POSSUM accurately predicts morbidity for pancreatic resection

Surgical Outcome Research POSSUM accurately predicts morbidity for pancreatic resection Wande Pratt, Saju Joseph, MD, Mark P. Callery, MD, FACS, and Charles M. Vollmer, Jr., MD, FACS, Boston, Mass

Background. The Physiologic and Operative Severity Score for the enUmeration of Mortality and Morbidity (POSSUM) is a predictive scoring system for postoperative morbidity. While numerous studies validate its application to major abdominal surgery, few exclusively consider pancreatic resections, whose unique complications are costly and problematic. We examined whether POSSUM could accurately reflect clinical and economic outcomes in pancreatic resection. Methods. 326 consecutive pancreatic resections (227 pancreaticoduodenectomies, 87 distal, 7 central, and 5 total pancreatectomies) were performed between October 2001 and January 2007. POSSUM score was prospectively calculated for each case, and patients were stratified to quintiles of morbidity risk: ⱕ 20%, 20-40%, 40-60%, 60-80%, ⱖ 80%. Actual clinical and economic outcomes were compared across the groups. Predictive risk assessment was further evaluated independently within each resection type. Logistic regression analysis was performed to identify specific POSSUM parameters predictive of postoperative morbidity. Results. Observed and Expected morbidity rates were equivalent (53.1% vs 55.5%) with an overall O/E ratio of 0.96. Although no patients presented with POSSUM scores below 20%, a relatively equal distribution was assigned to the remaining risk cohorts. Clinical and economic outcomes progressively worsened with escalations in POSSUM scores. Increasing morbidity risk was associated with significantly longer hospital stays, higher rates of complications, and more blood transfusions, ICU management, and discharge to rehabilitation facilities. This had considerable economic impact, as mean hospital costs rose from $19,951 in the 20-40% risk cohort, to $31,281 in the ⱖ 80% group. Breakdown by operation type demonstrates that POSSUM definitively predicts morbidity following both proximal and distal resection, but more accurately forecasts the need for ICU management and rehabilitation placement when pancreatoduodenectomy is performed. Multivariate analysis revealed that onehalf of POSSUM parameters were significant contributors for postoperative morbidity, with age, preoperative hemoglobin concentration, and intraoperative blood loss demonstrating the strongest correlations. Conclusion. POSSUM is a valuable perioperative scoring system for evaluating variance in pancreatic surgical methods and outcomes, and can be employed to guide management decisions that impact postoperative recovery. (Surgery 2008;143:8-19.) From the Department of Surgery, Beth Israel Deaconess Medical Center, Harvard Medical School

Surgical quality improvement initiatives are receiving increasing scrutiny, most notably in highacuity surgery, where baseline patient risk is often high, procedures are intrinsically expensive, and

This research was conducted with support from the Clinical Research Fellowship Program at Harvard Medical School offered by the Doris Duke Charitable Foundation, and the Harvard PASTEUR Program and Office of Enrichment Programs. Accepted for publication July 8, 2007. Reprint requests: Charles M. Vollmer, Jr, MD; Department of Surgery, Beth Israel Deaconess Medical Center, 330 Brookline Avenue, ST 9, Boston, MA 02215. E-mail: [email protected]. 0039-6060/$ - see front matter © 2008 Mosby, Inc. All rights reserved. doi:10.1016/j.surg.2007.07.035

8 SURGERY

complications are frequent and costly. Process and systems improvements have, to date, successfully advanced the quality of patient care in the operative and postoperative settings.1 Yet, the climate of surgery now affords unprecedented opportunities to understand better the impact of preoperative morbidity on postoperative outcomes. Recent data suggest that various prognostic factors (eg, age, disease severity, and comorbidity) influence surgical outcomes. Numerous risk scoring systems based on these identified risk factors have been devised, with the aim of delineating those patients most suitable for surgical intervention. These methods might also form the basis for rigorous appraisal of surgical outcomes within specialty centers and across medical institutions.2-4

Surgery Volume 143, Number 1

The Physiologic and Operative Severity Score for the enUmeration of Mortality and Morbidity (POSSUM) has gained acceptance as a reliable scoring system for estimating morbidity risk in major operations and to quantify the impact of complications.5-9 Using commonly measured parameters, this surgical scoring system accounts for patients’ preoperative risks, as well as the nature and conduct of the operation, in order to predict the likelihood of developing a postoperative complication. While POSSUM has been criticized for overestimating mortality, particularly among lower acuity patients,10,11 it has been validated in numerous studies as an accurate predictor of postoperative morbidity in major abdominal surgery, most notably in the colorectal, hepatic, and vascular specialties.12-16 Pancreatic conditions, with their inherent burdens and negative impact on quality of life, provide an ideal setting to apply such principles. Achievement of benchmark outcomes is especially important in the setting of pancreatic cancer, where fewer than a quarter of patients are resectable for cure, the margin for operative success is narrow, and long-term survival is rare.17,18 Although benign disease offers a better prognosis, chronic manifestations of pain, new onset diabetes, exocrine insufficiency, and biliary obstruction are similarly debilitating.19-21 Over the last 3 decades, refinements in operative technique and advancements in postoperative care have gradually contributed to significantly lower rates of mortality after elective pancreatectomy. However, postoperative complications, primarily pancreatic fistula, abscess, and delayed gastric emptying still occur with 30-60% frequency, often resulting in catastrophic events such as hemorrhage, intrabdominal sepsis, and single or multi-system organ failure.22-25 The consequences of these adverse events are increased utilization of intensive care facilities and resources, prolonged hospital stays, higher reoperative and readmission rates, and ultimately greater hospital costs.26,27 Despite the development and general acceptance of the POSSUM scoring system, there, so far, lacks any demonstration of its utilization in pancreatic surgery. This analysis, therefore, tests the value of POSSUM as a predictive index for morbidity and mortality after pancreatic resection, and determines whether it accurately reflects other clinical and economic metrics. PATIENTS AND METHODS Patients. Two surgeons (MPC, CMV) performed 326 consecutive pancreatic resections from October 2001 to January 2007, including

Pratt et al. 9

227 pancreatoduodenectomies, 87 distal, 7 central, and 5 total pancreatectomies. All patients were managed according to a standardized carepath for pancreatic resection implemented at Beth Israel Deaconess Medical Center.1 This pathway was designed to outline a standardized management approach for all patients. It provides a guide for preoperative planning, thromboembolic and antibiotic prophylaxis, perioperative pain management, and the removal of central venous catheters, nasogastric tubes, urinary catheters, and intra-abdominal drains. Furthermore, this clinical pathway standardizes patients’ perioperative fluid resuscitation, alimentation, and diagnostic testing. Management in the intensive care unit (ICU) upon transfer from the operating room or after recovery in our institution’s postoperative anesthesia care unit.was not a part of our standardized carepath, but was considered at the discretion of the operating surgeon or when adverse events dictated further intervention. Data collection. In accordance with guidelines for human subjects research, approval for analysis was obtained from the institutional review board at Beth Israel Deaconess Medical Center. Data on preoperative, intraoperative, and postoperative care were prospectively collected for each case. Preoperative parameters included patient demographics, presenting symptoms, comorbidities, vital signs, laboratory tests, prior imaging studies, and any other diagnostic or therapeutic studies performed. Intraoperative variables included total operative time, blood loss, fluid administration, and blood transfusions. Final disease pathology was determined following each case. Postoperative events and clinical outcomes were also recorded prospectively and graded by an independent research associate according to the Clavien complication scheme.28 Total hospital costs for each patient were obtained using the institution’s Casemix TSI data system, and are defined as costs from the initial operation to hospital discharge, plus any costs accrued during hospital readmissions within 30 days postoperatively. These costs reflect actual total hospital costs and not billing costs or charges. Morbidity risk classification. A detailed analysis was performed in order to assess whether expected morbidity (POSSUM) accurately reflects actual clinical and economic outcomes following pancreatectomy. Analyses were performed for the total cohort, as well as separately for each type of resection. Expected morbidity. Expected morbidity was estimated for each of the 326 consecutive patients in

10 Pratt et al.

Surgery January 2008

Table I. Physiologic score for enumeration of preoperative risk severity; operative score for enumeration of operative risk severity Physiologic score parameters 1 Age (years) Cardiac Signs

Chest radiograph findings Respiratory history Chest radiograph findings Blood pressure (systolic, mmHg) Pulse (per min) Glasgow coma score Hemoglobin (g/dL) White cell count (1000 per mm3) Urea (mg/L) Sodium (mEq/L) Potassium (mEq/L) Electrocardiogram findings

ⱕ 60 No failure

No dyspnea 110-130 50-80 15 13.0-16.0 4.0-10.0 ⱕ 22.5 ⱖ 136 3.5-5.0 Normal

2

4

8

ⱖ 71 peripheral edema; warfarin therapy

-raised jugular venous pressure

Borderline cardiomegaly

Cardiomegaly

Dyspnea on exertion Mild COPD 100-109 or 131-170

Limiting dyspnea Moderate COPD 90-99 or ⱖ 171

Dyspnea at rest Fibrosis, consolidation ⱕ 89

40-49 or 81-100 12-14 11.5-12.9 or 16.117.0 3.1-4.0 or 10.1-20.0

101-120 9-11 10.0-11.4 or 17.1-18.0

ⱕ 39 or ⱖ 121 ⱕ8 ⱕ 9.9 or ⱖ 18.1

ⱕ 3.0 or ⱖ 20.1

--

30.1-45.0 126-130 2.9-3.1 or 5.4-5.9 atrial fibrillation

ⱖ 45.1 ⱕ 125 ⱕ 2.8 or ⱖ 6.0 Any other abnormal rhythm; ⱖ 5 ectopics/min; Q waves; ST/T wave changes

61-70 Diuretic, digoxin, antihypertensive, antianginal therapy

22.6-30.0 131-135 3.2-3.4 or 5.1-5.3

Operative score parameters Operative severity Multiple procedures Total blood loss (ml) Peritoneal soiling

1 minor 1 ⱕ 100 None

2 moderate -100-500 minor (serous fluid)

Presence of malignancy Mode of surgery

None Elective

primary only --

4 major 2 501-999 local pus nodal metastases operation ⬍ 24 hours after admission

8 major ⫹ ⬎2 ⱖ 1000 free bowel content, pus, or blood distant metastases emergency (surgery within 2 hours necessary)

Adopted from Copeland et al5

accordance with evaluation of the POSSUM, which relies on 12 physiologic and 6 operative variables (Table I).5 While all data points were prospectively collected, POSSUM scores were retrospectively calculated at the completion of each case in accordance with scoring criteria. Physiological variables included patient age, Glasgow coma score, the presence of cardiac and respiratory symptoms, vital signs (systolic blood pressure and pulse), serum biochemistry evaluation (urea nitrogen, sodium, potassium), hematological parameters (white blood cell count and hemoglobin), electrocardiograph, and chest radiograph findings. Operative

variables are the magnitude of the operation, number of operations performed within 30 days, intraoperative blood loss, degree of peritoneal contamination, presence of malignancy, and the timing of surgical intervention. Scores for each variable were assigned to one of four grades based on exponentially increasing levels of severity (1, 2, 4, and 8). Separate physiologic and operative severity scores were then calculated, with minimum and maximum scores for each (physiologic: 12 to 88; operative: 6 to 48). Applying these derived severity scores, the risk of developing a postoperative complication (from 0% to 100%) was predicted for

Surgery Volume 143, Number 1

each patient using the following POSSUM score equation (PS indicates physiologic score; OS indicates operative score): The predicted risk of morbidity for the overall practice, as well that for each cohort, was then estimated by the mean POSSUM score. Expected morbidity, defined as the number of patients expected to develop a postoperative complication, was then calculated as follows: Observed morbidity. The incidence of actual postoperative complications (observed morbidity) was also determined for all patients undergoing pancreatic resection within our practice and separately for each cohort. In addition to general postoperative adverse events (ie, wound, respiratory, urinary tract, cardiac, renal, and neurologic complications), those specific to pancreatectomy were considered, including pancreatic fistula, delayed gastric emptying, biliary leak, abscess, gastrointestinal bleed, hemorrhage, ileus, and sepsis. As mentioned previously, observed morbidity was graded according to the Clavien complication scheme, a reliable clinical scoring system for postoperative morbidity, which assigns complications of escalating severity to five grades based on the use and/or requirement of therapeutic interventions. Grade I complications correspond to any deviation from the normal postoperative course without the need for pharmacologic treatment, surgical, endoscopic, or radiological intervention, but may require antiemetics, antipyretics, analgesics, diuretics, electrolytes, or physiotherapy. Examples include noninfectious diarrhea or atelectasis. Grade II complications include those that require pharmacologic interventions, including blood transfusions, total parental nutrition, or antibiotics. Examples include deep vein thrombosis or urinary tract infections. Grade III events consist of complications that require surgical, endoscopic, or radiological intervention, such as intra-abdominal fluid collections (biloma, pancreatic fistula, abscess) requiring percutaneous drainage, wound infection leading to eventration of small bowel, closure of dehiscent noninfected wounds, or anastomotic leakage requiring relaparotomy. Grade IV complications are life-threatening complications resulting in single or multiple organ dysfunction, or those requiring management in intensive care settings. Examples include lung failure requiring intubation, brain hemorrhage causing ischemic stroke, or necrotizing pancreatitis. Finally, the death of a patient represents a grade V complication. Validation of POSSUM. All patients were classified to 1 of 5 strata based on their individual POSSUM scores and subsequent risk of morbidity: (1)

Pratt et al. 11

less than 20%; (2) 20-40%; (3) 40-60%; (4) 60-80%; and (5) more than 80%. The mean POSSUM score within each stratum was calculated, and multiplied by the number of patients in each cohort to predict the number of patients expected to develop a postoperative complication. The frequency of observed and expected complications (ie, O/E ratio) was compared across cohorts. Analysis of the POSSUM scoring system was accomplished by comparing the severity of clinical outcomes across each stratum. The following clinically relevant parameters were considered: ICU utilization, duration of stay, patient discharge disposition, and hospital readmission. The need for therapeutic intervention (ie, transfusion, antibiotics, hyperalimentation, image-guided percutaneous drainage, surgical exploration) was also reviewed. Further validation of the scoring system was achieved by evaluating economic parameters across each risk stratum. Total hospital costs included those from the day of the index operation until hospital discharge, plus any costs incurred during readmissions within 30 days of discharge. Total hospital costs differ from index admission costs in that they include any hospital readmissions within 30 days postoperatively, and better represent the full economic impact of complications after pancreatic resection. Average total hospital costs were calculated and compared across the risk strata. Risk factors for morbidity. Regression analysis was performed to identify parameters from the POSSUM scoring system associated with development of complications after pancreatic resection. The 12 physiologic parameters and six operative factors were analyzed and correlated with the incidence of complications. Each was initially analyzed using univariate logistic regression analysis. Factors with P ⬍ .250 were retained for multivariate analysis. Factors demonstrating statistical significance (P ⬍ .050) on multivariate analysis were considered verifiable risk factors for prediction of morbidity according to the POSSUM scoring system. Mortality prediction. Expected mortality was also estimated utilizing the Portsmouth POSSUM (P-POSSUM), a modified method that is widely recognized as a more reliable scoring system for predicating mortality than the original POSSUM equation.10,16 It similarly utilizes the Physiologic and Operative Severity scores to estimate the risk of operative mortality (from 0% to 100%) for each patient using the following scale (PS indicates physiologic score; OS indicates operative score): Similar to the morbidity analysis, the predicted risk of mortality was estimated by the mean P-

12 Pratt et al.

Surgery January 2008

Table II. Demographics for patients presenting for pancreatic resection All resections Patients (%) Age (years) Cardiovascular symptoms Respiratory symptoms Vital signs Pulse (per min) SBP (mmHg) Biochemistry evaluation Sodium (mEq/L) Potassium (mEq/L) BUN (mg/dL) Hematological evaluation WBC (per mm3) Hemoglobin (g/dL) Electrocardiogram (%) Normal Abnormal Physiologic score

PD

Distal

Central

TP

227 (69) 62.4 118 (52) 54 (24)

87 (27) 57.1 40 (46) 19 (22)

7 (2) 53.0 4 (57) 1 (14)

5 (2) 54.4 2 (40) 1 (20)

76 137

75 137

76 135

70 149

78 135

.720 .447

139 4.1 16

139 4.1 16

139 4.0 16

140 4.4 28

138 4.2 24

.554 .028 .003

7,366 13.0

7,357 12.9

7,390 13.2

7,100 13.0

7,740 12.6

.967 .296

196 (60) 130 (40) 20.0

130 (57) 97 (43) 20.3

60 (69) 27 (31) 18.5

4 (57) 3 (43) 22.4

2 (40) 3 (60) 25.0

.214

326 60.6 164 (50) 75 (23)

P value† -.008 .741 .925

.013

All continuous variables are represented by the mean value. PD, pancreatoduodenectomy; TP, total pancreatectomy; SBP, systolic blood pressure; BUN, blood urea nitrogen; WBC, white blood cell count. †, indicates P values are for comparison of pancreatoduodenetomy (PD), distal, central, and total pancreatectomy

POSSUM score for the overall practice, as well that for each cohort. Statistical analysis. Categorical variables were compared using the chi-squared (␹2), Fischer’s exact tests, and univariate logistic regression statistics; continuous variables were compared using analysis of variance (ANOVA), the Student t tests for independent variables, and simple linear regression when appropriate. Differences between observed and expected morbidity were assessed with the ␹2 tests, using the method described by Hosmer and Lemeshow to test the goodness of fit.29 Factors associated with morbidity were calculated based on cross-tabulations using ␹2 statistic and the Pearson correlation test. Statistical significance was accepted at a P value ⬍ .050. Statistical computations were performed using Statistical Package for the Social Sciences 14.0 for Windows (SPSS, Inc., Chicago, IL) and STATA.8.2 for Windows (StataCorp LP, College Station, TX). RESULTS Patient demographics. Males and females underwent resection with relatively equal frequency (47% vs 53%). The mean age was 60.6 years (range, 23-90). All patients presented with normal Glasgow coma scores (equal to 15). Prior to definitive resection, more than one-third (n ⫽ 117) were controlled with diuretic, anti-anginal, or antihypertensive medications, and a significant proportion (14%) presented with extensive cardiac disease (congestive or valvular

heart disease, coronary artery disease, or prior myocardial infarction). While the majority of patients demonstrated normal findings on electrocardiography, 40% had abnormal findings. Respiratory symptoms were less common, present in only 75 patients (23%). Pulse, systolic blood pressure, biochemical and hematological evaluations were, on average, within normal limits (Table II). Overall, the mean Physiologic Severity score was 20.0, and ranged from 12 to 40. Patients presenting for pancreatoduodenectomy (PD) were significantly older than those presenting for distal pancreatectomy (62.4 vs 57.1 years, P ⫽ .019), and central pancreatectomy was more often performed for patients with elevated plasma chemistry levels (potassium, urea). All other parameters were roughly equivalent among the resection types. Subtle differences in age, potassium, and urea levels, however, are reflected in significantly higher Physiologic Severity scores prior to PD (mean, 20.3), central (22.4) and total pancreatectomy (25.0) when compared to distal pancreatectomy (18.6, P ⫽ .013). Operative factors. All pancreatic resections were performed electively, and were considered Major ⫹ in accordance with the POSSUM scoring system.5 Final pathology revealed that more than half of the patients harbored malignancies (53%), primarily pancreatic ductal adenocarcinoma (n ⫽ 106), and other periampullary malignancies (n ⫽ 43). Patients undergoing pancreatoduodenectomy most

Pratt et al. 13

Surgery Volume 143, Number 1

Table III. Operative outcomes during pancreatic resection All resections Blood loss (mean, ml) Operative time (mean, min) TNM classification (%) Benign Malignant Nodal Metastatic‡ Operative score

PD

502 367

525 417

153 (47) 76 (23) 92 (28) 5 (2) 15.9

79 (35) 63 (28) 81 (36) 4 (2) 16.2

Distal

Central

448 239

457 311

470 278

.918 ⬍.001

6 (86) 1 (14) 0 (0) 0 (0) 15.3

1 (20) 2 (40) 2 (40) 0 (0) 16.8

⬍.001

67 (77) 10 (12) 9 (10) 1 (1) 15.2

TP

P value†

.006

†, indicates P values are for comparison of pancreatoduodenetomy (PD), distal, central, and total pancreatectomy. ‡, Pancreatic resection for metastatic renal cell carcinoma (3 patients), metastatic melanoma (1), and metastatic pancreatic endocrine tumor.

often had ductal adenocarcinoma (n ⫽ 91, 40%), while those subjected to distal pancreatectomy largely had cystic disease (n ⫽ 35, 40%). Central pancreatectomy was most frequently performed for cystic neoplasms (n ⫽ 4) or neuroendocrine tumors (n ⫽ 3). Four (80%) total pancreatectomies were performed for pancreatic ductal adenocarcinoma, all in the setting of intraductal papillary mucinous neoplasia. The mean blood loss following all pancreatectomies was 502 ml, and was statistically equivalent among the resection types (Table III). Mean operative time equaled overall 367 minutes, but was significantly longer for pancreatoduodenectomy (417 min), when separately compared to that for distal (239 min, P ⬍ .001), central (311 min, P ⫽ .023), and total pancreatectomy (278 min, P ⫽ .007). The mean Operative Severity score was 15.9, and ranged from 13 to 27. Deeper analysis indicates that pancreatoduodenectomy was a significantly more severe operation than distal pancreatectomy (16.2 vs 15.2, P ⫽ .005). Central pancreatectomy was not significantly different in severity from either PD, distal or total pancreatectomy. Postoperative outcomes. Overall, 173 patients (53%) developed complications of any severity. Among the total cohort, 69 patients (21%) had grade I complications, 50 grade II (15%), 32 grade III (10%), and 18 grade IV (6%). The most common complications were oliguria and/or hypotension (31%), wound infection (14%), clinically relevant pancreatic fistula (14%), postoperative ileus (8%), and urinary tract infection (7%). Few patients (5%) required management in intensive care settings at any time during their hospitalizations, with 4 deaths (1.2%) occurring overall. Antibiotic therapy (not including antibiotics administered for prophylaxis) was required for 96 patients (29%), but postoperative blood transfusion (17%) and hyperalimentation (12%) were infrequently employed. The postoperative mean

hospital duration after pancreatic resection was 9.3 days (range, 1-48) for all patients. At the time of hospital discharge, 86% of patients returned home directly and 14% required additional care in rehabilitation facilities. Rates of hospital readmission and reoperation were similarly infrequent (10% and 5%, respectively). The average total hospital cost for all cases was $24,088. Postoperative outcomes across all resection types have been previously described.30 ANALYSIS OF POSSUM All resections. While clinical outcomes met or exceeded benchmark standards for pancreatectomy, predictive risk assessment demonstrated that POSSUM provides a valid estimate of absolute morbidity after pancreatic resection. The average POSSUM score for all patients undergoing pancreatic resection within our practice was 55.5%, thus, predicting (expected) morbidity in 181 of 326 patients. This risk assessment proved accurate, as 173 patients (53.1%) actually developed postoperative complications, for an overall observed-to-expected morbidity ratio of 0.96. Detailed analysis further validates POSSUM as a predictive scoring system for postoperative morbidity. Although no patients presented with POSSUM scores less than or equal to 20%, classification to all other strata occurred with equal frequency. Table IV lists the number of observed and expected complications for each group. Observed-to-expected morbidity ratios ranged from 0.73 to 1.42, with this ratio being highest for the 20-40% morbidity risk cohort, and lowest for the ⱖ 80% morbidity risk cohort. The predictive value of POSSUM was strongest for the 60-80% (O/E ratio 0.94) and 40%-60% (O/E ratio 0.93) morbidity risk groups. However, when ␹2 analysis was used to compare actual morbidity with estimated morbidity across each risk stratum, there was no significant lack of fit (␹2 ⫽ 4.57; 3 d.f.; P ⫽ .206), indicating that POSSUM

14 Pratt et al.

Surgery January 2008

Table IV. Morbidity risk stratification for resective pancreatic surgery POSSUM score ⱕ 20% 20-40% 40-60% 60-80% ⱖ 80% Overall POSSUM score ⫽

Patients

Predictive risk

Expected morbidity

Observed morbidity

O/E ratio

-102 79 85 60 326

-29.8% 48.8% 70.3% 87.1% 55.5%

-30 39 60 52 181

-43 36 56 38 173

-1.42 0.93 0.94 0.73 0.96

1

1⫹e关5.91⫺共0.16⫻PS兲 ⫺ 共0.16⫻PS兲兴 ␹2 ⫽ 4.57; 3 d.f.; p ⫽ .206 PS, physiologic score; OS, operative score.

accurately estimates outcome after pancreatic resection. POSSUM also correlates well with clinical and economic outcomes (Table V). In addition to significantly higher rates of actual morbidity, increasing patient acuity translated to more complications per patient (.62 to 1.35, P ⬍ .001), higher rates of blood transfusion, greater ICU utilization, and progressively longer hospital durations as POSSUM scores escalated. Moreover, requirements for rehabilitation placement steadily increased from 4% to 30% (P ⬍ .001). Rates of reoperation and hospital readmission, however, were not statistically different between the groups. Finally, total hospital costs significantly increased as morbidity risk increased (Fig 1). Pancreatoduodenectomy. The average POSSUM score for pancreatoduodenectomy was 58.0%. Risk assessment in this cohort was also accurate, as observed (126 patients) and expected (132) morbidity were approximately equal—O/E ratio of 0.96 (␹2 ⫽ 1.97; 3 d.f.; P ⫽ .579). Stratification according to individual POSSUM scores further predicted the clinical and economic impact of pancreatoduodenctomy in patients of varying acuity. Rates of morbidity paralleled escalations in POSSUM scores with good fidelity (P ⫽ .016), increasing from 43% for patients with low risk of morbidity (20-40%), to 70% for those with higher risk (⬎ 80%). Similarly, the mean number of complications per patient progressively increased from .63 to .56 (P ⫽ .001). Average hospital duration and costs were also associated with rising patient acuity (Fig 2). Distal pancreatectomy. Patients presenting for distal pancreatectomy had lower POSSUM scores on average (47.8%). Nevertheless, actual (39 patients) and estimated (42) morbidity had good correlation (␹2 ⫽ 1.36; 3 d.f.; P ⫽ .714), as the overall O/E ratio (0.96) resembled that for PD. This risk assessment, however, demonstrated considerably less predictive value for clinical and economic out-

comes in distal pancreatectomy. There were no significant differences across all measurable outcomes, particularly rates of complications, ICU utilization, and reoperation. Central pancreatectomy. Among patients presenting for central pancreatectomy, the mean POSSUM score was 59.8%. Postoperative complications occurred among all patients, though none required intensive care management. Moreover, there were no significant differences in hospital duration or costs among the various morbidity strata. However, central pancreatectomy was performed in only seven patients, precluding any accurate statistical analysis of POSSUM for this infrequently performed operation. Total pancreatectomy. The mean POSSUM score was highest for total pancreatectomy (71.6%), but was not significantly different from that for either PD (P ⫽ .990), distal (P ⫽ 106), or central pancreatectomy (P⫽ 1.000). Three of the five patients had POSSUM scores that exceeded 80%; yet, only one patient developed a postoperative complication and none required intensive care management or reoperation. Predictive analysis was not performed for this uncommon operation. Risk factors for morbidity. Regression analysis indicates that five physiologic parameters and two operative parameters are associated with the development of complications following pancreatic resection. Univariate analysis demonstrates that history of cardiac or respiratory disease, systolic blood pressure, the presence of finding on electrocardiogram, and the presence of malignancy are associated with postoperative morbidity, but were not significant on multivariate analysis. On multivariate analysis, factors associated with complications included patient age, hemoglobin level, and the amount of intraoperative blood loss. Each 10-year increase in patient age increases the odds of morbidity by 22% (OR 1.22, 95% CI 1.00 to 1.49, P ⫽ .049), while each 1 g/dL decrease below

Pratt et al. 15

Surgery Volume 143, Number 1

Table V. Postoperative outcomes for all pancreatic resections given morbidity stratification POSSUM score

20-40%

40-60%

60-80%

ⱖ 80%

P value

Patients (%) Actual morbidity (%) Number of complications (%) 1 complication ⱖ 2 complications No. of complications per patient Therapeutic interventions (%) Transfusion Antibiotics Hyperalimentation Percutaneous drainage Reoperation (%) ICU Utilization (%) Duration of stay (days) Discharge disposition (%) Home Rehabilitation facilities Hospital readmission (%) Mortality (%)

102 (31) 43 (42)

79 (24) 36 (46)

85 (26) 56 (66)

60 (18) 38 (63)

-.002

30 (29) 13 (13) 0.62

19 (24) 17 (22) 0.94

33 (39) 23 (27) 1.28

21 (35) 17 (28) 1.35

.008 ⬍.001

5 (5) 27 (26) 9 (9) 5 (5) 2 (2) 1 (1) 8.1

12 (15) 23 (29) 8 (10) 3 (4) 5 (6) 5 (6) 9.2

22 (26) 27 (32) 12 (14) 6 (7) 3 (4) 2 (2) 9.8

15 (25) 19 (32) 11 (18) 3 (5) 5 (8) 7 (12) 10.7

.001 .850 .298 .822 .279 .037 .002

98 (96) 4 (4) 12 (12) 0 (0)

69 (91) 7 (9) 6 (8) 3 (3.8)

69 (81) 16 (19) 9 (11) 0 (0)

41 (70) 18 (30) 7 (12) 1 (1.7)

⬍.001 .811 .914

All continuous variables are represented by the mean value.

Fig 1. Economic impact of increasing patient acuity (as measured by POSSUM) in all pancreatic resections.

normal hemoglobin concentration levels (13-16 g/dL) is associated with an 18% increase in morbidity (OR 1.18, 95% CI 1.01 to 1.38, P ⫽ .035). Increased blood loss, however, was the most significant factor predisposing patients to complications. Each additional unit (375 ml) of blood loss increases the odds of morbidity by 45% (OR 1.45, 95% CI 1.00 to 2.12, P ⫽ .007). One-half of all POSSUM score parameters were not significant for postoperative morbidity on univariate or multivariate analysis. These 9 parameters consisted of preoperative heart rate, Glasgow Coma score, white cell count, serum biochemistry (urea nitrogen, sodium and potassium concentrations), the magnitude of the operation, number of oper-

Fig 2. Impact of increasing patient acuity (as measured by POSSUM) in pancreatoduodenectomy.

ations performed within 30 days, degree of peritoneal contamination, and the timing of surgical intervention. Mortality. The mean POSSUM and P-POSSUM scores for mortality were 16.3% and 6.5%, respectively, for all resection types. Breakdown of each resection type revealed a significantly higher mean P-POSSUM score for total pancreatectomy when compared to distal pancreatectomy (15.7% vs 4.6%, P ⫽ .014), but all other comparisons were statistically equivalent. Predictive assessment markedly overestimated actual mortality (21 deaths) after pancreatic resection, as only 4 deaths (1.2%) occurred overall (3 PD; 1 distal pancreatectomy) for an O/E mortality ratio of 0.19.

16 Pratt et al.

Surgery January 2008

Table VI. Four postoperative deaths Age/ gender 67 F

Comorbidities

Atrial fibrillation; mitral valve insufficiency; prior acute pancreatitis 74 F NIDDM; Cirrhosis; Mandatory transfusion abstinence (Jehovah’s Witness) 58 M CAD/Myocardial infarction; IDDM; cardiac stent placement 81 F NIDDM; HTN; prior uterine cancer

Diagnosis

POSSUM P-POSSUM (morbidity) (mortality)

Operation

Complications

Expired

POD 9 Heparin induced thrombocytopenia; mesenteric venous thrombosis; bowel infarction Refractory postPOD 1 Distal operative pancreatectomy hemorrhage with splenectomy

Ampullary cancer

47.0%

2.4%

PPPD

Pancreatic cancer

88.3%

17.5%

Chronic pancreatitis

58.2%

3.9%

PPPD

Ampullary cancer

41.6%

2.1%

PPPD

POD 4 Upper gastrointestinal hemorrhage POD 2; aspiration event; coma Myocardial POD 36 infarction POD 2

PPPD, Pylorus-preserving pancreaticoduodenectomy; NIDDM, Noninsulin dependent diabetes mellitus; CAD, coronary artery disease; IDDM, insulin dependent diabetes mellitus; HTN, hypertension; POD, postoperative day.

When these 4 deaths are examined closely, all were presumably unpredictable (Table VI). One patient developed portal and mesenteric venous thrombosis, as a consequnce of heparin-induced thrombocytopenia, on postoperative day three follwing pancreatoduodenectomy. A second patient with cirrhosis refused blood transfusion on religious grounds and died secondary to refractory postoperative hemorrhage following distal panceratectomy for malignacy. A third patient suffered an aspiration event due to gastrointestinal hemorrhage. The fourth patient, an 81-year-old patient with type II diabetes mellitus, hypertension, and a diagnosis of ampullary adenocarcinoma, had myocardial infarction on postoperative day 2, with a gradual clinical decline over the ensuing month, despite a normal preoperative cardiovascular workup. DISCUSSION POSSUM is increasingly being applied to a number of high-acuity operations, most notably for colorectal operations; however, a paucity of studies have evaluated the potential value of POSSUM for pancreatic resections specifically.5-16 A recent study by Kocher et al31 was among the first and largest (177 cases) to examine POSSUM explicitly in patients undergoing major elective hepatopancreatobiliary surgical procedures. Yet, owing to its

random selection design, this study considered only 26 pancreatic operations (22 PDs, 2 distal and 2 total pancreatectomies) over a 10-year period. The authors concluded that such operations are distinct from other hepatopancreatobiliary procedures, largely because of their high propensity for excessive blood loss and gastrointestinal and respiratory complications.31 Although operative mortality after pancreatic resection is now considered an infrequent event in the hands of specialists, morbidity remains a significant problem for all surgeons performing these complex operations. Most clinical series report rates between 30% and 60%, and these complications are frequently associated with potentially catastrophic outcomes.19-26 Successful recovery after pancreatic resection depends on the ability to reduce the incidence of postoperative complications, as well as to mitigate their impact when they do occur. Currently, many specialty centers rely on multidisciplinary management approaches to achieve these objectives; however, there is an increasing interest in understanding the impact of baseline preoperative morbidity in pancreatic surgery. Therefore, using a large, contemporary series of consecutive patients, we sought to validate POSSUM in pancreatic surgery, and to determine whether this unique scoring system accurately re-

Surgery Volume 143, Number 1

flects clinical and economic outcomes in that domain. This study also differs from prior investigations in that it examines POSSUM over the full spectrum of pancreatic resections: pancreatoduodenectomy, distal, central, and total pancreatectomy. In this current analysis, POSSUM accurately estimates overall postoperative morbidity. When patients were classified to 1 of 5 morbidity risk groups, based on individual POSSUM scores, rates of complications significantly increased in parallel with rising predicted risk of morbidity. Patients not only suffered multiple complications but frequently encountered more severe clinical sequelae. Hospital stays were longer, and patients tended to require more aggressive management, including blood transfusion, ICU utilization, and discharge to rehabilitation facilities. Equally impressive was the economic impact associated with patient acuity, as increasing POSSUM scores were linked with progressively greater total hospital costs. Despite these findings, POSSUM neither predicts the need for reoperations nor the incidence of hospital readmission. While these particular outcomes are significant, their infrequent occurrence may partially explain the lack of association with POSSUM in our clinical series. This scoring system was not originally designed to predict the risk of reoperation or readmission. Finally, reoperation and hospital readmission may depend on other factors such as partial vessel resection during the index operation, the type and severity of complication, and the duration of the initial hospitalization.27,30 Breakdown by each resection type shows that POSSUM accurately predicts occurrence of morbidity for both pancreatoduodenectomy and distal pancreatectomy, but not for central or total resections. This latter conclusion, however, is tempered by a small number of patients undergoing central or complete pancreatectomy. When clinical and economic outcomes were also compared within each resection type, we observed that POSSUM best reflects the clinical and economic impact of PD, particularly in regards to the rate of complications, hospital duration, and total hospital costs. For distal pancreatectomy, these same measures were not associated with patient acuity level, as mean duration of stay and total hospital costs were equivalent for each morbidity level. This distinction suggests that POSSUM is more accurate for pancreatoduodenectomy than it is for distal pancreatectomy. There are two possible explanations for this observation. First, PD is traditionally felt to represent

Pratt et al. 17

a more severe operation than distal pancreatectomy, owing to its more extensive resection of bowel, pancreas, and lymphatics, its challenging reconstruction, and its propensity for complications with more significant clinical sequelae. Accordingly, we observed significantly higher rates of complications, longer hospital stays, and greater hospital costs for PD. Because the potential for morbidity and mortality is greater after PD, this risk scoring system demonstrates more predictive value in this operation. Similar findings have been observed in other studies on the topic, where POSSUM has significantly less predictive value for lower acuity procedures.11,13,14,31 Second, while distal pancreatectomy is frequently performed for benign conditions, PD is more often indicated for definitive treatment of pancreatic adenocarcinoma and other periampullary malignancies. Thus preoperative morbidity and physiologic profile are more variable in patients presenting for PD. This scenario provides a more accurate assessment of the impact of variance in patient acuity on postoperative outcome. While the current study specifically examines the utility of POSSUM for pancreatic resections, there are inherent limitations of this retrospective analysis. First, this study was conducted within a single, specialty practice at a high-volume institution, and did not allow for comparative assessment across multiple practices using POSSUM. Second, all pancreatic resections were performed by 2 fellowship-trained hepatopancreatobiliary surgeons who have a unified concept of operative technique and postoperative approach for these operations. Therefore, it remains unclear what specific management or technical approaches might contribute to different outcomes for patients with varying POSSUM scores. Third, because central and total pancreatectomies were rarely performed, this current analysis does not provide any specific conclusions about these unique operations. Finally, our analysis of operative mortality was limited by its infrequent occurrence rate in our clinical series. Death after pancreatic resection was markedly overestimated in our series, and did not correlate with individual POSSUM or P-POSSUM scores, indicating that these unfortunate events were largely unpredictable. These findings correspond with the current understanding of mortality after high-acuity surgery. A number of studies demonstrate POSSUM and P-POSSUM exhibit less predictive value for mortality, as opposed to morbidity, after elective operations.10,12,14,31 Furthermore, many current reports in the surgical and medical literature link low mortality rates with other factors,

18 Pratt et al.

including hospital volume, experience of the surgeon, improvements in critical care anesthesia and intensive care management, and the development of multidisciplinary care teams.1,32-36 All of these have the potential to mitigate the influence of a patient’s baseline physiologic risk, and while severe complications may occur, they may be contained to the extent that they do not ultimately result in death. Despite these limitations, POSSUM reflects the variable impact of preoperative morbidity on postoperative recovery after pancreatic resection, and predicts clinical and economic outcomes after pancreatoduodenectomy. Applications of POSSUM have the potential to enhance the process of comparative quality assessments and systems improvement initiatives (eg, clinical care pathways) both between and within individual practices.1,37 This is particularly important given the rise of pay-for-performance systems and their anticipated impact on health care reimbursements. POSSUM has had further implications within our practice. It is currently employed as a counseling tool to obtain informed patient consent during the preoperative consultation. However, we propose that individual POSSUM scores should not preclude pancreatic resection, but instead help the physician and patient alter expectations of postoperative outcomes. Based on the findings of our multivariate analysis, we have developed four process improvement measures to help guide management scenarios in our pancreato-biliary surgical practice. First, as our results suggest that elderly age is associated with an increased incidence of postoperative complications and other negative clinical sequelae, we have developed several perioperative interventions that target older patients undergoing pancreatic resection. All patients older than age 75 years that are scheduled for pancreatectomy now undergo a thorough preoperative evaluation by a dedicated group of gerontologists at our institution, who provide recommendations for cardiac and respiratory management, perioperative pain control, nutritional support, and discharge planning. Second, patients with high Physiologic POSSUM Scores are provided additional preoperative interventions— hyperalimentation, antibiotics, endoscopic drainage, biliary stenting—whenever permitted in order to improve physiologic parameters. This is particularly useful, and is often indicated, when patients present with malignant obstructive jaundice, comorbid illness (cardiac or respiratory illness), preoperative diabetes, or malnutrition.1

Surgery January 2008

Third, our standardized carepath dictates that all patients receive preoperative evaluation of hemoglobin concentration, the partial thromboblastin time (PTT), and the international normalized ratio (INR). Given the findings of this study, we now tend to provide additional intraoperative blood transfusion to patients with low hematocit levels. Abnormal coagulation studies (PTT, INR) are addressed to control intraoperative blood loss, and patients are frequently given vitamin K in the setting of hyperbilirubinemia. Fourth, our analysis demonstrates that intraoperative blood loss is the most variable and influential factor affecting the risk of morbidity. This finding underscores our conviction that meticulous operative technique is essential to reduce blood loss and to optimize postoperative recovery.37 In summary, the POSSUM scoring system is a useful adjunct for predicting and evaluating outcomes in pancreatic surgery, and can be helpful in guiding management decisions that impact postoperative recovery. REFERENCES 1. Vollmer CM, Pratt W, Vanounou T, Maithel SK, Callery MP. Quality assessment in high-acuity surgery: Volume and mortality are not enough. Arch Surg 2007;142:371-80. 2. Hartley MN, Sagar PM. The surgeon’s ‘gut feeling’ as a predictor of post-operative outcome. Ann R Coll Surg Engl 1994;76(Suppl):277-8. 3. Sutton R, Bann S, Brooks M, Sarin S. The Surgical Risk Scale as an improved tool for risk-adjusted analysis in comparative surgical audit. Br J Surg 2002;89:763-8. 4. Brooks MJ, Sutton R, Sarin S. Comparison of Surgical Risk Score, POSSUM, and p-POSSUM in higher-risk surgical patients. Br J Surg 2005;92:1288-92. 5. Copeland GP, Jones D, Walters M. POSSUM: a scoring system for surgical audit. Br J Surg 1991;78:356-60. 6. Copeland GP, Jones D, Wilcox A, Harris PL. Comparative vascular audit using the POSSUM scoring system. Ann R Coll Surg Engl 1993; 75:175-177. 7. Copeland GP, Sagar P, Brennan J, et al. Risk adjusted analysis of surgeon performance. Br J Surg 1995;82:408-11. 8. Sagar PM, Hartley MN, MacFie J, Taylor BA, Copeland GP. Comparison of individual surgeon’s performance. Dis Colon Rectum 1996;38:654-8. 9. Copeland GP. Assessing the surgeon: 10 years experience with the POSSUM system. J Clin Excellence 2000;2:187-90. 10. Prytherch DR, Whiteley MS, Higgins B, et al. POSSUM and Portsmouth POSSUM for predicting mortality. Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity. Br J Surg 1998;85:1217-20. 11. Law WL, Lam CM, Lee YM. Evaluation of outcome of laparoscopic colorectal resection with POSSUM, Portsmouth POSSUM, and colorectal POSSUM. Br J Surg 2006;93:94-9. 12. Midwinter MJ, Tytherleigh MJ, Ashley S. Estimation of mortality and morbidity risk in vascular surgery using POSSUM and the Portsmouth predictor equation. Br J Surg 1999;471-4. 13. Copeland GP. The POSSUM system of surgical audit. Arch Surg 2002;137:15-9.

Pratt et al. 19

Surgery Volume 143, Number 1

14. Lam CM, Fan ST, Yuen WC, Law WL, Poon K. Validation of POSSUM scoring systems for audit of major hepatectomy. Br J Surg 2004;91:450-4. 15. Slim K, Panis Y, Alves A, et al. Predicting postoperative mortality in patients undergoing colorectal surgery. World J Surg 2006; 30:100-6. 16. Whiteley MS, Prytherch DR, Higgins B, Weaver PC, Prout WG. An evaluation of the POSSUM scoring surgical system. Br J Surg 1996; 83:812-5. 17. Janes RH, Neiderhuber JE, Chmiel JS, et al. National patterns of care for pancreatic cancer. Results of a survey by the Commission on Cancer. Ann Surg 1996;223:261-72. 18. Lillemoe KD. Current management of pancreatic carcinoma. Ann Surg 1995; 221:133-6. 19. Lillemoe KD, Kaushal S, Cameron JL, Sohn TA, Pitt HA, Yeo CJ. Distal pancreatectomy: indications and outcomes in 235 patients. Ann Surg 1999;229:693-700. 20. Sohn TA, Yeo CJ, Cameron JL, et al. Intraductal papillary mucinous neoplasms of the pancreas: an updated experience. Ann Surg 2004;239:788-99. 21. Efron DT, Lillemoe KD, Cameron JL, Yeo CJ. Central pancreatectomy with pancreaticogastrostomy for benign pancreatic pathology. J Gastrointest Surg 2004;8:532-8. 22. Yeo CJ, Cameron JL, Sohn TA, et al. Six hundred fifty consecutive pancreaticoduodenectomies in the 1990s: Pathology, complications, and outcomes. Ann Surg 1997;226:248-57. 23. Niedergethmann M, Farag Soliman M, Post S. Postoperative complications of pancreatic cancer surgery. Minerva Chir 2004;59:175-83. 24. Tsiotos GG, Farnell MB, Sarr MG. Are the results of pancreatectomy for pancreatic cancer improving? World J Surg 1999;23:913-9. 25. Traverso LW, Shichi H, Low DE. Useful benchmarks to evaluate outcomes after esophagectomy and pancreaticoduodenectomy. Am J Surg 2004;187:604-8. 26. Pratt WB, Maithel SK, Vanounou T, Huang ZS, Callery MP, Vollmer CM. Clinical and economic validation of the Inter-

27.

28.

29. 30.

31.

32.

33.

34.

35.

36.

37.

national Study Group of Pancreatic Fistula (ISGPF) classification scheme. Ann Surg 2007;245:443-51. Emick DM, Riall TS, Cameron JL, et al. Hospital readmission after pancreaticoduodenectomy. J Gastrointest Surg 2006;10:1243-53. Dindo D, Demartines N, Clavien PA. Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg 2004;240:205-13. Hosmer DW, Lemeshow S. Applied Logistic Regression. 2nd ed. New York: John Wiley & Sons Inc; 2000. Pratt WB, Maithel SK, Vanounou T, Callery MP, Vollmer CM. Postoperative pancreatic fistulas are not equivalent after proximal, distal, and central pancreatectomy. J Gastrointest Surg 2006;10:1264-79. Kocher HM, Tekkis PP, Gopal P, et al. Risk-adjustment in hepatobiliarypancreatic surgery. World J Surg 2005;11: 2450-5. Cameron JL, Riall TS, Coleman J, Belcher KA. One thousand consecutive pancreaticoduodenectomies. Ann Surg 2006;244:10-5. Birkmeyer JD, SiewersAE, Finlayson EVA, et al. Hospital volume and surgical mortality in the United States. N Engl J Med 2002;346:1128-37. Birkmeyer JD, Finlayson EVA, Birkmeyer CM. Volume standards for high-risk surgical procedures: potential benefits of the Leapfrog initiative. Surgery 2001;130:415-22. Gordon TA, Burleyson GP, Tielsch JM, et al. The effects of regionalization on cost and outcome for one general highrisk surgical procedure. Ann Surg 1995;222:638-45. Sosa JA, Bowman HM, Gordon TA, et al. Importance of hospital volume in the overall management of pancreatic cancer. Ann Surg 1998;228:429-38. Vanounou T, Pratt WB, Fischer JE, Vollmer CM, Callery MP. Deviation Based Cost Modeling (DBCM): a generalizable model to evaluate the clinical and economic impact of clinical pathways. J Am Coll Surg 2007;204;570-9.