Defining a Hospital Volume Threshold for Minimally Invasive Pancreaticoduodenectomy in the United States | Gastrointestinal Cancer | JAMA Surgery | JAMA Network
[Skip to Navigation]
Sign In
Figure.  Smoothed Restricted Cubic Spline Plot of the Adjusted Log Odds Ratio of Experiencing Any Complication vs the Annual Number of Minimally Invasive Pancreaticoduodenectomy Cases Performed per Hospital
Smoothed Restricted Cubic Spline Plot of the Adjusted Log Odds Ratio of Experiencing Any Complication vs the Annual Number of Minimally Invasive Pancreaticoduodenectomy Cases Performed per Hospital

The curved line with long dashes represents the regression line in the change point estimation. The 2 lighter dotted curves represent the 95% CIs. The black dots correspond to the location of 3 knots used in the model. The intersection at the value of 22 cases per year (dotted blue vertical line) is the cutoff identified by the model, with adjustment for the effects of patient age, sex, race/ethnicity, comorbidities, year of diagnosis, and clinical diagnosis.

Table 1.  Demographic, Clinical, and Hospital Characteristics of Patients Undergoing Minimally Invasive Pancreaticoduodenectomy
Demographic, Clinical, and Hospital Characteristics of Patients Undergoing Minimally Invasive Pancreaticoduodenectomy
Table 2.  Unadjusted Patient Outcomes and Inflation-Adjusted Costs by Hospital Volume Status
Unadjusted Patient Outcomes and Inflation-Adjusted Costs by Hospital Volume Status
Table 3.  Summary of Adjusted Outcomes for Patients Treated at Low-Volume (≤22 Cases per Year) vs High-Volume (>22 Cases per Year) Hospitalsa
Summary of Adjusted Outcomes for Patients Treated at Low-Volume (≤22 Cases per Year) vs High-Volume (>22 Cases per Year) Hospitalsa
1.
Gagner  M, Pomp  A.  Laparoscopic pylorus-preserving pancreatoduodenectomy.  Surg Endosc. 1994;8(5):408-410.PubMedGoogle ScholarCrossref
2.
Cameron  JL, Riall  TS, Coleman  J, Belcher  KA.  One thousand consecutive pancreaticoduodenectomies.  Ann Surg. 2006;244(1):10-15.PubMedGoogle ScholarCrossref
3.
Hyder  O, Dodson  RM, Nathan  H,  et al.  Influence of patient, physician, and hospital factors on 30-day readmission following pancreatoduodenectomy in the United States.  JAMA Surg. 2013;148(12):1095-1102.PubMedGoogle ScholarCrossref
4.
Adam  MA, Choudhury  K, Dinan  MA,  et al.  Minimally invasive versus open pancreaticoduodenectomy for cancer: practice patterns and short-term outcomes among 7061 patients.  Ann Surg. 2015;262(2):372-377.PubMedGoogle ScholarCrossref
5.
Croome  KP, Farnell  MB, Que  FG,  et al.  Total laparoscopic pancreaticoduodenectomy for pancreatic ductal adenocarcinoma: oncologic advantages over open approaches?  Ann Surg. 2014;260(4):633-638.PubMedGoogle ScholarCrossref
6.
Clinical Outcomes of Surgical Therapy Study Group.  A comparison of laparoscopically assisted and open colectomy for colon cancer.  N Engl J Med. 2004;350(20):2050-2059.PubMedGoogle ScholarCrossref
7.
Ohtani  H, Tamamori  Y, Noguchi  K,  et al.  A meta-analysis of randomized controlled trials that compared laparoscopy-assisted and open distal gastrectomy for early gastric cancer.  J Gastrointest Surg. 2010;14(6):958-964.PubMedGoogle ScholarCrossref
8.
Vanounou  T, Steel  JL, Nguyen  KT,  et al.  Comparing the clinical and economic impact of laparoscopic versus open liver resection.  Ann Surg Oncol. 2010;17(4):998-1009.PubMedGoogle ScholarCrossref
9.
Sharpe  SM, Talamonti  MS, Wang  CE,  et al.  Early national experience with laparoscopic pancreaticoduodenectomy for ductal adenocarcinoma: a comparison of laparoscopic pancreaticoduodenectomy and open pancreaticoduodenectomy from the National Cancer Data Base.  J Am Coll Surg. 2015;221(1):175-184.PubMedGoogle ScholarCrossref
10.
Fong  ZV, Chang  DC, Ferrone  CR, Lillemoe  KD, Fernandez Del Castillo  C.  Early national experience with laparoscopic pancreaticoduodenectomy for ductal adenocarcinoma: is this really a short learning curve?  J Am Coll Surg. 2016;222(2):209.PubMedGoogle ScholarCrossref
11.
American Medical Association. 2015 Annual meeting: reports of the board of trustees. https://www.ama-assn.org/sites/default/files/media-browser/public/hod/a15-bot-reports.pdf. Accessed December 4, 2016.
12.
Tran  TB, Dua  MM, Worhunsky  DJ, Poultsides  GA, Norton  JA, Visser  BC.  The first decade of laparoscopic pancreaticoduodenectomy in the United States: costs and outcomes using the Nationwide Inpatient Sample.  Surg Endosc. 2016;30(5):1778-1783.PubMedGoogle ScholarCrossref
13.
Tran Cao  HS, Lopez  N, Chang  DC,  et al.  Improved perioperative outcomes with minimally invasive distal pancreatectomy: results from a population-based analysis.  JAMA Surg. 2014;149(3):237-243.PubMedGoogle ScholarCrossref
14.
Ejaz  A, Sachs  T, He  J,  et al.  A comparison of open and minimally invasive surgery for hepatic and pancreatic resections using the Nationwide Inpatient Sample.  Surgery. 2014;156(3):538-547.PubMedGoogle ScholarCrossref
15.
Deyo  RA, Cherkin  DC, Ciol  MA.  Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases.  J Clin Epidemiol. 1992;45(6):613-619.PubMedGoogle ScholarCrossref
16.
Charlson  ME, Pompei  P, Ales  KL, MacKenzie  CR.  A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.  J Chronic Dis. 1987;40(5):373-383.PubMedGoogle ScholarCrossref
17.
US Department of Labor. Bureau of Labor Statistics CPI inflation calculator. http://www.bls.gov/data/inflation_calculator.htm. Accessed November 7, 2015.
18.
Adam  MA, Thomas  S, Youngwirth  L,  et al.  Is there a minimum number of thyroidectomies a surgeon should perform to optimize patient outcomes? [published online March 8, 2016].  Ann Surg. doi:10.1097/SLA.0000000000001688PubMedGoogle Scholar
19.
Desquilbet  L, Mariotti  F.  Dose-response analyses using restricted cubic spline functions in public health research.  Stat Med. 2010;29(9):1037-1057.PubMedGoogle Scholar
20.
Adam  MA, Pura  J, Goffredo  P,  et al.  Presence and number of lymph node metastases are associated with compromised survival for patients younger than age 45 years with papillary thyroid cancer.  J Clin Oncol. 2015;33(21):2370-2375.PubMedGoogle ScholarCrossref
21.
Birkmeyer  JD, Dimick  JB.  Potential benefits of the new Leapfrog standards: effect of process and outcomes measures.  Surgery. 2004;135(6):569-575.PubMedGoogle ScholarCrossref
22.
Zureikat  AH, Moser  AJ, Boone  BA, Bartlett  DL, Zenati  M, Zeh  HJ  III.  250 Robotic pancreatic resections: safety and feasibility.  Ann Surg. 2013;258(4):554-559.PubMedGoogle Scholar
23.
Boone  BA, Zenati  M, Hogg  ME,  et al.  Assessment of quality outcomes for robotic pancreaticoduodenectomy: identification of the learning curve.  JAMA Surg. 2015;150(5):416-422.PubMedGoogle ScholarCrossref
24.
Speicher  PJ, Nussbaum  DP, White  RR,  et al.  Defining the learning curve for team-based laparoscopic pancreaticoduodenectomy.  Ann Surg Oncol. 2014;21(12):4014-4019.PubMedGoogle ScholarCrossref
25.
Shakir  M, Boone  BA, Polanco  PM,  et al.  The learning curve for robotic distal pancreatectomy: an analysis of outcomes of the first 100 consecutive cases at a high-volume pancreatic centre.  HPB (Oxford). 2015;17(7):580-586.PubMedGoogle ScholarCrossref
26.
Birkmeyer  JD, Siewers  AE, Finlayson  EV,  et al.  Hospital volume and surgical mortality in the United States.  N Engl J Med. 2002;346(15):1128-1137.PubMedGoogle ScholarCrossref
Original Investigation
April 2017

Defining a Hospital Volume Threshold for Minimally Invasive Pancreaticoduodenectomy in the United States

Author Affiliations
  • 1Department of Surgery, Duke University Medical Center, Durham, North Carolina
  • 2Department of Biostatistics, Duke University, Durham, North Carolina
  • 3Duke Clinical Research Institute, Durham, North Carolina
JAMA Surg. 2017;152(4):336-342. doi:10.1001/jamasurg.2016.4753
Key Points

Question  What is the minimum number of minimally invasive pancreaticoduodenectomy (MIPD) cases performed by a hospital that is associated with the lowest risk for complications?

Findings  In this review of 865 cases of MIPD, we found that increasing hospital procedural volume of MIPD is associated with improved outcomes up to 22 cases per year.

Meaning  The identified threshold of 22 cases per year may serve as a foundation for protocols aimed at safer implementation of MIPD at the national level and may have implications for surgical education and training.

Abstract

Importance  There is increasing interest in expanding use of minimally invasive pancreaticoduodenectomy (MIPD). This procedure is complex, with data suggesting a significant association between hospital volume and outcomes.

Objective  To determine whether there is an MIPD hospital volume threshold for which patient outcomes could be optimized.

Design, Setting, and Participants  Adult patients undergoing MIPD were identified from the Healthcare Cost and Utilization Project National Inpatient Sample from 2000 to 2012. Multivariable models with restricted cubic splines were used to identify a hospital volume threshold by plotting annual hospital volume against the adjusted odds of postoperative complications. The current analysis was conducted on August 16, 2016.

Main Outcomes and Measures  Incidence of any complication.

Results  Of the 865 patients who underwent MIPD, 474 (55%) were male and the median patient age was 67 years (interquartile range, 59-74 years). Among the patients, 747 (86%) had cancer and 91 (11%) had benign conditions/pancreatitis. Overall, 410 patients (47%) had postoperative complications and 31 (4%) died in-hospital. After adjustment for demographic and clinical characteristics, increasing hospital volume was associated with reduced complications (overall association P < .001); the likelihood of experiencing a complication declined as hospital volume increased up to 22 cases per year (95% CI, 21-23). Median hospital volume was 6 cases per year (range, 1-60). Most patients (n = 717; 83%) underwent the procedure at low-volume (≤22 cases per year) hospitals. After adjustment for patient mix, undergoing MIPD at low- vs high-volume hospitals was significantly associated with increased odds for postoperative complications (odds ratio, 1.74; 95% CI, 1.03-2.94; P = .04).

Conclusions and Relevance  Hospital volume is significantly associated with improved outcomes from MIPD, with a threshold of 22 cases per year. Most patients undergo MIPD at low-volume hospitals. Protocols outlining minimum procedural volume thresholds should be considered to facilitate safer dissemination of MIPD.

Introduction

The introduction of minimally invasive surgical techniques in the realm of pancreatic surgery has lagged behind, and this is especially true for pancreaticoduodenectomy.1 Pancreaticoduodenectomy is one of the more complex surgical procedures, with increased risks for postoperative morbidity and mortality.2,3 The complexity of this procedure stems from the difficulty in accessing the retroperitoneal location of the pancreas and the need for multiple anastomotic reconstructions. Pancreaticoduodenectomy anastomotic reconstruction via a minimally invasive approach is technically challenging even for experienced pancreatic surgeons.1 Anastomotic pancreatic leak and/or fistula can lead to life-threatening complications such as hemorrhage, sepsis, and death.

There has been an increasing interest in the use of minimally invasive pancreaticoduodenectomy (MIPD).4,5 This enthusiasm is driven by the potential benefits from minimally invasive techniques seen in other gastrointestinal procedures such as esophageal, gallbladder, and colon surgical procedures.6-8 While retrospective data from hospitals that perform high numbers of MIPDs may suggest the safety of minimally invasive vs open surgery,5 the generalizability of these outcomes to hospitals that perform a small number of minimally invasive cases has recently been questioned. National-level studies that included low- and high-volume hospitals have demonstrated that MIPD is associated with increased 30-day mortality when the procedure is done at hospitals that perform a small number of MIPD cases.4,9 Evidence suggesting an association between lower hospital procedural volume and compromised outcomes from MIPD4,5,9 has prompted debate about the safety of the procedure and a call for potentially restricting the procedure to high-volume hospitals.10 The shift of the health care system from a fee-for-service to a value-based model that links hospitals’ reimbursement to the quality of care provided has made this debate even more revelant.11 Therefore, determining a minimum number of MIPDs that potentially defines a high-volume hospital would seem essential. This study sought to determine whether there is a minimum number of MIPD cases performed by a hospital that is associated with a reduced likelihood of postoperative complications.

Methods
Data Source

Conducted on August 16, 2016, this retrospective analysis was performed from hospital discharge data of patients undergoing minimally invasive (laparoscopic and robotic) pancreaticoduodenectomy between 2000 and 2012 in the Health Care Utilization Project National Inpatient Sample data sets (HCUP-NIS). The HCUP-NIS is maintained by the Agency for Healthcare Research and Quality and represents a stratified 20% sample of inpatient discharges to acute care hospitals across the United States. The data set includes information on patient demographic, insurance, disease, treatment, and hospital characteristics.

Study Cohort

The study included adult patients who underwent MIPD for benign or malignant pancreatic disease from HCUP-NIS using the International Classification of Diseases, Ninth Revision, Clinical Modification procedure codes 52.7 and 52.51. Laparoscopic and robotic modifier codes were used to identify patients who underwent MIPD.12-14

Patient demographic characteristics, such as age, sex, race/ethnicity, and payer were obtained from the data set. Pancreatic diagnoses were determined using the corresponding International Classification of Diseases, Ninth Revision diagnosis code(s), which were then grouped into 3 major categories: pancreatic cancer, pancreatitis, and other benign conditions. Patient comorbidities were characterized using a modified Charlson/Deyo scoring system and categorized into 3 groups: 0, 1, and 2 or greater.15,16 Annual hospital volume was calculated for each hospital as the number of MIPDs reported by a hospital per year.

In-hospital complications were defined from secondary International Classification of Diseases, Ninth Revision diagnosis and procedure codes corresponding to the index surgical admission. Complications were grouped into bleeding, cardiopulmonary, thromboembolic, gastrointestinal, urologic, infection, wound, reoperation, in-hospital mortality, and any complication. Hospital costs were calculated by multiplying hospital charges from the discharge records by the specific HCUP-NIS hospital cost-to-charge ratios. Costs were then adjusted to 2014 US dollars using rates from the Bureau of Labor Statistics Consumer Price Index inflation calculator.17 This study was granted exempt status by the Duke University institutional review board.

Statistical Analysis

Our primary end point was occurrence of 1 or more in-hospital complications. Secondary outcomes included the incidence of individual complications, hospital length of stay, and inflation-adjusted costs.

A multivariable logistic regression model with restricted cubic splines (RCSs) was used to specify and estimate the functional form of annual hospital volume with respect to the incidence of any complication.18 The RCS statistical method provides a flexible model to examine the adjusted effect of a continuous predictor on an outcome and allows for visualization of the relationship without prior knowledge of the functional form of the association.19 Using this model, we identified a range of annual hospital volumes that corresponded to a change in the relative log odds of experiencing any postoperative complication. A bootstrap simulation that incorporated a Monte Carlo Markov Chain procedure was conducted to estimate the point that corresponded to the maximum change from this range of annual hospital volumes.20 The following factors were controlled for in this multivariable model: age, sex, race/ethnicity, comorbidities, year of the procedure, and pancreas-related diagnosis.

Based on the identified annual hospital volume threshold, the cohort was dichotomized into 2 groups: high-volume hospitals (>threshold hospital volume) and low-volume hospitals (≤threshold hospital volume). Multivariable logistic regression was used to examine the adjusted association between low- vs high-volume hospitals and incidence of any complication, and negative binomial regression was used to examine the association with hospital length of stay and costs. These models were built in the generalized estimating equation framework to account for the correlation of outcomes for patients treated at the same hospital.

Sensitivity Analysis

To examine the potential effect of hospital volume for open pancreaticoduodenectomy cases on the hospital volume threshold for MIPD, we ran a separate RCS model in which hospital volume for open cases was added, in addition to other covariates. Hospital volume for open cases was represented in the model as follows: low-volume hospitals (<11 cases per year) and high-volume hospitals, based on the Leapfrog-defined threshold.21

A 2-sided significance level of .05 was used for all statistical tests. Statistical analyses were performed using SAS version 9.4 (SAS Institute).

Results

The study included 865 patients who underwent MIPD between 2000 and 2012. Most patients (n = 747; 86%) underwent the procedure for a cancer diagnosis. Details of patient demographic and clinical characteristics are described in Table 1.

Hospital Procedural Volume

There were 229 hospitals that performed MIPD during the study period. The number of MIPD cases performed by a hospital ranged from 1 to 60 cases per year, with a median of 6 cases per year (interquartile range, 2-15 cases). The median number of open pancreaticoduodenectomies performed by these hospitals was 27 cases per year (interquartile range, 9-62 cases), with most cases (71%) performed at high-volume hospitals (≥11 open cases per year).

Overall, use of MIPD increased by more than 400% over the study period, from 14 cases in 2000 to 126 cases in 2012. Most cases (57%) were reported from hospitals that performed fewer than 10 cases per year, and 20% of cases were reported from hospitals that performed just 1 case per year.

Hospital Volume and Complications

In total, 410 patients (47%) experienced a postoperative complication; of these, 41% experienced 1 complication, 34% had 2 complications, and 24% had at least 3 complications (Table 2). After adjustment for patient demographics, comorbidities, and clinical diagnosis, hospital procedural volume was significantly associated with decreasing odds of experiencing a postoperative complication (overall association P < .001). The adjusted RCS plot demonstrated a nonlinear association between hospital procedural volume and postoperative complications, which confirmed the existence of a possible hospital volume threshold corresponding to the lowest risk of experiencing a postoperative complication. This plot showed that increasing hospital procedural volume was significantly associated with decreasing odds of patients experiencing a postoperative complication for up to 22 cases per year (Figure).

The threshold identified by the adjusted RCS model was then internally validated using a bootstrap analysis, in which the RCS model was reexamined in 1000 random replicates of the initial data sets. A Markov chain Monte Carlo simulation incorporated with bootstrapping was then used, which identified a best threshold for hospital procedural volume at 22 cases per year (95% CI, 21-23 cases per year).

Outcomes by Hospital Volume Status

Minimally invasive pancreaticoduodenectomy hospital volume status was then defined based on the threshold identified by the RCS and simulation models: (1) low-volume hospitals (≤22 cases per year) and high-volume hospitals (>22 cases per year). Most patients (n = 717; 83%) underwent MIPD at low-volume hospitals, and only 17% of patients (n = 148) underwent surgery at high-volume hospitals. Patient demographic and clinical characteristics by hospital volume status are characterized in Table 1. Compared with patients treated at high-volume hospitals, patients treated at low-volume hospitals were more often nonwhite (14% vs 7%; P = .002). Patient comorbidities and clinical diagnosis were not significantly different between the low- and high-volume groups. All cases in the minimally invasive high-volume group were performed at high-volume hospitals for open pancreaticoduodenectomy. Of the low-volume group, 65% of the minimally invasive cases were performed by open pancreaticoduodenectomy high-volume hospitals (Table 1). The study included different hospital types. Most patients were treated at teaching hospitals (n = 686; 79%); however, all patients in the high-volume group were treated at teaching centers, while just 75% of patients (n = 538) in the low-volume group were treated at teaching centers.

Compared with patients treated at high-volume hospitals, those who were treated at low-volume hospitals more often experienced delayed gastric emptying (7.7% for low-volume vs 2.7% for high-volume hospitals; P = .03), fistula formation (11.8% vs 6.1%; P = .04), cardiopulmonary complications (19.1% vs 12.2%; P = .046), or any complication (50.3% vs 33.1%; P < .001). The number of postoperative complications also was higher in patients treated at low- vs high-volume hospitals (Table 2). Median length of hospital stay was longer for patients in the low-volume group compared with those in the high-volume group (10 vs 9 days, respectively; P < .001). Inflation-adjusted hospital costs were not different between groups.

After adjustment for patient demographic and clinical characteristics, undergoing MIPD at low- vs high-volume hospitals was associated with increased odds of experiencing postoperative complications (odds ratio, 1.74; 95% CI, 1.03-2.94; P = .04) and a trend toward longer length of hospital stay (+18% increase; 95% CI, −3% to +44%; P = .09) (Table 3).

Sensitivity Analysis

After adjustment for patient demographics, comorbidities, clinical diagnosis, and hospital procedural volume for open pancreaticoduodenectomy cases, hospital volume for minimally invasive cases was significantly associated with decreasing odds of experiencing a postoperative complication up to 22 cases per year. The threshold identified by the adjusted RCS model was then internally validated, identifying a best threshold for hospital procedural volume at 22 cases per year (95% CI, 21-23 cases per year).

Discussion

To our knowledge, this is the first study to objectively determine a hospital procedural volume threshold for MIPD at the national level. After adjustment for patient demographic and clinical characteristics, higher hospital procedural volume was significantly associated with decreasing odds of patients experiencing a postoperative complication. This adjusted analysis identified a hospital procedural volume threshold of 22 cases per year that appears to be associated with the lowest risk for experiencing a postoperative complication. Patients undergoing surgery at low-volume hospitals (≤22 cases per year) were more likely to experience postoperative complications. Although use of MIPD increased more than 400% between 2000 and 2012, most cases are still performed at low-volume hospitals.

Minimally invasive pancreaticoduodenectomy is a technically challenging and complex surgical procedure, with significant variation in postoperative outcomes described across different hospitals. Published retrospective series from hospitals that perform a large number of MIPD cases may suggest that the procedure is safe.5,22 In a single-surgeon experience with 108 laparoscopic pancreaticoduodenectomies, Croome et al5 reported that the procedure is as safe as open pancreaticoduodenectomy and associated with shorter hospitalization, faster recovery, and more frequent use of adjuvant chemotherapy. However, these favorable outcomes from MIPD could not be generalized to hospitals performing a fewer number of cases.4,9 In a nationwide analysis, Adam et al4 compared outcomes from minimally invasive vs open pancreaticoduodenectomy performed mostly at low-volume hospitals; this analysis demonstrated that minimally invasive surgery was associated with increased 30-day mortality, without the benefit of shorter hospital length of stay.

Our study is aligned with other reports4,9 that have demonstrated a significant association between hospital volume and improved outcomes from MIPD. While the underlying mechanisms behind the hospital volume–outcomes association have not been fully determined, this association seems to be clinically intuitive. High-volume hospitals potentially have more experienced pancreatic and minimally invasive surgeons, developed intensive care units, structured processes for postoperative care, and readily available resources.5,23-25 These factors likely would facilitate better patient selection for the procedure, a lower likelihood of technical errors, and an enhanced ability to recognize postoperative complications and rescue patients experiencing major complications.26

Published data have demonstrated that MIPD is associated with inferior outcomes when the procedure is performed at low-volume hospitals, with increased 30-day mortality.4,9 However, the definition of a hospital volume threshold remains unclear. Thus, identification of a minimum number of MIPD cases that are needed to define a high-volume hospital is important for facilitating the safe implementation of the procedure.

While the volume threshold of 22 cases per year seems to be higher than high-volume definitions used in previous studies examining MIPD,4,9 it is imperative to highlight that our threshold number of cases was objectively determined based on the adjusted association between hospital volume and postoperative outcomes. Prior studies have not specifically examined a volume threshold for MIPD; instead, volume status definitions were arbitrarily assigned.4,9,12 We used 2 rigorous statistical methods to increase the validity of our identified threshold.

Although our sensitivity analysis demonstrated a minimal effect for hospital volume for open pancreaticoduodenectomy cases, it important to note that there was a significant correlation between hospital volume for minimally invasive and open pancreaticoduodenectomy. This may emphasize the importance of experience with open pancreaticoduodenectomy before implementation of MIPD. Therefore, interpretation of the threshold number of cases identified by this study should be contingent on adequate experience with open pancreaticoduodenectomy.

The threshold of 22 cases per year likely reflects the complexity of MIPD, which can be more challenging than open pancreaticoduodenectomy. The Leapfrog evidence–based hospital referral standards recommend a minimum of 11 cases per year for (open) pancreatectomy.21 Therefore, the hospital volume threshold for MIPD might be expected to be more than 11 cases per year, as suggested by other reports.23,24 Speicher et al24 reported their initial experience among 56 patients who underwent laparoscopic pancreaticoduodenectomy at a single institution by a team of fellowship-trained laparoscopic and pancreatic surgeons. After 50 cases, operative times and estimated blood loss were lower than those for open cases.24 Another study reported the first experience with 200 robotic pancreaticoduodenectomy cases performed by a team of surgeons at a single center. The learning curve necessary to achieve proficiency for robotic pancreaticoduodenectomy was determined to be 80 cases over 6 years.23 While these numbers reported from institutional series appear to be higher than our threshold of 22 cases per year, it is important to point out that these numbers represent unique institutional experiences over a number of years.

While surgeon volume has been shown to be associated with improved surgical outcomes for some surgical procedures, hospital volume may be more meaningful for complex procedures such as MIPD. Hospital volume often is associated with surgeon volume, but it also reflects, at least in part, the structure of support within hospitals that incorporate systems issues such as critical care and postoperative management. Birkmeyer et al26 examined the effect of hospital volume on surgical mortality from 14 different types of cardiovascular and major cancer procedures, including pancreatectomy. They found that the association between hospital volume and decreased mortality varied by type of surgery. Absolute differences in adjusted mortality rates between the lowest- and highest-volume hospitals was the highest (12%) for pancreatectomy, superseding other complex procedures such as esophagectomy (5%), pneumonectomy (5%), or gastrectomy (2%-5%).26

Limitations and Strengths

There are several limitations to our study, including the possibility of coding errors. Tumor characteristics are not captured by the HCUP-NIS database; therefore, they could not be analyzed. The study is retrospective in nature, with concerns for selection bias between low- vs high-volume hospitals. However, it is expected that high-volume hospitals would take on more complex cases, which may negatively affect their outcomes compared with low-volume hospitals. This study was based on data mostly representing the earliest experiences with MIPD in the United States, which could affect interpretation of the results of our study. Determination of a hospital volume threshold may evolve as surgeons gain more experience with the procedure. Although our sensitivity analysis accounted for the effect of hospital volume for open cases, there was a significant correlation between hospital volume for open and minimally invasive cases. This emphasizes the need for experience with open cases before implementation of MIPD. The HCUP-NIS represents a 20% sample of hospital discharges; therefore, these data may not reflect the absolute number of minimally invasive cases performed in the United States. Data about surgeon volume were not complete; thus, the independent effect of surgeon vs hospital volume could not be examined. While the rates of overall complications and mortality from minimally invasive surgery are consistent with published series, the incidence of postoperative pancreatic fistula was low; however, it is reasonable to assume that the underreporting of this complication would be universal across volume groups.

Strengths of the study include the large sample size and the inclusion of low- and high-volume hospitals at the national level.

Conclusions

This large study provides valuable information with regard to the importance of hospital volume in implementation of MIPD. While it is known that hospital volume is associated with improved outcomes from MIPD, our study determined the minimum number of procedures that is associated with the lowest risk for postoperative complications. This finding is timely and relevant, given the ongoing debate regarding safe implementation of this complex procedure and the current shift toward value-based health care reimbursement models.10 The identified threshold of 22 cases per year may serve as a foundation for protocols aimed at safer implementation of MIPD at the national level and may have implications for surgical education and training.

Back to top
Article Information

Corresponding Author: Julie A. Sosa, MD, MA, Section of Endocrine Surgery, Department of Surgery, Duke University Medical Center, No. 2945, Durham, NC 27710 (julie.sosa@duke.edu).

Accepted for Publication: August 29, 2016.

Published Online: December 28, 2016. doi:10.1001/jamasurg.2016.4753

Author Contributions: Drs Adam and Sosa had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Adam, Youngwirth, Pappas, Roman, Sosa.

Acquisition, analysis, or interpretation of data: Adam, Thomas, Youngwirth, Roman, Sosa.

Drafting of the manuscript: Adam, Youngwirth.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Adam, Thomas, Youngwirth.

Administrative, technical, or material support: Roman, Sosa.

Study Supervision: Youngwirth, Pappas, Roman, Sosa.

Conflict of Interest Disclosures: Dr Sosa is a member of the Data Monitoring Committee of the Medullary Thyroid Cancer Consortium Registry supported by Novo Nordisk, GlaxoSmithKline, AstraZeneca, and Eli Lilly. No other disclosures were reported.

Funding/Support: This work was supported in part by grant P30CA014236 from the National Institutes of Health.

Role of the Funder/Sponsor: The funder had a role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, and approval of the manuscript; and decision to submit the manuscript for publication.

Previous Presentation: A portion of the data was presented at the Society of Surgical Oncology Annual Cancer Symposium; March 3, 2016; Boston, Massachusetts.

Additional Contributions: We acknowledge the help of Mathias Worni, MD (Duke University), in preparation of coding for the analysis. He did not receive compensation.

References
1.
Gagner  M, Pomp  A.  Laparoscopic pylorus-preserving pancreatoduodenectomy.  Surg Endosc. 1994;8(5):408-410.PubMedGoogle ScholarCrossref
2.
Cameron  JL, Riall  TS, Coleman  J, Belcher  KA.  One thousand consecutive pancreaticoduodenectomies.  Ann Surg. 2006;244(1):10-15.PubMedGoogle ScholarCrossref
3.
Hyder  O, Dodson  RM, Nathan  H,  et al.  Influence of patient, physician, and hospital factors on 30-day readmission following pancreatoduodenectomy in the United States.  JAMA Surg. 2013;148(12):1095-1102.PubMedGoogle ScholarCrossref
4.
Adam  MA, Choudhury  K, Dinan  MA,  et al.  Minimally invasive versus open pancreaticoduodenectomy for cancer: practice patterns and short-term outcomes among 7061 patients.  Ann Surg. 2015;262(2):372-377.PubMedGoogle ScholarCrossref
5.
Croome  KP, Farnell  MB, Que  FG,  et al.  Total laparoscopic pancreaticoduodenectomy for pancreatic ductal adenocarcinoma: oncologic advantages over open approaches?  Ann Surg. 2014;260(4):633-638.PubMedGoogle ScholarCrossref
6.
Clinical Outcomes of Surgical Therapy Study Group.  A comparison of laparoscopically assisted and open colectomy for colon cancer.  N Engl J Med. 2004;350(20):2050-2059.PubMedGoogle ScholarCrossref
7.
Ohtani  H, Tamamori  Y, Noguchi  K,  et al.  A meta-analysis of randomized controlled trials that compared laparoscopy-assisted and open distal gastrectomy for early gastric cancer.  J Gastrointest Surg. 2010;14(6):958-964.PubMedGoogle ScholarCrossref
8.
Vanounou  T, Steel  JL, Nguyen  KT,  et al.  Comparing the clinical and economic impact of laparoscopic versus open liver resection.  Ann Surg Oncol. 2010;17(4):998-1009.PubMedGoogle ScholarCrossref
9.
Sharpe  SM, Talamonti  MS, Wang  CE,  et al.  Early national experience with laparoscopic pancreaticoduodenectomy for ductal adenocarcinoma: a comparison of laparoscopic pancreaticoduodenectomy and open pancreaticoduodenectomy from the National Cancer Data Base.  J Am Coll Surg. 2015;221(1):175-184.PubMedGoogle ScholarCrossref
10.
Fong  ZV, Chang  DC, Ferrone  CR, Lillemoe  KD, Fernandez Del Castillo  C.  Early national experience with laparoscopic pancreaticoduodenectomy for ductal adenocarcinoma: is this really a short learning curve?  J Am Coll Surg. 2016;222(2):209.PubMedGoogle ScholarCrossref
11.
American Medical Association. 2015 Annual meeting: reports of the board of trustees. https://www.ama-assn.org/sites/default/files/media-browser/public/hod/a15-bot-reports.pdf. Accessed December 4, 2016.
12.
Tran  TB, Dua  MM, Worhunsky  DJ, Poultsides  GA, Norton  JA, Visser  BC.  The first decade of laparoscopic pancreaticoduodenectomy in the United States: costs and outcomes using the Nationwide Inpatient Sample.  Surg Endosc. 2016;30(5):1778-1783.PubMedGoogle ScholarCrossref
13.
Tran Cao  HS, Lopez  N, Chang  DC,  et al.  Improved perioperative outcomes with minimally invasive distal pancreatectomy: results from a population-based analysis.  JAMA Surg. 2014;149(3):237-243.PubMedGoogle ScholarCrossref
14.
Ejaz  A, Sachs  T, He  J,  et al.  A comparison of open and minimally invasive surgery for hepatic and pancreatic resections using the Nationwide Inpatient Sample.  Surgery. 2014;156(3):538-547.PubMedGoogle ScholarCrossref
15.
Deyo  RA, Cherkin  DC, Ciol  MA.  Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases.  J Clin Epidemiol. 1992;45(6):613-619.PubMedGoogle ScholarCrossref
16.
Charlson  ME, Pompei  P, Ales  KL, MacKenzie  CR.  A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.  J Chronic Dis. 1987;40(5):373-383.PubMedGoogle ScholarCrossref
17.
US Department of Labor. Bureau of Labor Statistics CPI inflation calculator. http://www.bls.gov/data/inflation_calculator.htm. Accessed November 7, 2015.
18.
Adam  MA, Thomas  S, Youngwirth  L,  et al.  Is there a minimum number of thyroidectomies a surgeon should perform to optimize patient outcomes? [published online March 8, 2016].  Ann Surg. doi:10.1097/SLA.0000000000001688PubMedGoogle Scholar
19.
Desquilbet  L, Mariotti  F.  Dose-response analyses using restricted cubic spline functions in public health research.  Stat Med. 2010;29(9):1037-1057.PubMedGoogle Scholar
20.
Adam  MA, Pura  J, Goffredo  P,  et al.  Presence and number of lymph node metastases are associated with compromised survival for patients younger than age 45 years with papillary thyroid cancer.  J Clin Oncol. 2015;33(21):2370-2375.PubMedGoogle ScholarCrossref
21.
Birkmeyer  JD, Dimick  JB.  Potential benefits of the new Leapfrog standards: effect of process and outcomes measures.  Surgery. 2004;135(6):569-575.PubMedGoogle ScholarCrossref
22.
Zureikat  AH, Moser  AJ, Boone  BA, Bartlett  DL, Zenati  M, Zeh  HJ  III.  250 Robotic pancreatic resections: safety and feasibility.  Ann Surg. 2013;258(4):554-559.PubMedGoogle Scholar
23.
Boone  BA, Zenati  M, Hogg  ME,  et al.  Assessment of quality outcomes for robotic pancreaticoduodenectomy: identification of the learning curve.  JAMA Surg. 2015;150(5):416-422.PubMedGoogle ScholarCrossref
24.
Speicher  PJ, Nussbaum  DP, White  RR,  et al.  Defining the learning curve for team-based laparoscopic pancreaticoduodenectomy.  Ann Surg Oncol. 2014;21(12):4014-4019.PubMedGoogle ScholarCrossref
25.
Shakir  M, Boone  BA, Polanco  PM,  et al.  The learning curve for robotic distal pancreatectomy: an analysis of outcomes of the first 100 consecutive cases at a high-volume pancreatic centre.  HPB (Oxford). 2015;17(7):580-586.PubMedGoogle ScholarCrossref
26.
Birkmeyer  JD, Siewers  AE, Finlayson  EV,  et al.  Hospital volume and surgical mortality in the United States.  N Engl J Med. 2002;346(15):1128-1137.PubMedGoogle ScholarCrossref
×