Trends and Disparities in Regionalization of Pancreatic Resection
- First Online:
- Cite this article as:
- Riall, T.S., Eschbach, K.A., Townsend, C.M. et al. J Gastrointest Surg (2007) 11: 1242. doi:10.1007/s11605-007-0245-5
- 171 Views
The current recommendation is that pancreatic resections be performed at hospitals doing >10 pancreatic resections annually.
To evaluate the extent of regionalization of pancreatic resection and the factors predicting resection at high-volume centers (>10 cases/year) in Texas.
Using the Texas Hospital Inpatient Discharge Public Use Data File, we evaluated trends in the percentage of patients undergoing pancreatic resection at high-volume centers (>10 cases/year) from 1999 to 2004 and determined the factors that independently predicted resection at high-volume centers.
A total of 3,189 pancreatic resections were performed in the state of Texas. The unadjusted in-hospital mortality was higher at low-volume centers (7.4%) compared to high-volume centers (3.0%). Patients resected at high-volume centers increased from 54.5% in 1999 to 63.3% in 2004 (P = 0.0004). This was the result of a decrease in resections performed at centers doing less than five resections/year (35.5% to 26.0%). In a multivariate analysis, patients who were >75 (OR = 0.51), female (OR = 0.86), Hispanic (OR = 0.58), having emergent surgery (OR = 0.39), diagnosed with periampullary cancer (OR = 0.68), and living >75 mi from a high-volume center (OR = 0.93 per 10-mi increase in distance, P < 0.05 for all OR) were less likely to be resected at high-volume centers. The odds of being resected at a high-volume center increased 6% per year.
Whereas regionalization of pancreatic resection at high-volume centers in the state of Texas has improved slightly over time, 37% of patients continue to undergo pancreatic resection at low-volume centers, with more than 25% occurring at centers doing less than five per year. There are obvious demographic disparities in the regionalization of care, but additional unmeasured barriers need to be identified.
KeywordsPancreatic resectionVolume–outcome relationshipRegionalization of care
As for many complex surgical procedures, a strong volume–outcome relationship has been demonstrated in patients undergoing pancreatic resection. Whereas the definition of “high-volume” has varied, multiple studies have shown that surgical mortality, length of stay, hospital charges/costs, and long-term mortality are all decreased when such procedures are performed at high-volume centers1–11 or by high-volume surgeons.2,12,13
Because of the strongly observed volume–outcome relationship, pancreatic resection is often evaluated despite it being a relatively uncommon surgical procedure. As pointed out by Birkmeyer,14 the heavy scrutinization of pancreatic resection is also partly attributable to the high baseline risks associated with the procedure and its usefulness as a prototype for other complex surgical procedures.
As health care reform becomes an increasingly important issue, regionalization of care to high-volume centers specializing in specific complex procedures will be a topic of debate. Regionalization is defined as the delivery of care at a limited number of selected provider sites. Based on the volume-outcomes data for pancreatic resection,1–10 regionalization has been recommended for this procedure. The Leapfrog group, which is a coalition of greater than 150 large public and private health care purchasers, is making efforts to concentrate selected surgical procedures in centers that have the best results.15 In January 2004, pancreatic resection was added to Leapfrog group’s list of procedures targeted for evidence-based referral. For pancreatic resection, the Leapfrog group’s standard for evidence-based referral is strictly based on the process measure of annual volume of procedures performed. They recommend a minimum volume of greater than 10 cases per year.
In the studies evaluating the volume–outcome relationship for pancreatic resection, the percentage of patients resected at low-volume centers ranges from 24% to 77%.1–4,6 As the data supports regionalization and a large percentage of patients are still being resected at low-volume centers, we sought to evaluate trends and disparities in regionalization of pancreatic resection subsequent to the introduction of the concept in the mid-1990s.
This study uses the Texas Hospital Inpatient Discharge database to evaluate temporal trends in the percentage of patients undergoing pancreatic resection at high-volume hospitals throughout the state over the time period 1999 through 2004. Texas was chosen as it serves as a good model for regionalization throughout the United States. Texas has the largest rural population in the United States,16 the highest percentage of people without health insurance,17 and no ethnic majority. One-fifth of the state’s population lives in counties where the whole county has been designated by the U.S. Health Resources and Services Administration as medically underserved.18 As a result, patients often travel large distances to medical centers. We confirmed the volume–outcome relationship for pancreatic resection in Texas by comparing the in-hospital mortality, perioperative lengths of stay, and total charges between low and high-volume hospitals. In addition, we evaluated geographic patterns of referral and regionalization to high-volume centers and performed a multivariate analysis to determine the factors that predict resection at high-volume centers.
Data from the Texas Hospital Inpatient Discharge Public Use Data File (PUDF) for the years 1999 through 2004 are used for this study. The data are collected by the Texas Department of State Health Services, Texas State Health Care Information Center (THCIC), Center for Health Statistics to develop administrative reports on the use and quality of hospital care in Texas.19 The database includes all discharge records for 466 participating non-federal hospitals in Texas. It has 205 data fields in a base data file and 13 data fields in a detailed charges file. The data include patient demographics, hospital information, lengths of stay, ICD-9 diagnosis codes, ICD-9 procedure codes, hospital day of procedure, hospital charges, payer information, and discharge status.
Study Population/Patient Characteristics
ICD-9 Procedure and Diagnosis Codes
ICD-9 Procedure code
Pancreatectomy (total) with synchronous duodenectomy
Pancreaticoduodenectomy, radical (one-stage; two-stage)
Proximal pancreatectomy (head; with part of body; with synchronous duodenectomy)
Distal pancreatectomy (tail; with part of body)
Radical / subtotal pancreatectomy
Pancreatectomy / Pancreaticoduodenectomy partial NEC
ICD-9 Diagnosis code
Malignant neoplasm of the duodenum
Malignant neoplasm of gallbladder
Malignant neoplasm of extrahepatic bile ducts
Malignant neoplasm of ampulla of Vater
Malignant neoplasm other specified sites of gallbladder and extrahepatic bile ducts
Malignant neoplasm of head of pancreas
Malignant neoplasm of body of pancreas
Malignant neoplasm of tail of pancreas
Malignant neoplasm of pancreatic duct
Malignant neoplasm of other specified sites of pancreas
Malignant neoplasm of pancreas, part unspecified
To evaluate trends in regionalization in Texas, patients living out of state (or country) were excluded from the analysis. In addition, patients less than 18 years of age were excluded from the analysis. Age was defined as age groups based on the available data: 18–44 years, 45–54 years, 55–64 years, 65–74 years, and >75 years. These inclusion and exclusion criteria provided a cohort of 3,189 patients who underwent pancreatic resection in Texas between 1999 and 2004, inclusive.
For all patients with zip code data available (n = 3,161), we calculated the following distances: 1) the distance to the hospital at which the surgery was performed, 2) the distance to the nearest high-volume hospital, and 3) the distance to the nearest low-volume hospital.
Independent variables examined in the analysis included patient age group, gender, race/ethnicity (Hispanic, non-Hispanic white, non-Hispanic black, and other), diagnosis (periampullary cancer or other), procedure (distal pancreatectomy, pancreatic head resection, vs. other), year of diagnosis, admission type (emergent or elective), insurance status (uninsured, Medicare/Medicaid, other insurance), and distance to nearest high-volume facility. To control for patients’ comorbidities we used a variable included in the discharge data public use file called “Severity of Illness.” This variable is based on the all-patient refined diagnosis-related grouper (DRG) and considers comorbidity, age, and certain procedures to calculate the “severity of illness” (on a 0–4 scale), with 4 being the most severe. As only two patients had illness severity scores of 0, these were combined with the scores of 1 for the purpose of the analysis.
Hospital Volume/Hospital Characteristics
A Texas hospital was included in the analysis if at least one pancreatic resection was performed there in the 6-year time period. Pancreatic resections performed on patients from out of the state or country were included when determining a hospital’s volume status. Hospitals were then classified into high-volume and low-volume providers based on the 2004 Leapfrog criteria,14 greater than 10 cases per year.
The number of pancreatic resections performed by each hospital each year was examined. The criteria to qualify as a high-volume provider were a minimum volume of more than 10 pancreatic resections per year for 3 of the 6 years of the study and an average volume during the 6-year period of >10 pancreatic resections. Only two hospitals did greater than 10 cases per year for 3 years, but did not meet the average volume requirements to be considered high-volume hospitals. Hospital volume was determined before removing non-Texas residents.
To provide more detail for some analyses on the distribution of pancreatic resections throughout the state, the volume criteria were further subdivided into hospitals that performed less than five resections per year, 5–10 resections per year, 11–19 resections per year, and >20 resections per year. Besides resection at a high-volume center, other outcome variables of interest included in-hospital mortality, the lengths of hospital stay (total and postoperative), and the total hospital charges. Given the nature of the dataset, 30-day mortality could not be determined.
SAS Statistical Software, version 9.1.3 (Cary, NC) was used for all statistical analyses. The percentage of patients undergoing surgical resection at high-volume hospitals each year was calculated. Trends were evaluated for statistical significance using the Cochran-Armitage test for trend.
The patient characteristics, hospital characteristics, and outcome variables were compared between high- and low-volume providers. The primary outcome variable of interest was resection at a high-volume center. Bivariate analyses were used to determine which independent variables were associated with resection at a high-volume center. Significance was accepted at the p < 0.05 level. All means are expressed as mean + standard deviation and all proportions are expressed as percentages. Chi-square analysis was used to compare proportions for all categorical data and t tests were used to compare all continuous variables between the high- and low-volume providers.
A logistic regression model was used to estimate the odds ratio for receipt of surgical resection at a high-volume center. Year and distance to the nearest high-volume center were modeled as a continuous variable. Patient age group, gender, race, diagnosis, illness severity, admission status, insurance status, type of resection, and distance to a high-volume center were used as covariates to determine the independent predictors of surgical resection at a high-volume center. Categorical variables were modeled as a series of binary variables referenced to a single group specified for each variable.
Overall Cohort (n = 3,189)
Severity of illness
Score = 1
Score = 2
Score = 3
Score = 4
Other malignant disease
Other benign disease
Type of operation
Pancreatectomy not otherwise specified
The most common reason for pancreatic resection was periampullary adenocarcinoma, in 57.8% of patients, followed by chronic pancreatitis in 13.6%, other benign disease processes in 14.5%, and other malignant disease processes in 14.1%. 71.6% of resections were performed electively. Distal pancreatectomy was performed in 24.5% of patients (ICD-9 Procedure Code 52.52), whereas pancreaticoduodenectomy was performed in the remaining 75.5% (ICD-9 Procedure Codes 52.51, 52.53, 52.59, 52.6, 52.7; see Tables 1 and 2).
Trends in Resection at High-Volume Centers
Comparison of High- and Low-Volume Centers
Bivariate Comparison of Low- and High-Volume Centersa
Low Volume %
High Volume %
Unadjusted in-hospital mortality
Total length of stay (median)
Total hospital charges
Severity of illness
Score = 1
Score = 2
Score = 3
Score = 4
Type of operation
Pancreatectomy not otherwise specified
Patients undergoing resection at high-volume centers were more likely to be male (50.3 vs. 45.9%, P = 0.02), non-Hispanic white (66.7% vs. 57.6%, P < 0.0001), have non-federal insurance (52.4% vs. 46.5%, P < 0.0001), undergo pancreatic head resection (71.2% vs. 64.1%, P < 0.0001), and to be undergoing elective procedures (79.5% vs. 60.6%, P < 0.0001, Table 3). They were less likely to have periampullary cancer (56.2% vs. 59.9%, P = 0.039). Patients resected at high-volume hospitals had higher “severity of illness” scores (P = 0.0012).
Distance to High-Volume Hospital and Hospital of Surgical Procedure
To further explore factors affecting utilization of high-volume hospitals, we stratified patients by whether they lived within 75 mi of a high-volume hospital. When evaluating the 73.7% of patients who lived within 75 mi (n = 2,329) of a high-volume hospital, 34% (n = 792) were resected by at a low-volume hospital and 66% (n = 1,537) were resected at a high-volume center. Patients resected at a high-volume center traveled further than patients resected at low-volume centers (mean 32.9 vs. 13.1 mi, median 17.8 vs. 7.4 mi, P < 0.0001). In addition, patients resected at high-volume centers often traveled to a high-volume center that was not the closest to their home, with a mean distance of 32.9 + 42.6 mi to the hospital performing the surgery and mean distance of 22.7 + 20.3 mi to the nearest high-volume hospital.
Of the 832 patients who lived more than 75 mi from a high-volume center, they traveled a mean distance of 86.1 + 111.2 mi (median = 37.3 mi) to get to the hospital performing their surgery. The mean distance to a high-volume hospital was 191.7 + 109.8 mi (median = 145.6 mi). Only 36% were resected at high-volume hospitals. Those resected at low-volume hospitals traveled 27.8 + 54.8 mi (median = 8.6 mi) to have their surgery, whereas those resected at high-volume hospitals traveled 188.4 + 111.2 mi (median = 150.6). On average, patients resected at high-volume centers lived closer to the nearest high-volume center that those resected at low-volume centers (149.2 mi vs. 215.9 mi, P < 0.0001).
Multivariate Logistic Regression Analysis
Logistic Regression Analysisa
95% Confidence Interval
Year of diagnosis
Distance to nearest high-volume hospital (by 10 mile increment increases)
Severity of illness
Score = 1
Score = 2
Score = 3
Score = 4
All other diagnoses
Type of operation
Pancreatectomy not otherwise specified
We tested for interactions between “distance to a high-volume hospital” and other covariates, and none were significant. As no significant interactions were identified, we did not stratify patients by distance in the multivariate models.
Regionalization of medical and surgical procedures, especially those procedures that involve large costs and require considerable technical and professional skills, can be expected to improve the quality of medical care and save money. However, the benefits of regionalization must be weighed against the potential detriments including inconvenience to patients (increased travel costs, loss of time from work, constraints on the places where one can receive care),20 the potential for overwhelming of high-volume centers, increased mortality at low-volume hospitals as a result of regionalization, the decreasing quality of urgent related procedures at low-volume hospitals, and reduced access to surgical care if low-volume hospitals cannot recruit qualified surgeons.14
As discussed in his editorial, Birkmeyer14 points out that these concerns are “not very persuasive in the case of pancreaticoduodenectomy” or pancreatic resection in general. Pancreatic resection is an ideal model for regionalization of care for several reasons. First, there is a well-demonstrated, strong volume–outcome relationship. Second, the volume of pancreatic resections performed in this country in a given year are low enough such that high-volume centers would not be overwhelmed. Similarly, the volume lost from shifting these procedures away from low-volume centers would not be detrimental to the low-volume centers, as they occur so infrequently and often cost the hospitals money.
In Texas, the regionalization of pancreatic resection has improved between 1999 and 2004. Of the patients, 54.5% had their pancreatic resection performed at a high-volume center (>11 cases/year) in 1999, and this percentage increased to 63.3% by 2004. This extent of regionalization of pancreatic resection to high-volume centers is similar to the rates seen elsewhere.1,3,6,7,21 The studies are difficult to compare as the volume cutoffs vary. In a 2000 paper by Gouma and colleagues,7 40% of patients were resected at a hospital performing fewer than five pancreatic resections per year. Likewise, in a Maryland study by Gordon et al,1 45.9% of patients were resected at hospitals performing fewer than 20 resections per year. Worse, in a 2003 study of the California and Florida data by Ho and colleagues,3 77% of resections were performed in hospitals doing fewer than 10 resections per year. The Netherlands experience has been similar, with 65% of patients in 1994–1995 undergoing surgery at centers performing 10 or fewer resections per year. In the Netherlands, a plea for regionalization was made, but they were only able to decrease the percentage of patients resected at low-volume centers (<10 cases/year) to 57% in the time period 2000–2003. In a 2002 analysis of the Nationwide Inpatient Sample (NIS),6 the mean number of resections performed at any given hospital in the sample was only 1.5 cases per year. In a more recent analysis of the NIS,21 34.4% of patients were resected at a hospital performing fewer than five resections per year.
Whereas regionalization has increased significantly over the time period of the study, it is still a matter of concern that 26.6% of pancreatic resections in Texas in 2004 were performed at centers doing fewer than five cases per year, and 36.7% were performed at hospitals doing 10 or fewer resections per year. In addition, 19% of patients who were operated on at a low-volume center traveled farther than the distance to the nearest high-volume center.
We also observed interesting geographic patterns in the likelihood of traveling to high-volume centers (see Fig. 5), which are likely applicable to the United States as a whole. The 14 high-volume providers are located in six of Texas’s 254 counties: Dallas, Tarrant, Bell, Harris, Bexar, and Galveston. However, within these counties, high-volume centers have varying levels of monopoly, with people living in Galveston, Bell, and Dallas counties being the most likely to get resected at high-volume centers. The dip and then rise in percentage of patients undergoing resection at a high-volume center based on distance seen in Fig. 4 is likely real. We theorize that big counties with high-volume providers also generate more low-volume providers. As a result, they may be more likely to go to or be referred to one of these providers. However, in the mid-distance suburbs, where fewer low-volume providers exist, referring physicians may be more likely to tell patients that they don’t do such complex procedures and refer them to the high-volume centers in surrounding counties. Therefore, both distance, and referral patterns affect the extent of regionalization.
Outside the principal counties, there are four different situations: (1) suburban rings around the principal counties, (2) middle distance places (such as Texarkana, east Texas, Austin Hill County), (3) remote places (such as south Texas and San Angelo), and (4) very remote west Texas. For the suburbs and middle-distance places, the existence of a local middle-volume provider is key (5–10 cases/year). For example, in Brownsville (South Texas) there is no middle-volume provider. As a result, Brownsville patients are more likely to travel to high-volume providers. McAllen, close to Brownsville, has a middle-volume provider, and few of their people travel to high-volume hospitals. In addition, Brownsville does not refer to McAllen despite its proximity. Beaumont, in East Texas, has a similar situation to Brownsville, with no middle- or high-volume provider, and these patients tend to travel. From San Angelo west, low-volume providers in El Paso, San Angelo, and Lubbock monopolize the market. Here, the very long distance to high-volume providers seems to be a key factor.
Many studies use quartiles or quintiles to establish the volume cutoffs so as to have equal group sizes for statistical analysis when comparing outcomes such as in-hospital mortality, charges, and lengths of stay. For our analysis, we chose to use the Leapfrog group’s minimal volume cutoff for pancreatic surgery to evaluate the extent of regionalization, as this is the current recommendation by a large coalition of payers. Based on the definition we used for “high-volume”, only two hospitals shared the time period with fewer than 11 cases per year, but met the high-volume criteria. Several cities within Texas have middle-volume providers that do not meet the minimum volume requirements, but are clearly referral centers for the geographic area (e.g., Tyler, Lubbock). In these areas, concentration of patients from surrounding very low volume centers would likely bring these centers up to minimum volume standards.
Although we based our volume standards on all resections performed in the state, we eliminated patients who were not from Texas in analyzing the trends in regionalization (although these resections were included in determining a hospital’s volume status). By virtue of the fact that these patients traveled out of state (or country) to have the procedure performed, it is implied that they were an inherently different group of patients. There is also a potential for bias if patients in Texas traveled to nearby cities outside of the state to have their pancreatectomy performed, which we cannot identify. For example, it may be closer for patients in west Texas to travel to Denver, Albuquerque, or Oklahoma City instead of an in-state high-volume provider.
Despite the evidence that regionalization of pancreatic resection is warranted, Texas (and likely other states) are only achieving partial regionalization of care, with greater then 25% of patients being operated on at low-volume centers. In the multivariate analysis of the Texas Data, Hispanic patients were less likely to be resected at high-volume centers. This may be a result of a lack of education regarding the importance of volume for this procedure to this largely Spanish-speaking or bilingual population. The same was not true for Blacks. Older age also seemed to be a barrier to regionalization. Older people may be more reluctant to travel even a minimally further distance to get the best care. However, in the elderly population it is even more critical for the procedure to be performed at high-volume centers so as to minimize complications.
Patients with a diagnosis of periampullary cancer were less likely to be resected at high-volume centers, despite the fact that the highest volume center in Texas is a designated cancer center. This may be the result of hurried decision making and concern of delay when this uncommon diagnosis is made at low-volume centers. The same holds true for emergent procedures, which are far more likely to occur at low-volume centers. Especially in patients with cancer, regionalization to specialized centers will improve both their short- and long-term outcomes. Moreover, there are very few urgent or emergent indications for pancreatic resection and such resections should be minimized.
One of the biggest barriers to regionalization seems to be the distance to a high-volume center. Interestingly, this distance need not be great to influence the choice of hospital. Using the Medicare claims data, Birkmeyer et al.22 have demonstrated that, if not set too high, hospital volume standards could be implemented without imposing unreasonable travel burdens on individuals. This is true for the Texas Discharge Data with a cutoff of 11 or more procedures per year, as 75% of patients live within 75 mi of a high-volume center. However, our study demonstrates that even when the excess travel distance required for surgery at a high-volume center is short, many patients do not go to the high-volume centers. The etiology, however, is unclear, and both patient preference and referral patterns (such as those observed in Brownsville and McAllen) likely influence whether patients go to high-volume centers.
Texas serves as a good model for the regionalization of pancreatic resection. Unlike smaller states in which all patients can easily travel to a high-volume center, Texas is large with many rural areas distant from high-volume centers and would serve as a good model for regionalization to the high-volume centers throughout the U. S. Our data demonstrate that regionalization is feasible and the detailed analysis of the barriers to successful regionalization will aid in achieving this goal. To succeed in regionalizing care for pancreatic resection, we need to change referral patterns such that the 34% of patients living within 75 mi go to high-volume versus low-volume centers. In cities throughout the Texas or the U.S. with medium volume referral centers, cases done by very low-volume providers need to be concentrated at these medium-volume providers. In addition, we can help implement process measures at middle-volume centers that improve outcomes to the level of the high-volume centers (if needed), thereby removing the travel burden for patients.
Most of the volume–outcomes literature focuses on mortality as the endpoint. Although volume is clearly important for good outcomes after pancreatic resection, it is not the whole picture. A recent study by Meguid and colleagues23 demonstrated that volume explained less than 2% of the variance in the data on perioperative death after pancreatic resection. Other endpoints such as length of stay and total charges are also available in some of the published studies, but data on complication rates, morbidity rates, and readmission rates are not readily available. The next step is to further examine “high-volume” providers and measure outcomes more specifically, including the evaluation of common complications after pancreatic surgery including pancreatic fistula,24–33 delayed gastric emptying,24–27,34 intraabdominal abscess formation,24–27 wound infections, bleeding, and others. The goal would then be to standardize care at high-volume institutions through the implementation of critical pathways (which focus on the process measures in the care of patients) designed based on the practices at the institutions with the best outcomes. These data need to be made widely available such that referring physicians, payers, and patients can make informed decisions regarding where to have their pancreatic surgery performed.