Value of E-PASS models for predicting postoperative morbidity and mortality in resection of perihilar cholangiocarcinoma and gallbladder carcinoma

Abstract

Background

It has previously been reported that a general risk model, Estimation of Physiologic Ability and Surgical Stress (E-PASS), and its modified version, mE-PASS, had a high predictive power for postoperative mortality and morbidity in a variety of gastrointestinal surgeries. This study evaluated their utilities in proximal biliary carcinoma resection.

Methods

E-PASS variables were collected in patients undergoing resection of perihilar cholangiocarcinoma and gallbladder carcinoma in Japanese referral hospitals.

Results

Analysis of 125 patients with gallbladder cancer and 97 patients with perihilar cholangiocarcinoma (n = 222). Fifty-six patients (25%) underwent liver resection with either hemihepatectomy or extended hemihepatectomy. The E-PASS models showed a high discrimination power to predict in-hospital mortality; areas under the receiver operating characteristic curve (95% confidence intervals) were 0.85 (0.76–0.94) for E-PASS and 0.82 (0.73–0.91) for mE-PASS. The predicted mortality rates correlated with the severity of postoperative complications (Spearman's rank correlation coefficient: ρ = 0.51, P < 0.001 for E-PASS; ρ = 0.47, P < 0.001 for mE-PASS).

Conclusions

The E-PASS models examined herein may accurately predict postoperative morbidity and mortality in proximal biliary carcinoma resection. These models will be useful for surgical decision-making, informed consent, and risk adjustments in surgical audits.

Introduction

Hepatopancreatobiliary malignancy is a challenging area for surgeons due its high morbidity and mortality rates. Postoperative mortality in perihilar cholangiocarcinoma is still an important issue as mortality rates ranged between 0% and 15%.^{1, 2, 3} According to the American College of Surgeons National Surgical Quality Improvement Program data base, postoperative morbidity rates for cholangiocarcinoma surgery were 29.4% for hepatectomy, 43.5% for biliary-enteric anastomosis (BEA) and 47.8% for hepatectomy and biliary-enteric anastomosis.³ 30-day mortality rates were 5.2% for hepatectomy and 12.0% for hepatectomy and BEA.³ Regarding gallbladder carcinoma resection, Nishino et al.⁴ reported that postoperative mortality rate were 2.8% when the hepatectomy was less than right hemihepatectomy, while the rate increased up to 12.5% when the hepatectomy was right hemihepatectomy or more. Therefore, predicting the postoperative risk is important for both clinicians and patients in this field.

Overwhelming surgical stress that exceeds a patient's physiological ability may result in the disruption of homeostasis in vital organs, leading to postoperative complications in many organs. Based on this hypothesis, we previously constructed a prediction model for postoperative morbidity and mortality in elective gastrointestinal surgeries, which was designated as Estimation of Physiologic Ability and Surgical Stress (E-PASS).⁵ Several cohort studies demonstrated reproducible outcomes for predicting postoperative morbidity and mortality.^{6, 7, 8, 9} We further modified this model to preoperatively predict mortality rates using a reduced number of variables, and designated this modified version as mE-PASS.⁹ Our prospective cohort study, which targeted a wide variety of gastrointestinal surgeries, demonstrated that mE-PASS had a higher discriminative power than the well-known prediction models, Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity (POSSUM)¹⁰ and its modified version, Portsmouth-POSSUM (P-POSSUM).¹¹ Moreover, the E-PASS models were shown to be useful in many specialty surgical fields, including surgeries for colorectal carcinoma,¹² gastric carcinoma,¹³ liver carcinoma,¹⁴ and choledochocystolithiasis.¹⁵ Regarding proximal biliary carcinoma resection, Wang et al. demonstrated that E-PASS accurately predicted postoperative mortality in the surgical treatment of hilar carcinoma.¹⁶ No further study has been performed in this field. Preoperative risk tools are needed for liver and biliary surgery in proximal bile duct tumors as these procedures are associated with high risks of postoperative morbidity and mortality. Therefore, this study was undertaken to determine whether the E-PASS models were capable of predicting postoperative outcomes in resection of proximal biliary carcinoma.

Patients and methods

Study design and patient selection

This study was performed according to the ethical guidelines on epidemiological research developed by the Ministry of Health, Labor, and Welfare of Japan. This study analyzed two subsets of patients (Cohort 1, 1991–2007; Cohort 2, 2007–2013). Cohort 1 was composed of patients in previously conducted multicenter cohort studies in 16 referral hospitals in Japan.^{5, 6, 7, 8, 9} Cohort 2 was then set to obtain a sufficient sample size composed of patients in 5 referral hospitals in Japan. Most participating hospitals in both cohorts are medium-volume surgical centers and typically manage approximately 30–80 liver resections annually. In both cohorts, the patients were those who underwent elective resection of perihilar cholangiocarcinoma or gallbladder carcinoma. Perihilar cholangiocarcinoma is defined as an apparent malignant biliary stricture originating in the common hepatic duct or in the hepatic duct confluence (Bismuth types 1 to 4), or in the left or right hepatic duct.¹⁷ Perihilar cholangiocarcinoma was classified according to Bismuth classification.¹⁸ Patients who underwent palliative bypass or exploratory laparotomy were excluded from this study. Patients, who underwent cholecystectomy for stone disease and were postoperatively diagnosed with gallbladder carcinoma, were also excluded from this study.

Surgical procedure

The standard operative procedure for perihilar cholangiocarcinoma is bile duct resection with regional lymph node dissection plus hepatectomy, as recommended in Japanese guidelines.¹⁹ Extrahepatic bile duct resection alone is also selected in patients with Bismuth type I carcinoma for whom it is judged that curative resection can be achieved after a strict diagnosis of the local extension of the cancer.¹⁹

The standard operative procedure for gallbladder carcinoma is open cholecystectomy with regional lymph node dissection plus partial hepatectomy. Extrahepatic bile duct resection is simultaneously carried out when a tumor is thought to be very close to or invaded the extrahepatic bile duct on preoperative assessment or intraoperative inspection. Concomitant hemihepatectomy or extended hemihepatectomy is indicated for invasive tumors.

Type of liver resection was classified according to the Brisbane classification.²⁰

E-PASS models

We investigated E-PASS variables and evaluated postoperative courses. E-PASS development has previously been described in detail.⁵ Shortly, the original E-PASS consists of the preoperative risk score (PRS), the surgical stress score (SSS), and the comprehensive risk score (CRS) in which PRS and SSS is combined. Using the CRS, we generated equations for predicted in-hospital mortality rates.⁸

• (1)PRS – Preoperative risk score

PRS = −0.0686 + 0.00345 X₁ + 0.323 X₂ + 0.205 X₃ + 0.153 X₄ + 0.148 X₅ + 0.0666X₆,

where X₁ is age; X₂, the presence (1) or absence (0) of severe heart disease; X₃, the presence (1) or absence (0) of severe pulmonary disease; X₄, the presence (1) or absence (0) of diabetes mellitus; X₅, the performance status index (range, 0–4); and X₆, the ASA physiological status classification (range, 1–5).

Severe heart disease is defined as heart failure meeting the criteria of New York Heart Association Class III or IV, or severe arrhythmia requiring mechanical support. Severe pulmonary disease is defined as any condition with a percent vital capacity of less than 60% and/or a percent forced expiratory volume in 1 s of less than 50%. Diabetes mellitus is defined according to the World Health Organization criteria.

• (2)SSS – Surgical stress score

SSS = −0.342 + 0.0139 X₁ + 0.0392 X₂ + 0.352 X₃

Where X₁ is blood loss (in grams) divided by body weight (in kilograms); X₂, the operation time (in hours); and X₃, the extent of the skin incision (0 indicates a minor incision for laparoscopic or thoracoscopic surgery including laparoscopic- or thoracoscopic-assisted surgery; 1, laparotomy or thoracotomy alone; and 2, laparotomy and thoracotomy).

• (3)CRS – Comprehensive risk score

CRS = −0.328 + (0.936 × PRS) + (0.976 × SSS)

• (4)Equations for predicted in-hospital mortality rates (%)

$$CRS<0.159Y\fallingdotseq 0$$

$$0.159\le CRS<2.98Y=-0.465+1.192\left(CRS\right)+10.91{\left(CRS\right)}^2$$

$$CRS\ge 2.98Y=100$$

We subsequently modified this model, designated this system as modified E-PASS (mE-PASS).⁹ In mE-PASS, PRS is the same as the original E-PASS. As an indicator of surgical severity, we used the median value of SSS in each surgical procedure as shown in Table 1. This score is called the Surgical Stressed Score fixed (SSSf). When making SSSf, surgical procedures were classified according to the Japanese National Health Care reimbursement system, 2004 edition.⁹ Using PRS and SSSf, we constructed a new Comprehensive Risk Score fixed (CRSf) and an equation for postoperative mortality rate as shown below.

• (1)CRSf – Comprehensive risk score fixed

(CRSf) = 0.052 + 0.58(PRS) + 0.83(SSSf)

• (2)Equations for predicted in-hospital mortality rates

$$CRSf<\mathrm{0.326}Y\fallingdotseq 0$$

$$CRSf\ge 0.326Y=-0.0541\left(CRSf\right)+0.197{\left(CRSf\right)}^2-0.00328$$

Table 1.

Values of Surgical Stress Score fixed (SSSf) in mE-PASS

Surgical procedure	SSSf
Cholecystectomy with nodal dissection, with or without partial hepatectomy	0.309
Choledochal resection with nodal dissection, with or without partial hepatectomy	0.401
Liver segmentectomy	0.453
Hemihepatectomy	0.663
Extended hemihepatectomy	1.025

Surgical procedures were classified according to the Japanese National Health Care reimbursement system, 2004 edition.⁹ The values for SSSf were determined as the median values of Surgical Stress Score (SSS) of E-PASS scoring system in individual procedures.⁹

Endpoints

Endpoints of study were in-hospital mortality and 30-day morbidity. Postoperative complications were defined as described previously.²¹ Anastomotic leak is defined as conditions that require documentation by re-operation or by contrast study of leak from suture line in a viscus into a body cavity or to the skin.²¹ Liver failure was defined as 1) a serum bilirubin level of >10 mg/dL together with an NH₃ level of ≥100 mg/dL or a prothrombin time of <40%, or 2) a requirement for continuous hemodiafiltration or plasma exchange, as described previously.²² Bile leak is defined as Grade B (requiring a change in clinical management) or C (a repeat laparotomy is necessary) of the International Study Group of Liver Surgery (ISGLS) criteria.²³

The severity of postoperative complications was classified using the Clavien grading system, as described previously.²⁴

Statistical analysis

All statistical analyses were performed using the software, Stata version 12.0 (StataCorp LP, College Station, Texas). A two-tailed P value of less than 0.05 was considered significant. The Mann–Whitney U test was used to test for differences in non-parametric continuous variables. Categorical variables between groups were compared using the chi-square test with Yates correction for continuity or, where appropriate. The relationship between ordinal and continuous variables was analyzed by Spearman's rank correlation (ρ), the significance of which was determined by Spearman's rank sum test.

The discriminative power of a model to discriminate patients who died during hospitalization from those who did not was assessed by calculating the area under the receiver operating characteristic curve (AUC) with a 95% confidence interval (CI). AUC > 0.8 was considered as good discriminative power. The significance of AUCs between the models was determined by the chi-square test.

To analyze the relationship of postoperative morbidity and the E-PASS models, we created a stacked bar chart showing proportions with each Clavien's grade.²⁵ Patients were divided into 4 risk bands according to the order of the predicted mortality rates in each model. Each risk band contained the same number of patients (n = 55 or n = 56). The proportions of each grade (0–5) were calculated in each risk band.

In the analysis of CRS value on postoperative mortality rates, we determined the cut-off value of CRS by receiver operating characteristic curve analysis.

To analyze the independence of the predictors for postoperative complications, logistic regression analysis was performed. We included major liver resection, BEA, biliary drainage, cholangitis, and the CRS as covariables.

Results

Demographic data

Table 2 shows the demographic data of the patients examined. The median age was 71.5 years old with a range between 38 and 89 years old. 58% of the patients were male. The preoperative liver function data were obtained in 174 patients from the retrospective investigation. Elevated serum bilirubin levels greater than 1.5 mg/dL were observed in 7% (7/102) of the patients with gallbladder carcinoma and 51% (37/72) of those with perihilar cholangiocarcinoma. Preoperative drainage was performed in 6% (6/102) of the patients with gallbladder carcinoma and 49% (35/72) of those with perihilar cholangiocarcinoma. Preoperative cholangitis was observed in 10% (7/72) of patients with perihilar cholangiocarcinoma. On the other hand, no patient with gallbladder carcinoma showed preoperative cholangitis. Hepatectomy was performed in 81% of these patients (180/222). Regarding segmentectomy, we had 10 patients with perihilar cholangiocarcinoma undergoing central segmentectomy.

Table 2.

Demographic data of subjects

Diagnoses, No.	222
Perihilar cholangiocarcinoma	97
Bismuth type I	30
Bismuth type II	19
Bismuth type IIIa	7
Bismuth type IIIb	26
Bismuth type IV	15
Gallbladder carcinoma	125
Preoperative liver function, No.a	174
Elevated serum bilirubin levels greater than 1.5 mg/dL	44
Preoperative biliary drainage	41
Preoperative cholangitis	7
Surgical procedures, No.	222
Cholecystectomy with nodal dissection, with partial hepatectomy	85
Cholecystectomy with nodal dissection, without partial hepatectomy	6
Choledochal resection with nodal dissection, with partial hepatectomy	6
Choledochal resection with nodal dissection, without partial hepatectomy	38
Liver resection with choledochal resection with nodal dissection
Segmentectomy 4	1
Segmentectomy 2,3	1
Segmentectomy 4,5	29
Left hemihepatectomy − segmentectomy 1	5
Left hemihepatectomy + segmentectomy 1	20
Extended left hemihepatectomy + segmentectomy 1	5
Right hemihepatectomy − segmentectomy 1	4
Right hemihepatectomy + segmentectomy 1	3
Extended right hemihepatectomy − segmentectomy 1	9
Extended right hemihepatectomy + segmentectomy 1	10

^aPreoperative liver function was obtained in 174 patients from the retrospective investigation.

Treatment outcomes

Seventy-eight patients (35%) developed postoperative complications of Clavien grading III or greater, as shown in Table 3. The overall 30-day, 90-day, and in-hospital mortality rates were 5% (11/222), 9% (20/222), and 9% (21/222), respectively. The in-hospital mortality rate for major liver resection, hemihepatectomy or extended hemihepatectomy, with BEA was 23% (13/56), whereas that of BEA without major liver resection was 4% (3/71). The in-hospital mortality rates in major liver resection were listed in Table 4. Regarding the causes of in-hospital mortality, 9 out of 21 deceased patients had liver failure. Liver failure-unassociated mortality included death from hemoperitoneum (n = 4), anastomotic leakage (n = 2), disseminated intravascular coagulopathy (n = 2), abdominal abscess (n = 1), liver abscess (n = 1), pneumonia (n = 1), and an embolus in the portal vein (n = 1).

Table 3.

Postoperative complications

	Perihilar cholangio-carcinoma (n = 97)	Gallbladder carcinoma (n = 125)	Total (n = 222)
Bile leak	15 (15%)	5 (4%)	20 (9%)
Anastomotic leak	12 (12%)	6 (5%)	18 (8%)
Abdominal abscess	12 (12%)	6 (5%)	18 (8%)
Liver failure	9 (9%)	1 (1%)	10 (5%)
Hemoperitoneum	3 (3%)	6 (5%)	9 (4%)
Wound infection	4 (4%)	1 (1%)	5 (2%)
Respiratory failure	3 (3%)	1 (1%)	4 (2%)
DICa	2 (2%)	2 (2%)	4 (2%)
Sepsis	2 (2%)	1 (1%)	3 (1%)
Pneumonia	2 (2%)	1 (1%)	3 (1%)
Heart failure	2 (2%)	0 (0%)	2 (1%)
SBOb	2 (2%)	1 (1%)	3 (1%)
Liver abscess	1 (1%)	3 (2%)	4 (2%)
Others	8 (8%)	6 (5%)	14 (6%)
Any complication	51 (53%)	27 (22%)	78 (35%)

Data are shown as number of patients (%). Postoperative complications were defined as conditions of Clavien's grade III or greater.

^aDisseminated intravascular coagulopathy.

^bSmall bowel obstruction.

Table 4.

Postoperative in-hospital mortality rates in major liver resection

Surgical procedures	Mortality rates
Left hemihepatectomy − segmentectomy 1	0/5
Left hemihepatectomy + segmentectomy 1	5/20
Extended left hemihepatectomy + segmentectomy 1	2/5
Right hemihepatectomy − segmentectomy 1	2/4
Right hemihepatectomy + segmentectomy 1	0/3
Extended right hemihepatectomy − segmentectomy 1	2/9
Extended right hemihepatectomy + segmentectomy 1	2/10
Overall	23% (13/56)

Accuracy of E-PASS models

Fig. 1 shows the discriminative powers of the E-PASS models to detect in-hospital mortality. AUCs (95% CIs) were 0.85 (0.76–0.94) for E-PASS and 0.82 (0.73–0.91) for mE-PASS; there was no significant difference (P = 0.45). When E-PASS scores were evaluated according to the cause of death, the scores were similarly high in patients with and without liver failure (Table 5). AUCs (95% CIs) of E-PASS for liver failure-associated death was 0.90 (0.84–0.95), whereas that for liver failure-unassociated death was 0.75 (0.60–0.89).

Figure 1.

Receiver operating characteristic curve analysis of E-PASS models used to predict postoperative in-hospital mortality. E-PASS, Estimation of Physiologic Ability and Surgical Stress; mE-PASS, modified E-PASS

Table 5.

Characteristics of deceased patients

	With liver failure	Without liver failure	p value
n	9	12
Age, median (range)	73 (60–82)	72 (53–89)
%Male	78	75
PRS, median (range)	0.62 (0.33–0.70)	0.52 (0.24–0.88)	0.55
SSS, median (range)	0.83 (0.67–1.42)	0.76 (0.27–1.89)	0.52
CRS, median (range)	1.07 (0.80–1.45)	0.95 (0.40–1.90)	0.39

P values for continuous variables were determined by the Mann–Whitney U test.

Fig. 2 illustrates the relationship between the predicted in-hospital mortality rates of the E-PASS models and the severity of postoperative complications. More severe complications were frequently observed in both E-PASS and mE-PASS as the risk band increased. Clavien's grading correlated with the predicted mortality rates in E-PASS (ρ = 0.51, P < 0.001) as well as those in mE-PASS (ρ = 0.47, P < 0.001).

Figure 2.

Relationship between E-PASS models and postoperative morbidity. Four risk bands were set according to the order of the predicted in-hospital mortality rates of Estimation of Physiologic Ability and Surgical Stress (E-PASS) or its modified form, mE-PASS. The risk band 1 indicates the group of patients who are least likely to die, and the risk band 4 indicates the group of patients who are most likely to die. The Clavien grading (CG) system was used to classify postoperative morbidity

Fig. 3 displays the relationship between the CRS values and morbidity with or without major liver resections. In both groups, patients with CRS ≥ 0.8 had a significantly higher postoperative mortality rates than those with CRS < 0.8. There was a trend of higher postoperative complication rate at the CRS of 0.8 or greater in major liver resection, while the rate was significantly higher in procedures without major liver resection. We were able to obtain the data of ICG retention test or 99mTc-galactosyl human serum albumin test in 36 patients undergoing hepatectomy. All the patients showed normal or slightly abnormal values. There were 7 patients who died of postoperative complications. Four patients were associated with liver failure and 3 patients were not. Six out of 7 deceased patients showed the CRS of 0.8 or greater.

Figure 3.

Relationship between Comprehensive Risk Scores and postoperative morbidity according to surgical procedures. Comprehensive Risk Scores (CRS) of the E-PASS were analyzed with or without major liver resection. Major resection indicated hemihepatectomy or extended hemihepatectomy. White box indicates postoperative complications of Clavien's grade III or greater. Black box indicates in-hospital mortality

We further analyzed the relationship between the CRS and postoperative morbidity according to the type of carcinoma (Fig. 4). In both gallbladder carcinoma and perihilar cholangiocarcinoma, higher morbidity and mortality rates were observed at higher CRS risk bands.

Figure 4.

Relationship between Comprehensive Risk Scores and postoperative morbidity according to type of carcinoma. White box indicates postoperative complications of Clavien's grade III or greater. Black box indicates in-hospital mortality

To analyze if the CRS was an independent predictor of postoperative morbidity, we performed a logistic regression analysis (Table 6). We found that CRS was the only independent predictor for postoperative complications of Clavien's grade III or higher.

Table 6.

Logistic regression analysis of the predictors for postoperative complications

	OR	95% CI	p value
Biliary enteric anastomosis	1.58	0.53–4.69	0.41
Major liver resection	0.82	0.27–2.49	0.73
Preoperative biliary drainage	2.07	0.70–6.20	0.19
Preoperative cholangitis	0.51	0.08–3.34	0.49
CRS	9.35	2.09–42	0.0035

Postoperative complications were defined as conditions of Clavien's grade III or higher. Major liver resection indicates hemihepatectomy or more. OR, odds ratio; 95% CI, 95% confidence interval; CRS, Comprehensive risk score of the E-PASS scoring system.

Discussion

In the present study, we analyzed the predictive powers of the E-PASS models in the specialty field of proximal biliary carcinoma resection. The ability of the E-PASS models was high to predict postoperative morbidity and mortality. Similar findings were obtained from the People's Republic of China. Wang et al. demonstrated that E-PASS accurately predicted postoperative mortality in the surgical treatment of hilar carcinoma; the AUCs to detect in-hospital mortality was 0.842 for E-PASS.¹⁶ These findings support the E-PASS models having a high predictive power in this field.

The present study revealed that the E-PASS models predicted both liver failure-associated and -unassociated death. Similar findings were previously obtained for liver carcinoma surgery.¹⁴ A preoperative assessment of future remnant liver is routinely performed in Japan using quantitative liver function tests, such as the indocyanine green test or Tc scintigraphy.¹ When the extent of liver resection is properly selected using this assessment, postoperative liver failure may depend on the balance between a patient's general reserve capacity and the degree of surgical stress that they are subjected to. Overwhelming surgical stress that exceeds a patient's physiological ability to cope will disrupt homeostasis, primarily in the most damaged organ. Hepatectomy involving excessive blood loss and an extended surgical duration will damage the remnant liver, leading to liver failure, even if it preserves the functional capacity of the liver. Therefore, the E-PASS models will be of clinical value for surgical decision-making, in addition to quantitative liver function tests.

The results of the present study may lead to the clinical use of the E-PASS models in surgical decision-making and informed consent. In the preoperative conference, we can calculate the predicted mortality rates. Therefore, surgeons can discuss the presumable procedure according to the anatomical location and extension of carcinoma with its risk. Furthermore, surgeons may explain the risk of surgery to patients and families before operation. Moreover, we previously constructed an integer-based score of E-PASS, termed Total Risk Point (TRP).⁸ TRP can be calculated by just adding the points of E-PASS variables and significantly correlated with the CRS (r = >0.99).⁸ This simple score would enhance the clinical use of this model.

mE-PASS will also be useful for adjusting the risk to the population when assessing the quality of care between centers. If a crude postoperative mortality rate is used as a metric of surgical quality, a hospital that avoids high-risk patients will get a higher ranking, while a hospital welcoming high-risk patients will get a lower ranking. Therefore, case-mix adjustments of the patient population are essential for comparative surgical audits.¹⁰ mE-PASS will be useful for this purpose because it can predict postoperative mortality using physiological variables.⁹ We can use the observed-to-estimated mortality rates ratio (OE ratio) using mE-PASS in a hospital. If the OE ratio is less than 1, that means the observed mortality rate is smaller than the estimated mortality rates, suggesting the better surgical quality. In contrast, if the OE ratio is greater than 1, that means the poorer surgical quality. In this fashion, we could objectively estimate quality of care.

A limitation of this study is that the participating hospitals were mainly medium-volume hospitals. The in-hospital mortality rate of 9% in this study may have been higher than those of previous studies. Liver failure occurred in 9% of the patients undergoing perihilar cholangiocarcinoma resection. Hirano et al.¹ reported that in-hospital mortality was 3.4% after radical resection for hilar cholangiocarcinoma. Carroll et al.² reported that the observed in-hospital mortality rate for all patients undergoing resection for biliary tract cancers in the Nationwide Inpatient Sample 1998–2006 was 5.6%. Our results need to be validated in high-volume centers. Another limitation is that this study was only performed in Japan. Although the high predictive power of E-PASS for hilar carcinoma was previously demonstrated in China,¹⁶ validation studies are desirable in western countries.

In summary, we analyzed the predictive powers of the E-PASS models in biliary carcinoma resection. Our results suggest that the predictive values were promising. The E-PASS models will be useful for surgical decision-making, informed consent, and assessing the quality of care in this field. These efforts would improve quality of surgical performance.

Source of funding

Alliance Family Foundation.

Conflict of interest

None declared.