribution in stay for the pancreatitis patients. age.501 and a surgical coefficient of Ϫ0. for example by incorporating obesity (33). Sensitivity and specificity for predicting mortality have been reported as 63% and 76% for Ranson and 72% and 84% for the modified Glasgow criteria.7 days (survivors) and 2. The recently published BALI score (21). suggesting continuing physiologic deterioration in this initially lower severity subset (41) or late-onset complications perhaps related to infected necrosis (42). particularly as the intense inflammatory stimulus underlying severe acute pancreatitis cannot. presumably from their multiple organ failure. there were still not sufficient cases for this diagnostic category to receive a coefficient. However. Clearly. 35–37). ideally. need for surgery. Bornman PC: Acute pancreatitis. population. provided prognostic ability similar to Ranson. on hospital admission also needs to be compared against the 24-hr data reported herein. Nonsurvivors with high first 24-hr APACHE II scores (Ն23) died soon after admission (median 2 days). in general. based on amylase levels or computed tomography scans) or to evaluate all potential prognostic variables (e. 136:763–768 3. Both the APACHE II and the Simplified Acute Physiology Score systems were of limited clinical utility in the early prognostic evaluation of acute pancreatitis (30). Glasgow.K. By comparison. CONCLUSIONS Our retrospective analysis of a large database of patients admitted to U. such as the APACHE II (15) and III (26) scores and Simplified Acute Physiology Score (27). However. have been applied to cohorts of pancreatitis patients (28 –32). Jacobs ML. However.K. No. It will also considerably facilitate prospective studies and comparisons of performance. Kodesch R. in a non-U. being based on small data sets from single centers. respectively. Civetta JM. and APACHE II scores.8 days (nonsurvivors) compares with 1. In terms of comparability. The converse was seen with survivors. A score with good prognostic ability provides the possibility of identifying those patients who are more likely to benefit from new therapeutic interventions and management regimens. 7 of their cases classified as severe pancreatitis by the Atlanta system could be correctly predicted by an APACHE II score Ն8. The major strength of our study is the availability of a large data set from a large and representative sample of general ICUs in the United Kingdom. no attempts have been made at large-scale objective validation (34). appear high for a median APACHE II score of 17. Generic scores have also been modified for the prediction of severe acute pancreatitis.

Rabitti PG. Cho Y. 143:29 –36 Knaus WA. Br J Anaesth 2001. Ann Surg 1977. Risk prediction of hospital mortality for critically ill hospitalized adults. et al: Prospective multicentre survey on acute pancreatitis in Italy (ProInf-AISP): Results on 1005 patients. 8:R99 –R111 Young JD. Ann Surg 2006. et al: Prognostic factors in acute pancreatitis. Manes G. Ferguson JC. Naidich DP. 36. 33. Pancreas 1993. et al: Acute pancreatitis: Value of CT in establishing prognosis. Chest 1991. outcome and length of stay for admissions to adult. Lemeshow S: Goodness of fit tests for the multiple logistic regression model. Roses DF. 1992 Flint R. Lin SY. et al: The APACHE III prognostic system. Wagner DP. Pancreatology 2004. Crit Care 2004. Barie PS: Severity scoring for prognostication in patients with severe acute pancreatitis: Comparative analysis of the Ranson score and the APACHE III score. Br J Surg 1974. 174: 331–336 25. Johnson CD. Commun Stat 1980. Wales and Northern Ireland: The Intensive Care National Audit & Research Centre Case Mix Programme Database. BMJ 1998. Lancet 1989. A9:1043–1069 Hanley JA. 25:331–335 Johnson CD. 139:69 – 81 Imrie CW. UK. Williamson RC. Draper EA. et al: Acute pancreatitis: Prognostic value of CT. Knaus WA. Arch Surg 2002. 34. 41. Imrie CW. Frulloni L. randomised. et al: Double blind. Lee KK. Curtis JR. in the treatment and prevention of organ failure in predicted severe acute pancreatitis. et al: High early mortality rate from acute pancreatitis in Scotland. Barcia AM. 8. 35. 9. APACHE II and APACHE III scoring systems in acute pancreatitis. Robinson DL. 16. Ulrich CD II. placebo controlled study of a platelet activating factor antagonist. Radiology 1985. et al: Comparison of Ranson. Imrie CW. 16:1– 84 Mergener K. Acute pancreatitis: Analysis of factors influencing survival. Arch Surg 2004. general critical care units in England. 31. Curr Prob Surg 1979. Pasternack BS: Statistical method for quantifying the severity of clinical acute pancreatitis. Chapman & Hall. 91:89 –94 17. McNeil BJ: The meaning and use of the area under the receiver operating characteristics (ROC) curve. 86:1302–1306 Ranson JHC. Vlachonikolis IG. Baillie J: Acute pancreatitis. et al: Prediction of severity in acute pancreatitis: Prospective comparison of three prognostic indices. 22: 79 –91 18. 137:1136 –1140 Dominguez-Munoz JE.4. et al: Outcome and quality of life of patients with acute pancreatitis requiring intensive care. 8:682– 686 Eachempati SR. 4:1– 6 Bradley EL III: Atlanta redux. Goldfrad C. J Surg Res 1977. Roussomoustakaki M. Rifkind KM. 19:849 – 861 Venkatesan T. et al: A simplified acute physiology score for ICU patients. et al: Management of the critically ill patient with severe acute pancreatitis. No. Dig Liver Dis 2004. 48:62– 69 23. Crit Care Med 1984. 128:586 –590 McKay CJ. Rowan K: Case mix. Khan AA. Windsor JA: Early physiological response to intensive care as a clinically relevant approach to predicting the outcome in severe acute pancreatitis. Br J Surg 1981. New York. 38. Ranson JH. Balthazar EJ. 156:767–772 24. Br J Surg 1999. Megibow AJ. Le Gall JR. 61:539 –544 Ranson JHC: Acute pancreatitis. et al: Evaluation of the clinical usefulness of APACHE II and SAPS systems in the initial prognostic classification of acute pancreatitis: A multicenter study. Arch Surg 1993. Spitzer AL. Loirat P. Pancreas 2003. 15. Oxford. lexipafant. 185:43–51 Bradley EL III: A clinically based classification system for acute pancreatitis. Benjamin IS. 11:6049 – 6052 26. 13. 36:205–211 21. Pancreas 2002. Carter DC: Biliary surgery in the same admission for gallstone associated acute pancreatitis. Toh SK. Pancreas 2003. 25:1340 –1346 Nathens AB. Sinclair M. Kingsnorth AN. Evans S. 7 . 32. 139:438 – 443 Carnovale A. University of Oxford. Imrie CW. Campbell MJ: Combination of APACHE-II score and an obesity score (APACHE-O) for the prediction of severe acute pancreatitis. Cavallini G. 26:105–106 Imrie CW: Observations on acute pancreatitis. 5. Radiology 1990. 37. Gastroenterol Clin North Am 1990. 2:403– 407 20. 12:975–977 28. Moulton JS. Rowan K: Development and testing of a hierarchical method to code the reason for admission to intensive care units: The ICNARC Coding Method. 48 hours after hospital admission compared with the APACHE II score at admission. 7. 316:44 – 48 Steinberg WM: Predictors of severity of acute pancreatitis. 87:543–548 Hastie TJ. 42. Surg Gynecol Obstet 1974. 6:438 – 444 1708 Crit Care Med 2007 Vol. 1984 –1995. Schell MT. Cooper MJ. Draper EA. Osborne DH. Wagner DP. Crit Care Med 1985. Alperovitch A. 65:337–341 Blamey SL. 2:201–204 29. Ranson JH. et al: Prevalence and predictors of severity as defined by Atlanta criteria among patients presenting with acute pancreatitis. O’Neill J. et al: Mortality in acute pancreatitis: Is it an early or a late event? JOP 2005. 26:107–110 Rowan KM: Outcome comparisons of intensive care units in Great Britain and Ireland using the APACHE II method [DPhil thesis]. Leung TK. Lee CM. 243:380 –388 22. Hydo LJ. J Surg Res 2000. 100: 1619 –1636 27. Brady AR. Garcia MJ. 35. et al: A single centre double blind trial of trasylol therapy in primary acute pancreatitis. Bassi C. Lancet 1985. 14. 12. et al: Prognostic signs and the role of operative management in acute pancreatitis. 13: 818 – 829 Soran A. 32:2524 –2536 Harrison DA. 68:758 –761 19. Beale RJ. et al: Balthazar computed tomography severity index is superior to Ranson criteria and APACHE II scoring system in predicting acute pancreatitis outcome. Tibshirani R: Generalized Additive Models. et al: Improved prediction of outcome in patients with severe acute pancreatitis by the APACHE II score at 30. Balthazar EJ. Gut 2001. et al: Applying Ockham’s razor to pancreatitis prognostication: A four-variable predictive model. 137:730 –736 Chatzicostas C. 6. et al: APACHE II: A severity of disease classification system. Larvin M. World J Gastroenterol 2005. Parekh D. Crit Care Med 2004. 39. 1990 Hosmer DW. 10. Corfield AP. Radiology 1982. Br J Surg 1978. 11. McMahon MJ: APACHE-II score for assessment and monitoring of acute pancreatitis. Gut 1984. Chelluri L. Carballo F. 40. Arch Surg 2002.