Local Referral of High-Risk Pancreatectomy Patients to Improve Surgical Outcomes and Minimize Travel Burden

Abstract

Background:

Referring patients to high-quality hospitals for complex procedures may improve outcomes. This is most feasible within small geographic areas. However, access to specialized surgical procedures may be an implementation barrier. We sought to determine the availability of high-quality hospitals performing pancreatectomy and the potential benefit and travel burden of referral within small geographic areas.

Methods:

We identified elderly Medicare beneficiaries undergoing pancreatectomy between 2012 and 2014. Hospitals were stratified into quintiles of quality based on postoperative complication rates. Patient risk was assessed by modeling the predicted risk of developing a postoperative complication. The geographic unit of analysis was Metropolitan Statistical Area (MSA). Hospitals were categorized into MSA by zip-code. Travel distance was calculated using patient and hospital zip code.

Results:

Among high-risk patients, 40.7% received care at the lowest-quality hospitals even though 80% had a high-quality hospital in the same MSA. Shifting these patients from low- to high-quality hospitals would decrease serious complications from 46.6% to 21.9% (P<0.001) and mortality from 10.9% to 8.9% (P=0.047). Three quarters of high-risk patients treated at low- quality hospitals could reach a high-quality hospital by extending their travel < 5 miles, and nearly 60% traveled farther to a low-quality hospital than was necessary to reach a high-quality hospital.

Conclusions:

High-risk pancreatectomy patients often receive care at low-quality hospitals despite the availability of high-quality hospitals in the area or within an acceptable distance. Referral of high-risk patients to high-quality hospitals within small geographic areas may be an effective strategy to improve outcomes following pancreatic surgery.

Introduction

Postoperative mortality following pancreatectomy has significantly decreased over the past two decades.[1, 2] However, complication rates have remained relatively stable with 30% of patients experiencing a complication.[3, 4] These complications substantially increase spending for pancreatectomy during the index admission and lead to increased post-acute care utilization and disability.[5–7] For oncology patients, postoperative complications have significant repercussions even when they do not result in short-term mortality. Serious complications may delay or prevent receipt of chemotherapy and are associated with inferior long-term survival independent of adjuvant therapy.[8] Receiving surgical care at high-quality, low-morbidity hospitals is critical to improving pancreatectomy outcomes.

Shifting all pancreatectomy patients to high-quality hospitals may be impractical. Previous work evaluating inter-hospital referral for complex cancer surgery revealed that highrisk patients achieve the greatest benefit.[9, 10] This suggests that selective referral strategies should, at least initially, focus on patients at high risk for poor outcomes. However, the majority of studies evaluating selective referral have overlooked the potential travel burden associated with inter-hospital referral, including financial concerns and challenges with post-discharge follow-up.[11–15] Selective referral of high-risk pancreatectomy patients to high-quality hospitals would be most practical within small geographic regions. However, the feasibility, clinical impact, and travel burden of shifting pancreatectomy patients to local high-quality hospitals is not well understood.

To understand the benefits and drawbacks of selective local referral of pancreatectomy patients within small geographic regions, we studied Medicare beneficiaries undergoing elective pancreatectomy. We sought to determine the availability of high-quality hospitals, the distribution of high-risk patient among hospitals, and the clinical benefit and travel burden of shifting high-risk patients to high-quality hospitals within small geographic areas.

Materials and Methods

Data Source

We used data from the Medicare Provider Analysis and Review (MEDPAR) file during the years 2012–2014. Using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes, we included patients aged 65 to 99 years undergoing elective pancreatectomy (ICD-9 code: 5251, 5252, 5253, 5259, 525, 526, 526) for a diagnosis of malignancy. We excluded patients without concurrent Medicare Part A for at least 3 months before and 6 months after surgery. In order to capture only elective operations, we excluded patients with a preoperative length of stay greater than 1 day or an urgent/emergent admission. Hospitals were identified by provider number in the MEDPAR file and linked to data from the American Hospital Association’s (AHA) annual survey.

Metropolitan Statistical Areas

Metropolitan Statistical Areas (MSAs) are defined by the United States Office of Management and Budget as geographic areas with a core population density of >50,000 people and include surrounding counties that share a high degree of economic and social integration.[16] In this analysis, MSAs were defined as Division or Metropolitan Core-Based Statistical Area (CBSA). Hospitals in rural and micropolitan CBSAs were excluded from the analysis. Hospitals were classified into MSAs by their reported CBSA in the AHA dataset.

Hospital Quality Assessment

High-quality and low-quality hospitals were characterized using risk- and reliability- adjusted rates of serious complications within 30 days of the index operation. Postoperative complications were identified using ICD-9 CM codes and included anastomotic, cardiac, genitourinary, hemorrhagic, neurologic, obstruction, postoperative shock, pulmonary, splenic injury, thromboembolic, wound infection, and reoperation.[17] Serious complications were further defined as a postoperative complication associated with a hospital length of stay longer than the 75^th percentile. The length of stay criteria was included in characterizing serious complications, as done is previous studies, to identify complications that had a meaningful clinic impact and add face validity to the complication rate.[8, 18, 19] Serious complications were risk- and reliability-adjusted as detailed below. Hospitals were placed in rank order by adjusted serious complication rate and stratified into ordinal quintiles. Hospitals in the highest two quintiles of complications were defined as low-quality hospitals. Hospitals in the lowest two quintiles of complications were labeled high-quality hospitals.

Patient Risk Assessment

In order to assess how clinical outcomes varied for different patient populations, patients were stratified by their risk of developing a postoperative complication using a multivariable logistic regression model. Complications evaluated in the model included: pulmonary failure, pneumonia, myocardial infarction, deep venous thrombosis or pulmonary embolism, acute renal failure, hemorrhage, surgical site infection, gastrointestinal hemorrhage. Patient sex, procedure code of the index operation, and comorbid conditions were captured and included in the model.[20, 21] Patients in the lowest quintile of risk for developing a complication were labeled “low-risk” and those in the highest quintile were labeled “high-risk”.

Payment Data

In order to determine variation in spending between low- and high-quality for high-risk patients, we evaluated to total surgical episode payment using Medicare payment from the MEDPAR file. Total surgical episode was composed of the index hospitalization, readmission, physician services, and post-discharge ancillary care up to 30 days after the date of discharge. Payments were price-standardized to account for variations in Medicare payments based on care setting and geography.[22]

Assessment of Travel Burden

The distanced traveled to a low-quality hospital and the distance to reach the nearest high-quality hospital was assessed for high-risk patients who underwent pancreatectomy at low- quality centers. Distance was calculated between the geographic center of the patients’ home zip code and the zip code of the low-quality hospital treated at or to the zip code of the nearest high- quality hospital.

Statistical Analysis

We used a multivariable logistic regression model accounting for procedure ICD-9 code, year of the operation, patient age, sex, race, and comorbidities to calculate risk-adjusted rates of serious complications for each hospital.[23, 24] We used hierarchical modeling and empirical Bayes estimates to reliability adjust hospital-based complication rates.[25] The final quintiles of hospital quality were based on the risk- and reliability-adjusted hospital outcomes. High- and low-quality hospitals that were part of a MSA were compared with respect to hospital characteristics using chi-square and Wilcoxon rank-sum tests, as appropriate. Unadjusted perioperative outcomes between patient risk groups and high- and low-quality hospitals within individual MSAs were compared using Pearson χ² tests.

All statistical tests were 2-sided and statistical significance was defined as a P value of <0.05. Statistical analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC) and STATA version 14.0 (StataCorp, College Station, TX).

Results

We identified 18,988 Medicare beneficiaries undergoing pancreatectomy at 1,082 hospitals between 2012 and 2014. Of all pancreatectomy patients, 18,723 (98.6%) received care at one of the 1,014 (93.7%) hospitals located within a MSA (n=287). Among these MSAs, 170 (59.2%) contained a high-quality hospital, and 17,080 (91.2%) of patients were treated within a MSA containing a high-quality center. Low-quality hospitals provided care to 8,344 (44.6%) of all pancreatectomy patients, and 9,843 (52.6%) of patients were treated at high-quality hospitals. There was no substantial difference in hospital referral pattern based on patient risk. The proportion of high-risk patients (n=1,525; 40.7%) who were treated at low-quality hospitals was comparable to that of low-risk patients (n=1,778; 47.4%). Among high-risk patients who received care at a low-quality hospital, 79.1% (n=1,207) had a high-quality hospital within the same MSA. (Figure 1)

Figure 1:

Distribution of patients and hospitals within Metropolitan Statistical Areas

Abbreviation: MSA, Metropolitan Statistical Area; HQH, High-quality hospital; LQH, Low- quality hospital

Patient Characteristics

There were no clinically significant differences in patient age, gender, or race/ethnicity between high-quality and low-quality hospitals. (Table 1) However, high-quality hospitals were more likely to treat patients with multiple comorbidities. Among high-quality hospitals, 56.8% of patients had at least three comorbidities compared to 54.2% of patients at low-quality hospitals (P=0.002). A comparable number of patients at high-quality and low-quality hospitals underwent pancreaticoduodenectomy (61.9% vs 61.8%; P=0.953) and distal pancreatectomy (30.2% vs 30.8%; P=0.327).

Table 1 –

Patient Characteristics Among High- and Low-Quality Hospitals

	HQH	LQH	P
Patients, No.	9843	8344
Age, average years (SD)	73.6 (5.7)	73.7 (5.8)	0.179
Male, No. (%)	4,842 (49.2)	4,138 (49.6)	0.591
White, No. (%)	8,571 (87.1)	7,122 (85.4)	0.001
Comorbidities, No. (%)
0 or 1	2,066 (21.0)	1,839 (22.0)	0.002
2	2,189 (22.2)	1,981 (23.7)
≥3	5,588 (56.8)	4,524 (54.2)
Pancreaticoduodenectomy, No. (%)	6,090 (61.9)	5,159 (61.8)	0.953
Distal pancreatectomy, No. (%)	2,968 (30.2)	2,572 (30.8)	0.327
Other procedure, No. (%)	785 (7.9)	613 (7.4)	0.112

Abbreviation: SD, standard deviation

Hospital Characteristics

Hospitals were characterized based on their risk- and reliability-adjusted serious complication rate. For all patients, low-quality hospitals had a complication rate that was 1.2 times higher than high-quality hospitals (27.5% vs 22.6%; P <0.001) and a serious complication rate that was 1.7 times higher than high-quality centers (15.2% vs 8.9%; P <0.001). High-quality hospitals had significantly lower perioperative mortality compared to low-quality hospitals (3.5% vs 3.7%; P<0.001). (Table 2)

Table 2 -.

Characteristics of High- and Low-Quality Hospitals

	HQH	LQH	P
Hospitals, No.	406	405
Bed size, No. (%)
< 200	453 (4.6)	431 (5.2)	0.001
200–499	3,231 (32.8)	3,233 (38.7)
≥ 500	6,159 (62.6)	4,680 (56.1)
Teaching hospital, No. (%)	9,358 (95.1)	7,970 (95.5)	0.159
Nurse: bed ratio, Average No.	9.4	9.8	<0.001
ICU beds, Average No.	55.4	48.6	<0.001
Annual all-procedure surgical volume, Average No.	11,574	10,911	<0.001
Annual pancreatectomy-specific volume, Average No.	33	36	<0.001
Annual pancreaticoduodenectomy volume, Average No.	22	24	<0.001
*Complication rate, %	22.6	27.5	<0.001
*Serious complication rate, %	8.9	15.2	<0.001
*Mortality rate, %	3.5	3.7	<0.001

Abbreviation: HQH, High-quality hospital; LQH, Low-quality hospital

^*Postoperative outcomes are risk- and reliability-adjusted

High-quality and low-quality hospitals differed on several characteristics. High-quality hospitals were more likely to have a greater number of overall beds and ICU beds. High-quality hospitals also had a larger all-procedure surgical volume (including non-pancreatectomy cases). Low-quality hospitals had a higher pancreatectomy volume; however, the absolute difference between pancreatectomy volume was minimal with high-quality hospitals performing 33 per year and low-quality hospitals performing 36 per year. There was no clinically significant difference in number of pancreaticoduodenectomy performed between high-quality and low- quality hospitals. (Table 2)

Patient Risk, Hospital Quality, and Perioperative Outcomes

Patients were stratified according to their risk of developing a serious postoperative complication. A comparable number of high-risk patients underwent pancreaticoduodenectomy at high-quality and low-quality hospitals (1,157 patients vs 1,114 patients; P=0.851). Similarly, the number of low-risk pancreaticoduodenectomy patients at high-quality and low-quality hospitals did not significantly differ (942 patients vs 950 patients; P=0.108). For both high-risk and low-risk patients, treatment at low-quality hospitals was associated with higher complication and serious complication rates compared to high-quality hospitals. (Table 3) For low-risk patients, low-quality hospitals had a serious complication rate of 4.6% versus 1.7% at high- quality hospitals (P<0.001). Mortality for low-risk patients did not significant differ between high-quality and low-quality hospitals (1.0% vs 0.6%; P=0.159). Episode payments for low-risk patients undergoing pancreatectomy at low-quality hospitals was 3% higher than at high-quality hospitals ($23,551 vs $22837; P<0.001)

Table 3 –

Clinical Outcomes of High-Quality and Low-Quality Hospitals for High Risk and Low Risk Patients

	High-Risk Patients			Low-Risk Patients
	HQH (n=2,151)	LQH (n=l,525)	P	HQH (n=1859)	LQH (n=l,778)	P
Complication rate, No. (%)	1,002 (46.6)	977 (64.0)	<0.001	159 (8.5)	213 (11.9)	0.001
Serious complication rate, No. (%)	472 (21.9)	711 (46.6)	<0.001	31 (1.7)	82 (4.6)	<0.001
Mortality rate, No. (%)	191 (8.9)	166 (10.9)	0.047	18 (1.0)	10 (0.6)	0.159

Abbreviation: HQH, High-quality hospital; LQH, Low-quality hospital

While care delivered in low-quality hospitals was associated with worse outcomes for low-risk patients, the difference in outcomes for high-risk patients treated at high- versus low- quality hospitals was substantially larger. (Table 3) High-risk pancreatectomy patients treated at low-quality hospitals were 2.1 times more likely to experience a serious complication than patients treated at high-quality hospitals (46.6% vs 21.9%; P<0.001). Low-quality hospitals had a postoperative mortality rate of 10.9% for high-risk patients, which was significantly greater than mortality at high-quality hospitals (10.9% vs 8.9%; P=0.047). Episode payments for high- risk patients undergoing pancreatectomy at low-quality hospitals was 31% higher than at high- quality hospitals ($54,717vs $41,724; P<0.001).

Travel Burden

In order to assess the feasibility of selective local referral, we evaluated the distance high- risk patients treated at low-quality hospitals would need to travel to reach a high-quality hospital. The 1,525 high-risk patients who received care at a low-quality hospital traveled an average of 66.7 miles to reach the low-quality facility. A substantial majority of these patients (77.7%) could reach a high-quality hospital by extending their travel less than 5 miles. Notably, 55.9% of high-risk patients treated at a low-quality hospital traveled farther to reach the low-quality hospital than was necessary to reach the closest high-quality hospital. (Figure 2)

Figure 2:

Differences in distance from low-quality hospital to nearest high-quality hospital for all high-risk patients treated at low-quality hospitals

HQH; High-Quality Hospital. LQH: Low-Quality Hospital

A large proportion of MSAs contain high-quality hospitals. Therefore, we next evaluated how far high-risk patients treated at low-quality hospitals in MSAs that also contained a high- quality hospital would need to travel to reach a high-quality center. Of the 1,525 high-risk patients who received care at a low-quality hospital, 79.1% (n=1,207) had a high-quality hospital within the same small geographic area. For these 1,207 patients, 92.3% (n=1,114) could reach a high-quality hospital by extending their travel less than 5 mile. Further, 65.3% (n=788) of these patients could travel a shorter distance to reach the nearest high-quality hospital. (Figure 3) A small proportion of the high-risk patients treated at low-quality hospitals (n=318; 20.9%) would need to cross into another MSA to reach a high-quality hospital. For 19.8% of these patients, even though traveling to a different MSA was necessary, they could still travel a shorter distance to reach a high-quality hospital than was required to reach the low-quality hospital.

Figure 3:

Differences in distance from low-quality hospital to nearest high-quality hospital for high-risk patients treated at low-quality hospitals in MSAs with a high-quality hospital

MSA; Metropolitan Statistical Area, HQH; High-Quality Hospital. LQH: Low-Quality Hospita

Discussion

In our analysis of Medicare beneficiaries undergoing elective pancreatectomy, we found that 40% of the highest-risk patients received care at the lowest-quality hospitals. Interesting, the overwhelming majority of these patients had a high-quality hospital in the same small geographic area. Shifting high-risk patients from low-quality hospitals to local high-quality hospitals could decrease serious complications by 50% and mortality by 20%, highlighting the potential benefit of selective local referral for a large proportion of pancreatectomy patients. Our findings further demonstrate that approximately half of high-risk patients reside closer to a high- quality hospital than the low-quality hospital where they were treated and suggests that this strategy may be achievable without imposing excessive travel burdens on patients. Thus, referring high-risk pancreatectomy patients to local high-quality hospitals may be a feasible strategy to improve the delivery of complex surgical care.

A majority of patients elect to undergo high-risk cancer operations at nearby hospitals regardless of hospital quality. For example, half of pancreatectomy patients prefer receiving care at their local hospital even when the mortality risk is twice that of a regional referral center.[26] Further, a substantial proportion of patients report prohibitive barriers to traveling long distances to regional centers, including difficulty with follow-up care and lack of socioeconomic resources.[15] A portion of patients will choose to travel great distances to reach a specialty centers for surgery. However, our results indicate that long distance travel may not be necessary for patients seeking high-quality care, as over 75% of high-risk patients treated at low-quality hospitals could reach a high-quality hospital by traveling fewer than 5 additional miles. Our findings are consistent with previous work showing that many pancreatectomy patients bypass high-quality hospitals to reach centers where they received care.[11]

Serious complications have significant repercussions for patients and the healthcare system, even if they do not result in mortality.[5–8] Our results demonstrate that shifting high- risk pancreatectomy patients to local high-quality hospitals could decrease postoperative morbidity by up to 53%. Ideally, all patients would receive care at high-quality hospitals. However, attempting to implement such a large-scale strategy may be impractical. As high-risk patients achieve the largest benefit when referred to a high-quality center, a more feasible approach may be to focus selective referral strategies primarily on high-risk patients. It is in payers’ and policymakers’ interest to explore reimbursement structures to divert high-risk patients from low-quality centers. Although our data show that many patients can reach a high-quality hospital with minimal travel burden, there still may be cost associated with this referral strategy. This is especially true for the small proportion of patients required to travel longer distances to reach a high-quality center. Prior work demonstrated significantly reduced spending with selective local referral of colectomy patients, and significant savings from local referral of high-risk pancreatectomy patients would be expected.[27] Thus, reimbursement programs which mitigate travel expenses would more than pay for themselves and prove cost effective.

This analysis should be interpreted in the context of several limitations. First, we used the MEDPAR file which contains administrative claims data and is dependent on accurate and complete coding of the primary and all secondary diagnoses. This database may incompletely capture patient comorbidities and perioperative complications. Further, this database does not contain anatomic information, such as pancreas gland texture and pancreatic duct size, which may be related to the development of postoperative complications. However, it is unlikely that potentially uncaptured characteristics would account for the differences in outcomes among high-quality and low-quality hospitals. Second, estimates of travel burden were calculated as a straight-line distance between two zip codes and may underestimate true travel distances. Additionally, travel times may vary substantially between cities and the number of miles to reach a high-quality hospital may not fully represent travel burden. Finally, other barriers such as the ability of hospitals to accommodate additional high-risk patients and challenges with insurance coverage for patients with private insurance were not addressed in this study. These barriers will need to be evaluated to fully assess the effectiveness and feasibility of selective local referral for pancreatectomy.

Conclusion

The delivery system for complex oncology care in the United States is heterogenous, and a single strategy is unlikely to improve care for all individuals. For pancreatectomy patients, a local referral strategy focused on high-risk patients could significantly improve the quality of their surgical care. This benefit can be achieved without excessive travel burden on patients. While some financial burdens may remain, providing financial support to offset these costs would likely be entirely offset by cost savings to payers. Public and private payers should consider implementation of such a strategy in order to improve the value of complex oncologic surgery such as pancreatectomy.