Scandinavian Journal of Gastroenterology. 2012; 47: 28–35

ORIGINAL ARTICLE

Accuracy of clinical prediction rules in peptic ulcer perforation: an observational study

DAVID LEVARETT BUCK1, MORTEN VESTER-ANDERSEN2 & MORTEN HYLANDER MØLLER3

1Emergency Department, Holbæk Hospital, Denmark, 2Department of Anaesthesiology and Intensive Care Medicine, Copenhagen University Hospital Herlev, Denmark, and 3Department of Anaesthesiology and Intensive Care Medicine, Copenhagen University Hospital Bispebjerg, Denmark

Abstract Objective. The aim of the present study was to compare the ability of four clinical prediction rules to predict adverse outcome in perforated peptic ulcer (PPU): the Boey score, the American Society of Anesthesiologists (ASA) score, the Acute Physiology and Chronic Health Evaluation (APACHE) II score, and the sepsis score. Material and methods. Design: an observational multicenter study. Participants and settings: a total of 117 patients surgically treated for PPU between 1 January 2008 and 31 December 2009 in seven gastrointestinal departments in Denmark were included. Pregnant and breastfeeding women, non- surgically treated patients, patients with malignant ulcers, and patients with perforation of other organs were excluded. Primary outcome measure: 30-day mortality rate. Statistical analysis: the ability of four clinical prediction rules to distinguish survivors from non-survivors (discrimination ability) was evaluated by the area under the receiver operating characteristic curve (AUC), positive predictive values (PPVs), negative predictive values (NPVs), and adjusted relative risks. Results. Median age (range) For personal use only. was 70 years (25–92 years), 51% of the patients were females, and 73% of the patients had at least one co-existing disease. The 30-day mortality proportion was 17% (20/117). The AUCs: the Boey score, 0.63; the sepsis score, 0.69; the ASA score, 0.73; and the APACHE II score, 0.76. Overall, the PPVs of all four prediction rules were low and the NPVs high. Conclusions. The Boey score, the ASA score, the APACHE II score, and the sepsis score predict mortality poorly in patients with PPU.

Key Words: APACHE II score, ASA score, Boey score, clinical prediction rule, perforated peptic ulcer, prognosis, sepsis

Introduction shock, and perforation longer than 24 h [13]. In the original study by Boey et al., the in-hospital mortality Mortality and morbidity following perforated peptic proportion increased progressively with the number of Scand J Gastroenterol Downloaded from informahealthcare.com by Danmarks Natur-og on 01/08/12 ulcer (PPU) is substantial, and mortality proportions prognostic variables, being 0%, 10%, 45.5%, and of 25–30% have been reported in population- 100% in patients with none, one, two, or all three based studies [1–6]. variables, respectively. A large number of prognostic factors for morbidity The most commonly used preoperative clinical and mortality following PPU have been characterized prediction rule worldwide is the American Society [7] and a number of clinical prediction rules proposed of Anesthesiologists (ASA) score [14,15]. In the ASA [8–12]. scoring system, the patient’s preoperative health The most well-known prediction rule in PPU is the status, independent of the current surgical disease, Boey score which seeks to predict mortality based on is graded in five categories: (1) normal health, (2) the presence of major medical illness, preoperative mild systemic disease, (3) severe systemic disease, (4)

Correspondence: Morten Hylander Møller, MD PhD, Department of Anaesthesiology and Intensive Care Medicine, Copenhagen University Hospital Bispebjerg, Bispebjerg Bakke 23, DK–2400 Copenhagen NV, Denmark. Tel: +45 22555343. E-mail: [email protected]

(Received 30 May 2011; revised 31 October 2011; accepted 2 November 2011) ISSN 0036-5521 print/ISSN 1502-7708 online 2012 Informa Healthcare DOI: 10.3109/00365521.2011.639078 Accuracy of clinical prediction rules in PPU 29

severe systemic disease that is a constant threat to life, Data extraction and management and (5) the patient is not expected to survive without surgery [14]. The primary data extraction was performed by the Another commonly used perioperative prediction local principal investigators at each center using a rule is the Acute Physiology and Chronic Health standardized case report form. Data were subse- Evaluation (APACHE) II score, where the predicted quently validated and transferred to an electronic mortality risk is based on the sum of points from database by MHM. 12 physiologic measures, age, and health status [16]. The score has been sought to be validated in abdom- Protocol inal surgery in general [17], peritonitis in general [18], and PPU [11,19]. Risk stratification according to the four scores (the In a recently published paper, 30-day mortality in Boey score, the ASA score, the APACHE II score, PPU patients was significantly reduced by implement- and the sepsis score) was done when booking ing a perioperative care protocol, based on diagnos- for surgery was performed (preoperatively). Follow- tics, monitoring, and treatment according to the up time was 30 days. The included patients were part Surviving Sepsis Campaign [20]. Sepsis score [21] of a clinical controlled multicenter intervention trial, on admission was included in the protocol. recently published [20]. At present, clinical prediction rules are not rou- tinely used in PPU patients in everyday clinical prac- Outcome measures tice. Accurate and early identification of PPU patients with an increased risk of adverse outcome is needed to The primary outcome measure was 30-day mortality. plan and target the level of perioperative monitoring Secondary outcome measures were: (1) development and treatment. The aim of the present study was to of septic shock (according to the 2001 International compare the performance of four clinical prediction Sepsis Conference [21]) within 30 days of surgery and rules in PPU: the Boey score, the ASA score, the (2) postoperative admission to the intensive care unit APACHE II score, and the sepsis score. We hypoth- (ICU) within 30 days of surgery. esized that existing clinical prediction rules are poor predictors of adverse outcome in PPU. Sample size

For personal use only. Material and methods Receiver operating characteristics curve: we expected a discriminatory ability (area under the receiver Design and approval operating characteristic curve; AUC) of 0.75. Based on 80% power to detect a significant association This observational multicenter study was approved = by The Danish Data Protection Agency (no 2007- (p 0.05, two-sided) between the clinical prediction 41-0702) and did not require informed patient con- rule and the primary outcome, 80 patients were sent according to Danish law. The manuscript was required in the study [23]. prepared according to the STROBE statement [22]. Multiple logistic regression: with a binary response variable, nine covariates, ß = 0.80, a=0.05, and an anticipated medium effect size, 113 patients were

Scand J Gastroenterol Downloaded from informahealthcare.com by Danmarks Natur-og on 01/08/12 Participants and settings required in order to detect an association between the variables and endpoints [24,25]. A total of 117 consecutive patients surgically treated for PPU between 1 January 2008 and 31 December 2009 in seven gastrointestinal departments in Den- Statistical analysis mark were included. The participating hospitals were: Copenhagen University Hospital Herlev, Copenha- Data were analyzed using the statistical software gen University Hospital Bispebjerg, Copenhagen packages SPSS 19.0. University Hospital Hillerød, Copenhagen University The ability of the four clinical prediction rules to Hospital Helsingør, Aarhus University Hospital, distinguish survivors from non-survivors (discrimina- Odense University Hospital, and Kolding Hospital. tion ability) was evaluated by the AUC. The optimal Only patients with intraoperatively confirmed gastric AUC is close to 1 [23]. or duodenal PPU were included. Pregnant and The risk of dying within 30 days of surgery, given breastfeeding women, non-surgically treated patients, a positive test result (a score above a certain thresh- patients with malignant ulcers, and patients with old), is reported as the positive predictive value perforation of other organs were excluded. (PPV). Correspondingly, the risk of surviving within 30 D. L. Buck et al.

30 days of surgery, given a negative test result (a score patients were females, and 73% of the patients had at below a certain threshold), is reported as the negative least one of six co-morbid diseases. Fifty-three per- predictive value (NPV). The optimal PPV and NPV cent of the patients were categorized as ASA class ‡3. are close to 100%. The optimal cut-off value of the Alcohol abuse was present in 26% of the patients, and clinical prediction rules was calculated by means of 50% of the patients smoked daily (Table I). the Youden index (sensitivity + specificity – 1) [26]. Perioperative characteristics of the 117 patients are Based on existing literature, and on the Youden shown in Table II. indices, the following thresholds were chosen: Boey The 30-day mortality proportion was 17% (20/ score ‡2, ASA score ‡3, APACHE II score ‡12, and 117). In all, 26% (30/117) were diagnosed with septic sepsis score ‡3. Finally, adjusted estimates of the four shock [20], and 49 patients (42%) were admitted to prediction rules’ association to the outcome measures the ICU. are presented as relative risks (RRs) with 95% con- The AUCs of the four clinical prediction rules – fidence intervals (95% CIs). Adjustment for the fol- with respect to the primary outcome 30-day lowing confounders was done: age ‡65 years, gender, mortality – were as follows: the Boey score 0.63; co-morbidity, alcohol abuse, preoperative delay the sepsis score 0.69; the ASA score 0.73; and the >24 h, and smoking. Definitions of the variables are APACHE II score 0.76 (Figure 1). With respect to the found in Table I. secondary endpoint septic shock, the APACHE II Missingdataoccurredinlessthan5%ofthepatients. score and the sepsis score had the highest AUC The prevalence and pattern of missing values were (Figure 2), whereas in relation to ICU admission, evaluated; data were not “missing completely at the ASA score and the APACHE II score had the random” (MCAR). Consequently, multiple imputation highest AUC (Figure 3). for the missing values was performed [27,28]. Five Overall, the PPVs of all four prediction rules imputations were made, and the five datasets were were low: 30-day mortality (24–41%), septic shock analyzed using Markov Chain Monte Carlo fully con- (35–72%), and ICU admission (49–80%) (Table III). ditional specification [27], and the estimates were The NPVs with respect to 30-day mortality were on pooled for overall estimation. Furthermore, complete- the other hand high: ASA score ‡3, 93%; APACHE II case analysis was performed. The pooled imputed esti- score ‡12, 97%; Boey score ‡2, 87%; and sepsis score mates and the corresponding complete-case estimates ‡3, 90% (Table III). For septic shock and ICU were compared, and if there were any noticeable differ- admission, the NPVs were lower (Table III). For personal use only. ences, both results are presented. Otherwise, only the For 30-day mortality, the APACHE II score and pooled imputed estimates are presented. the ASA score had the highest RRs, 31.6 (1.8–545.2) and 21.5 (3.1–150.0), respectively. The sepsis score Results had the strongest association with septic shock: RR 14.6 (4.2–50.2). Only ASA score and sepsis score fi A total of 117 patients were included [20]. Median were signi cantly associated with ICU admission age (range) was 70 years (25–92 years), 51% of the (Table III).

Table I. Demographic characteristics. Discussion Characteristic n = 117 Scand J Gastroenterol Downloaded from informahealthcare.com by Danmarks Natur-og on 01/08/12 In this observational multicenter study in patients Age, median (range) 70 (25–92) Men, no (%) 57 (49) surgically treated for PPU, we found poor discrimi- Body mass index, median (range) 24 (15–65) natory abilities of all four evaluated clinical prediction Co-morbidity1, no (%) 85 (73) rules: the Boey score, the ASA score, the APACHE II ASA class2, no (%) score, and the sepsis score. This was the case for all I 12 (10) outcome measures. II 43 (37) III 45 (38) During the last decade, fast-track surgery and evi- IV 15 (13) dence based in-hospital care have been sought to be V 2 (2) implemented in various fields of surgery, leading to a Daily smoking, no (%) 59 (50) reduction in morbidity and mortality [29]. In patients 3 Alcohol abuse , no (%) 30 (26) with PPU, mortality has been reduced considerably 1One or more of the following diseases: diabetes, chronic obstruc- by the implementation of a standardized evidence tive pulmonary disease, heart disease, liver cirrhosis, active cancer based in-hospital care protocol [20]. The limited disease, AIDS or other chronically treated disease. number of ICU and high dependency unit beds 2American Society of Anesthesiologists class. 3More than 36 g/day (men) or 24 g/day (women). emphasizes the importance of individual risk Accuracy of clinical prediction rules in PPU 31

Table II. Perioperative characteristics. Characteristic (n = 117)

Time to admission, median (range), hours 10 (1–336) Time from admission to surgery, median (range), hours 5 (1–72) Duration of surgery, median (range), minutes 83 (30–240) Surgical procedure, no (%) Simple open closure ± omental patch 96 (83) Laparoscopic closure 12 (10) Resection 8 (7) Gastric site of perforation, no (%) 42 (36) Preoperative metabolic acidosis, no (%) 56 (48) Preoperative hemoglobin <6.0 mmol/L, no (%) 10 (8) Preoperative creatinine >110/130 mmol/L, no (%) 33 (28) Preoperative albumin <550 mmol/L, no (%) 56 (48) Use of NSAIDs at admission, no (%) 44 (38) Use of aspirin at admission, no (%) 36 (31) Use of steroids at admission, no (%) 17 (15) Use of anticoagulants at admission, no (%) 7 (6) Shock on admission*, no (%) 28 (24)

Abbreviation: NSAID = non-steroidal anti-inflammatory drug. *Blood pressure <90 mmHg and heart rate >100 per min.

stratification [30]. Early and accurate identification of Boey score, in-hospital mortality increases progres- patients with increased risk of adverse outcome is sively with the presence of major medical illness, needed to plan and target the level of perioperative preoperative shock, and perforation longer than monitoring and treatment. Thus, a clinical scoring 24 h [13]. The Boey score has been re-evaluated in system should be able to predict adverse outcome with a number of single center studies [3,10,32–35], a high degree of precision. Furthermore, the score though without being able to fully replicate the con- should be easy to calculate, preferably bedside. These vincing results found by Boey et al. This is both the characteristics have proven difficult to realize [31]. case for the PPV, the NPV, and the discriminatory

For personal use only. In the present study, the Boey score had the poorest function (AUC). When we apply the same cut-off discriminatory ability of survival, with an AUC of value as used in the present study (Boey score ‡2) 0.63. Furthermore, the PPV and NPV were low, to the original work of Boey et al. [13,] this results in a and the RR was statistically insignificant. In the substantially lower PPV and NPV, comparable to the

1.0 The ASA score (AUC 73%) The APACHE II score (AUC 76%) The sepsis score (AUC 69%) 0.8 The Boey score (AUC 63%) Reference line Scand J Gastroenterol Downloaded from informahealthcare.com by Danmarks Natur-og on 01/08/12

0.6

Sensitivity 0.4

0.2

0.0 0.0 0.2 0.4 0.6 0.8 1.0 1 – Specificity

Figure 1. Receiver operating characteristics (ROC) curves with area under the curve (AUC). 30-day mortality. 32 D. L. Buck et al.

1.0 The ASA score (AUC 67%) The APACHE II score (AUC 78%) The sepsis score (AUC 74%) 0.8 The Boey score (AUC 72%) Reference line

0.6

Sensitivity 0.4

0.2

0.0 0.0 0.2 0.4 0.6 0.8 1.0 1 – Specificity

Figure 2. Receiver operating characteristics (ROC) curves with area under the curve (AUC). Septic shock.

previously mentioned studies [32–35]. The Boey The ASA score has been criticized for its subjec- score is crude, consisting of only three parameters. tivity and the wide inter-observer variability [36]. Consequently, it does not include many of the other Nevertheless, it predicts postoperative mortality existing and well-examined prognostic factors for well [37]. In the present study, the AUC of ASA adverse outcome in PPU, e.g. old age, tachycardia, class was 0.73 (mortality), 0.67 (septic shock), and

For personal use only. and acute renal failure [7]. This might explain the 0.69 (ICU). The only other study examining the inferior performance in the present study. On the discriminatory ability of ASA score in PPU patients other hand, it is simple to calculate and was created found a somewhat higher AUC of 0.91 [34]. This specifically for patients with PPU. discrepancy is most likely caused by differences in

1.0 The ASA score (AUC 69%) The APACHE II score (AUC 72%) The sepsis score (AUC 64%) 0.8 The Boey score (AUC 64%) Reference line Scand J Gastroenterol Downloaded from informahealthcare.com by Danmarks Natur-og on 01/08/12

0.6

Sensitivity 0.4

0.2

0.0 0.0 0.2 0.4 0.6 0.8 1.0 1 – Specificity

Figure 3. Receiver operating characteristics (ROC) curves with area under the curve (AUC). Intensive care unit (ICU) admission. Accuracy of clinical prediction rules in PPU 33

Table III. Positive predictive values (PPVs), negative predictive values (NPVs), and relative risks (RRs) with 95% confidence intervals (CIs) of the four clinical prediction rules (n = 117). 30-day mortality Septic shock ICU admission

Score PPV NPV RR* (95% CI) PV NPV RR* (95% CI) PPV NPV RR* (95% CI) ASA score ‡3 26% 93% 21.5 (3.1–150.0) 35% 85% 5.6 (1.5–20.6) 56% 74% 4.5 (1.6–12.8) APACHE II score ‡12 24% 97% 31.6 (1.8–545.2) 35% 94% 10.0 (1.4–69.4) 49% 75% 2.7 (0.8–9.5) Boey score ‡2 24% 87% 6.4 (1.2–34.7) 40% 82% 5.1 (1.2–21.9) 55% 65% 2.3 (0.7–8.2) Sepsis score ‡3 41% 90% 7.7 (2.1–28.0) 72% 88% 14.6 (4.2–50.2) 80% 69% 10.2 (2.6–39.7)

Abbreviations: APACHE = Acute Physiology and Chronic Health Evaluation; ASA = American Society of Anesthesiologists; ICU = intensive care unit. *Adjusted for age ‡65 years, yes/no; male gender, yes/no; co-morbidity, yes/no; alcohol abuse, yes/no; preoperative delay >24 h, yes/no; and smoking, yes/no.

demographic characteristics and in the mortality preoperative delay, co-existing use of medications, rate. The PPVs and NPVs of other studies evaluating and low serum albumin blood level [7]. ASA score are of similar strengths as in our study It is well known that sepsis (systemic inflammatory [35,38–40]. The explanation for the inferior perfor- response syndrome and proven or highly suspected mance of the ASA score could be the fact that ASA infection) [41] is associated with adverse outcome in score is not developed within PPU patients and only PPU patients [7]. To our knowledge, the discrimina- accounts for the pre-morbid status. However, in our tory ability of the sepsis score in PPU has not been experience it is difficult to completely ignore the examined previously. In the present study, the sepsis current clinical condition of the patient when asses- score had low AUCs: 30-day mortality 0.69, septic sing the ASA score. Consequently, the performance shock 0.74, and ICU admission 0.64. Furthermore, we of the ASA score could in fact be overestimated. found low PPVs and NPVs. This might be explained Of the four clinical prediction rules evaluated, by the fact that the sepsis score is not specific for PPU the APACHE II score performed best overall with patients, it does not include co-existing diseases, and it AUCs of 0.76 (mortality), 0.78 (septic shock), and only takes into account few of the other known prog-

For personal use only. 0.72 (ICU). An APACHE II score ‡12 seems to imply nostic factors for PPU [7]. a nearly 32 times increased risk of 30-day mortality. However, only 24% of the patients predicted to die by Limitations the APACHE II score died (PPV 24%). On the other hand, all but four patients predicted to survive This study has limitations. The cohort was part of an (APACHE II score <12) survived (NPV 97%). Use interventional study [20], which potentially could of APACHE II score as a clinical prediction rule in affect the external validity of the study. Hence, PPU patients has been sparsely examined [8,10,11]. external or internal (boot-strapping) validation of Overall, PPVs and NPVs are of similar strengths as in the present results is warranted. The risk of con- the present study. The only other study examining founding was sought to be minimized by the use of the discriminatory ability of APACHE II score logistic regression, with adjustment for previously Scand J Gastroenterol Downloaded from informahealthcare.com by Danmarks Natur-og on 01/08/12 found a somewhat higher AUC of 0.87 [11]. This identified prognostic factors and suspected con- discrepancy is most likely caused by differences in founding factors (multivariate technique). However, demographic characteristics and in the mortality rate. one major problem is that confounding variables are The APACHE II score takes into account age, co- not always known or measurable (residual confound- existing diseases, the current clinical status, and cur- ing). The calculated 95% CIs are wide, indicating rent biochemical findings: in total, 14 variables [16]. low statistical precision. Choosing other definitions This is most likely the main reason why the APACHE and categorizations of the demographic variables and II score performs best among the four clinical pre- confounders, and choosing other cut-off values of diction rules evaluated in the present study. One the clinical prediction rules could lead to other possible explanation for the somewhat inadequate results. Despite using the preferred method of han- performance of the APACHE II score in general dling missing values that are not MCAR (multiple could be the fact that the score is not developed imputation) [27,28], missing values could affect within a PPU cohort. Thus, well-established impor- the validity of the results. The PPU patients in the tant prognostic factors for mortality in PPU patients present cohort have one particular type of case mix. are not included in the APACHE II score: The results of this study should not be applied 34 D. L. Buck et al.

uncritically to other categories of patients and/or to [11] Koc M, Yoldas O, Kilic YA, Gocmen E, Ertan T, PPU patients with different characteristics. Dizen H, et al. Comparison and validation of scoring systems in a cohort of patients treated for perforated peptic ulcer. Langenbecks Arch Surg 2007;392:581–5. Conclusion, clinical implications and perspectives [12] Egberts JH, Summa B, Schulz U, Schafmayer C, Hinz S, Tepel J. Impact of preoperative physiological risk profile on postoperative morbidity and mortality after emergency oper- In patients surgically treated for PPU, the Boey ation of complicated peptic ulcer disease. World J Surg 2007; score, the ASA score, the APACHE II score, and 31:1449–57. the sepsis score predict mortality, septic shock, and [13] Boey J, Choi SK, Poon A, Alagaratnam TT. Risk stratifica- admission to ICU poorly. The AUCs and PPVs were tion in perforated duodenal ulcers. A prospective validation – overall low, while the NPVs were 90–95%. Thus, in of predictive factors. Ann Surg 1987;205:22 6. [14] ASA. New classification of physical status. Anesthesiology the clinical setting, the scores can rule out mortality 1963;24:111. within 30 days of surgery with a high degree of [15] Saklad M. Grading of patients for surgical procedures. Anes- precision in a PPU cohort with a similar case mix. thesiology 1941;2:281–4. In order to be able to precisely predict adverse [16] Knaus WA, Draper EA, Wagner DP, Zimmerman JE. fi outcome in PPU patients, we suggest that a score APACHE II: a severity of disease classi cation system. Crit Care Med 1985;13:818–29. developed within PPU patients and including both [17] Wacha H, Hau T, Dittmer R, Ohmann C. Risk factors pre-morbid objective measures and current objective associated with intraabdominal infections: a prospective mul- measures is used. ticenter study. Peritonitis Study Group. Langenbecks Arch Surg 1999;384:24–32. Declaration of interest: The authors report no [18] Ohmann C, Wittmann DH, Wacha H. Prospective evalua- fl tion of prognostic scoring systems in peritonitis. Peritonitis con icts of interest. The authors alone are responsible Study Group. Eur J Surg 1993;159:267–74. for the content and writing of the paper. [19] Schein M, D’Egidio A. APACHE II scoring in surgical upper gastrointestinal emergencies. Dig Dis 1991;9:156–64. [20] Moller MH, Adamsen S, Thomsen RW, Moller AM; References and the PULP trial group. Multicentre trial of a perioperative protocol to reduce mortality in patients with peptic ulcer [1] Blomgren LG. Perforated peptic ulcer: long-term results after perforation. Br J Surg 2011;98:802–10. simple closure in the elderly. World J Surg 1997;21:412–14; [21] Levy MM, Fink MP, Marshall JC, Abraham E, Angus D, discussion 4–5. Cook D, et al. 2001 SCCM/ESICM/ACCP/ATS/SIS Inter- [2] Moller MH, Shah K, Bendix J, Jensen AG, Zimmermann- national Sepsis Definitions Conference. Crit Care Med 2003; For personal use only. Nielsen E, Adamsen S, et al. Risk factors in patients surgically 31:1250–6. treated for peptic ulcer perforation. Scand J Gastroenterol [22] von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, 2009;44:145–52. Vandenbroucke JP. Strengthening the Reporting of Obser- [3] Irvin TT. Mortality and perforated peptic ulcer: a case for risk vational Studies in Epidemiology (STROBE) statement: stratification in elderly patients. Br J Surg 1989;76:215–18. guidelines for reporting observational studies. BMJ 2007; [4] Christiansen C, Christensen S, Riis A, Thomsen RW, 335:806–8. Johnsen SP, Tonnesen E, et al. Antipsychotic drugs and [23] Hanley JA, McNeil BJ. The meaning and use of the area short-term mortality after peptic ulcer perforation: under a receiver operating characteristic (ROC) curve. Radi- a population-based cohort study. Aliment Pharmacol Ther ology 1982;143:29–36. 2008;28:895–902. [24] Broll S, Glaser S, Kreienbrock L. Calculating sample size [5] Christensen S, Riis A, Norgaard M, Sorensen HT, bounds for logistic regression. Prev Vet Med 2002;54:105–11. Thomsen RW. Short-term mortality after perforated or [25] Hsieh FY, Bloch DA, Larsen MD. A simple method of

Scand J Gastroenterol Downloaded from informahealthcare.com by Danmarks Natur-og on 01/08/12 bleeding peptic ulcer among elderly patients: a population- sample size calculation for linear and logistic regression. based cohort study. BMC Geriatr 2007;7:8. Stat Med 1998;17:1623–34. [6] Moller MH, Adamsen S, Wojdemann M, Moller AM. Per- [26] Youden WJ. Index for rating diagnostic tests. Cancer 1950;3: forated peptic ulcer: how to improve outcome? Scand J 32–5. Gastroenterol 2009;44:15–22. [27] Schafer JL. Multiple imputation: a primer. Stat Methods [7] Moller MH, Adamsen S, Thomsen RW, Moller AM. Pre- Med Res 1999;8:3–15. operative prognostic factors for mortality in peptic ulcer [28] Schafer JL, Graham JW. Missing data: our view of the state of perforation – a systematic review. Scand J Gastroenterol the art. Psychol Methods 2002;7:147–77. 2010;45:785–805. [29] Kehlet H, Wilmore DW. Evidence-based surgical care and the [8] Schein M, Gecelter G, Freinkel Z, Gerding H. APACHE II evolution of fast-track surgery. Ann Surg 2008;248:189–98. in emergency operations for perforated ulcers. Am J Surg [30] Sobol JB, Wunsch H. Triage of high-risk surgical patients for 1990;159:309–13. intensive care. Crit Care 2011;15:217. [9] Mishra A, Sharma D, Raina VK. A simplified prognostic [31] Rix TE, Bates T. Pre-operative risk scores for the prediction scoring system for peptic ulcer perforation in developing of outcome in elderly people who require emergency surgery. countries. Indian J Gastroenterol 2003;22:49–53. World J Emerg Surg 2007;2:16. [10] Lee FY, Leung KL, Lai BS, Ng SS, Dexter S, Lau WY. [32] Lunevicius R, Morkevicius M. Management strategies, early Predicting mortality and morbidity of patients operated on results, benefits, and risk factors of laparoscopic repair of for perforated peptic ulcers. Arch Surg 2001;136:90–4. perforated peptic ulcer. World J Surg 2005;29:1299–310. Accuracy of clinical prediction rules in PPU 35

[33] Chandra SS, Kumar SS. Definitive or conservative surgery [38] Kocer B, Surmeli S, Solak C, Unal B, Bozkurt B, for perforated gastric ulcer? - An unresolved problem. Int J Yildirim O, et al. Factors affecting mortality and morbidity Surg 2009;7:136–9. in patients with peptic ulcer perforation. J Gastroenterol [34] Lohsiriwat V, Prapasrivorakul S, Lohsiriwat D. Perforated Hepatol 2007;22:565–70. peptic ulcer: clinical presentation, surgical outcomes, and [39] Larkin JO, Bourke MG, Muhammed A, Waldron R, the accuracy of the Boey scoring system in predicting Barry K, Eustace PW. Mortality in perforated duodenal ulcer postoperative morbidity and mortality. World J Surg depends upon pre-operative risk: a retrospective 10-year 2009;33:80–5. study. Ir J Med Sci 2010;179:545–9. [35] Makela JT, Kiviniemi H, Ohtonen P, Laitinen SO. Factors [40] Forsmo HM, Glomsaker T, Vandvik PO. Perforated peptic that predict morbidity and mortality in patients with perfo- ulcer–a 12-year material. Tidsskr Nor Laegeforen 2005;125: rated peptic ulcers. Eur J Surg 2002;168:446–51. 1822–4. [36] Mak PH, Campbell RC, Irwin MG. The ASA Physical Status [41] Bone RC, Balk RA, Cerra FB, Dellinger RP, Fein AM, Classification: inter-observer consistency. American Society of Knaus WA, et al. Definitions for sepsis and organ Anesthesiologists. Anaesth Intensive Care 2002;30:633–40. failure and guidelines for the use of innovative therapies in [37] Wolters U, Wolf T, Stutzer H, Schroder T. ASA classifica- sepsis. The ACCP/SCCM Consensus Conference Commit- tion and perioperative variables as predictors of postoperative tee. American College of Chest Physicians/Society of Critical outcome. Br J Anaesth 1996;77:217–22. Care Medicine. Chest 1992;101:1644–55. For personal use only. Scand J Gastroenterol Downloaded from informahealthcare.com by Danmarks Natur-og on 01/08/12