Pancreatic cancer is a major cause of morbidity and mortality in the Medicare population. Whether the health care burden of pancreatic cancer has changed over the last decade is unknown.
Pancreatic cancer is a major cause of morbidity and mortality in the Medicare population. Whether the health care burden of pancreatic cancer has changed over the last decade is unknown.
The authors used Medicare data from 2000 to 2010 to identify beneficiaries aged ≥ 65 years who were hospitalized for the management of pancreatic cancer. Annual trends were estimated for the age-sex-race–adjusted initial hospitalization rate, the age-sex-race-comorbidity–adjusted 1-year mortality rate after initial hospitalization, age-sex-race-comorbidity–adjusted procedure rates, 1-year all-cause rehospitalizations after initial pancreatic cancer hospitalization, and mean inflation-adjusted Medicare payment for initial hospitalization.
A total of 130,728 patients had ≥ 1 hospitalizations for pancreatic cancer and were identified from 56,642,071 beneficiaries during the study period. The age-sex-race–adjusted rate of initial hospitalization for pancreatic cancer was 50 per 100,000 person-years in 2010, representing a 0.5% annual increase since 2000 (95% confidence interval [95% CI], 0.3%-0.7%). In the same period, the age-sex-race-comorbidity–adjusted 1-year mortality rate decreased by 4.4% (95% CI, 3.9%-4.9%), and the age-sex-race-comorbidity–adjusted surgical resection rate increased by 6.9% (95% CI, 6.4%-7.5%). The mean inflation-adjusted Medicare payment for the initial hospitalization decreased, from $14,118 in 2000 to $13,318 in 2010, and the number of 1-year all-cause rehospitalizations after the initial hospitalization increased from 0.75 per patient in 2000 to 0.82 per patient in 2009 (all P < .001).
For Medicare fee-for-service beneficiaries, initial pancreatic cancer hospitalization, surgical resection, and rehospitalization rates increased, but 1-year mortality rates declined over the last decade. Cancer 2014;120:1050–1058. © 2013 American Cancer Society.
Pancreatic cancer is common and lethal in the elderly population, accounting for 3% of all cancers and 6% of cancer deaths. In 2013, 45,220 new cases of pancreatic cancer are projected to occur in the United States, a relative increase of 3% from 2012.[1, 2] Effective strategies for the early detection of pancreatic cancer are lacking, and most patients are diagnosed at an advanced stage when surgical resection with curative intent is not possible. As a result, the prognosis of pancreatic cancer is poor: > 80% of patients die within a year of diagnosis, and 98% die within 5 years. The majority of patients with pancreatic cancer in the United States are elderly, with a median age at diagnosis of 72 years. The American Cancer Society reports that the pancreatic cancer incidence rate is increasing, with annual increases of 1.0% and 0.8% per year in women and men, respectively, reported since 1998. Continued increases in incidence are expected as the US population continues to age. In January 2013, the Recalcitrant Cancer Research Act was signed into law to accelerate funding for pancreatic cancer research.
To the best of our knowledge, little is known regarding how the health care burden has changed over the last decade for patients with pancreatic cancer, and there is a paucity of data regarding spatial variations in pancreatic cancer treatment and outcomes. Previous studies have reported older data with limited coverage of the entire Medicare population.[6-11] The last decade has brought a multitude of changes that affect pancreatic cancer care, including increased use of abdominal imaging, a general shift toward outpatient oncology care, a renewed emphasis on end-of-life care, and increased scrutiny of the quality and cost-effectiveness of cancer care.[12-16] The effects of these changes on clinical and economic outcomes in pancreatic cancer remain largely unknown.
In the current study, we used the 100% Medicare administrative claims data from the Centers for Medicare and Medicaid Services (CMS) to estimate trends in the burden of pancreatic cancer in the Medicare fee-for-service population from 2000 to 2010 across states/regions and age-sex-race subgroups. We evaluated trends in the initial pancreatic cancer hospitalization rate; the 1-year mortality rate after initial hospitalization; and patterns of care, including 1-year surgical or nonsurgical procedure rates, major discharge dispositions, length of stay (LOS), Medicare expenditures, and postdiagnosis rehospitalizations. The results of the current study describe recent changes in the patterns of care and outcomes for patients with pancreatic cancer and provide important context to help prioritize future effectiveness research for the treatment of this common and lethal malignancy.
Using Medicare beneficiary denominator files from the CMS, we identified all Medicare beneficiaries aged ≥ 65 years who were enrolled in the Medicare fee-for-service plan between January 2000 and December 2010. We calculated person-years for beneficiaries to account for new enrollment, disenrollment, or death during the study period. We linked beneficiaries with Medicare inpatient claims data from the CMS to identify patients with pancreatic cancer with a principal discharge diagnosis of 157.xx, according to the International Classification of Diseases, Ninth Revision, Clinical Modification. We identified all unique patients who had at least 1 hospitalization for pancreatic cancer between January 1, 2000 and December 31, 2010. If a patient had > 1 pancreatic cancer hospitalization during the entire study period, the first hospitalization was selected.
Patient demographic information included age, sex, race (white, black, or other), and the state of residence. Common clinical comorbidities for elderly patients were identified using the hierarchical condition categories method that uses principal and secondary diagnosis codes from hospitalizations within the 12 months before the index hospitalization. Data from 1999 were used for patients hospitalized with pancreatic cancer in 2000.
We defined the pancreatic cancer initial hospitalization rate by dividing the total number of patients with pancreatic cancer in each year by the corresponding person-years of Medicare fee-for-service beneficiaries for that year. We defined the 1-year mortality rate as all-cause deaths within 365 days from the date of the initial pancreatic cancer admission divided by the total number of patients with pancreatic cancer with an index hospitalization within a given year. Procedures included: 1) resections (total pancreatectomy; radical pancreaticoduodenectomy; and proximal, distal, radical subtotal, and partial pancreatectomy); 2) palliative surgical bypass (gastrojejunostomy, biliary-enteric bypass, hepaticojejunostomy, or gastrojejunostomy and biliary-enteric bypass together); and 3) stent/biliary drainage procedures (duct exploration for relief of obstruction other than calculus, insertion of a choledochohepatic tube for decompression, incision of other bile ducts for relief of obstruction, endoscopic insertion of a tube into a bile duct, replacement of a tube in a biliary or pancreatic duct, and biliary drainage [percutaneous or endoscopic]) according to the International Classification of Diseases, Ninth Revision, Clinical Modification codes listed in the online supporting information[18, 19] within 1-year from the initial date of pancreatic cancer hospitalization. Major discharge dispositions included discharge to home, homecare, intermediate-care facility/skilled nursing facility, hospice, and in-hospital death. LOS was defined as the difference between the discharge and admission dates plus 1. Patients with a LOS of > 100 days were considered as outliers and excluded from the LOS analysis. Medicare expenditures were calculated as the mean Medicare reimbursement per patient for the initial hospitalization, adjusting for the annual Consumer Price Index inflation rate reported by the Bureau of Labor Statistics of the United States Department of Labor (bls.gov/data/inflation_calculator.htm) and using the 2000 expenditure as a baseline. The number of rehospitalizations was defined as 1-year cumulative unplanned all-cause hospitalizations after the initial hospitalization. Data from 2010 were used for reporting the 1-year mortality rate, procedure rates, and number of rehospitalizations for patients hospitalized in 2009.
The hospitalization rate at the national and state levels was age-standardized to the 2000 Medicare population aged ≥ 65 years and expressed as per 100,000 person-years. We fitted a mixed-effects model with a Poisson link function and state-specific random intercepts to estimate the trend in the initial hospitalization rate, adjusting for age-sex-race. Time was modeled as an ordinal variable ranging from 0 to 10, corresponding to the years 2000 to 2010. The incidence rate ratio for the time variable was used to represent the age-sex-race–adjusted annual trend in the hospitalization rate. We fitted the mixed-effects model with a logit link function to estimate the trend in the 1-year mortality rate from the initial hospitalization, adjusted for age-sex-race and comorbidities. The time variable in the model was restricted to 0 to 9, corresponding to the years 2000 to 2009. We used the odds ratio of the time variable to represent the risk-adjusted annual trend in 1-year mortality. We repeated this model for each of the 3 surgical and nonsurgical procedure outcomes separately. Using the mixed models, we conducted stratified analyses to estimate trends in hospitalization, 1-year mortality, and 1-year procedure rates by age-sex-race subgroups.
To assess spatial trends and variations in outcomes, we fitted a mixed-effects model with a Poisson link function and state-specific random intercepts and state-specific random slope of the time variable to assess whether a state-specific annual trend in the pancreatic cancer hospitalization rate was significantly different from the national average, adjusting for age-sex-race. A state-specific annual trend was expressed as a deviation of the state-specific slope of the time variable from the national overall slope of the time variable. We used the mixed-effects model with a logit link function to assess state-specific annual trends in 1-year mortality and procedure rates after initial hospitalization, adjusting for age-sex-race-comorbidity and restricting the time variable to 0 to 9. We mapped the age-standardized pancreatic cancer initial hospitalization rates for the years 2000 and 2010 for each state. All analyses were conducted using SAS statistical software (version 9.3, 64-bit; SAS Institute Inc, Cary, NC). All statistical testing was 2-sided, at a significance level of .05. To facilitate data presentation, hospitalization, patient characteristics, mortality, and procedure rates were reported in 2-year intervals as 2000 to 2001, 2005 to 2006, and 2009 to 2010 (2009 for mortality and procedure rates), which represented the baseline, middle, and endpoints of the study period. The Human Investigation Committee at the Harvard School of Public Health determined that Institutional Review Board approval was not required for the current analysis.
From 2000 to 2010, we identified 130,728 unique patients with pancreatic cancer from a total of 56,642,071 Medicare fee-for-service beneficiaries. Over the 11-year period, the mean age at the time of the initial hospitalization increased slightly from 77.7 years (standard deviation [SD], 7.3 years) in 2000 to 77.8 years (SD, 7.7 years) in 2010. The percentage of female patients decreased from 56.2% in 2000 to 54.1% in 2010 (P < .001). See the online supporting information for a summary of patient characteristics and common comorbidities in the 2-year intervals.
The observed initial hospitalization rates increased by all race groups (see online supporting information). The age-standardized initial hospitalization rate increased from 47 per 100,000 person-years to 50 per 100,000 person-years between 2000 and 2010 (Table 1). Among demographic subgroups, black women had the highest hospitalization rate. Figure 1A (Left) shows the age-sex-race–adjusted annual trends, represented by incidence rate ratios, in pancreatic cancer initial hospitalization rates. There was an increase in the hospitalization rate across most subgroups; the risk-adjusted annual increases were 0.5% (95% confidence interval [95% CI], 0.3%-0.7%) for the entire population, 0.4% (95% CI, 0.2%-0.7%) for men, and 0.6% (95% CI, 0.3%-0.9%) for women.
|Patient Groups||Hospitalization Rate per 100,000 Person-Years (95% CI)||1-Year Mortality Rate, % (95% CI)|
|Overall||47 (46.5-47.6)||47 (46.7-47.9)||50 (49.5-50.7)||82.4 (81.8-82.8)||78.9 (78.4-79.4)||78.5 (77.8-79.2)|
|64 y<age<75 y||38 (37.0-38.5)||38 (36.9-38.3)||41 (39.8-41.3)||75.7 (74.8-76.6)||71.5 (70.5-72.4)||69.2 (67.9-70.6)|
|74 y<age<85 y||56 (55.1-57.3)||58 (56.7-58.8)||62 (61.1-63.4)||83.8 (83.1-84.6)||80.0 (79.2-80.8)||80.3 (79.2-81.4)|
|Age >84 y||60 (58.6-62.4)||59 (57.0-60.6)||59 (57.0-60.4)||91.5 (90.6-92.3)||90.4 (89.5-91.2)||91.2 (90.0-92.2)|
|Female||45 (44.1-45.6)||45 (44.7-46.2)||48 (46.8-48.4)||83.1 (82.5-83.8)||79.6 (78.9-80.3)||78.8 (77.8-79.8)|
|Male||50 (49.3-51.2)||50 (49.0-50.7)||53 (52.5-54.3)||81.3 (80.6-82.1)||78.0 (77.2-78.8)||78.1 (77.0-79.2)|
|Black||67 (64.6-69.5)||66 (63.3-68.1)||73 (70.3-75.4)||85.2 (83.8-86.6)||83.3 (81.8-84.8)||82.6 (80.5-84.6)|
|Other race||38 (35.2-39.9)||36 (33.7-37.8)||42 (39.5-43.8)||82.2 (79.7-84.6)||79.5 (77.0-81.9)||78.9 (75.5-82.0)|
|White||46 (45.2-46.4)||46 (45.8-47.0)||49 (48.0-49.3)||82.0 (81.4-82.5)||78.3 (77.8-78.9)||77.9 (77.1-78.7)|
|Black: female||69 (65.8-72.2)||68 (65.1-71.3)||74 (70.8-77.5)||84.0 (82.1-85.8)||82.9 (81.0-84.7)||82.8 (80.0-85.3)|
|Black: male||37 (33.7-39.8)||34 (31.6-37.0)||40 (37.1-42.7)||83.8 (80.4-86.7)||80.1 (76.7-83.3)||77.7 (72.8-82.1)|
|Other race: female||43 (42.2-43.8)||44 (43.2-44.7)||46 (44.8-46.4)||83.0 (82.2-83.7)||79.1 (78.4-79.9)||78.3 (77.2-79.4)|
|Other race: male||64 (60.0-67.9)||62 (58.2-65.6)||71 (67.0-75.0)||87.4 (85.0-89.5)||84.1 (81.6-86.4)||82.3 (78.8-85.5)|
|White: female||39 (35.2-42.5)||38 (34.5-41.0)||44 (40.7-47.4)||80.2 (75.9-83.9)||78.8 (74.9-82.3)||80.1 (75.2-84.5)|
|White: male||50 (48.8-50.8)||50 (48.7-50.6)||53 (51.6-53.6)||80.8 (79.9-81.6)||77.4 (76.5-78.2)||77.5 (76.3-78.7)|
The age-standardized hospitalization rates varied substantially by state and over time (Fig. 2), ranging from a low of 21 (Puerto Rico) to 62 (District of Columbia) per 100,000 person-years in 2000 (interquartile range [IQR], 42-48) and from 24 (Wyoming) to 67 (Pennsylvania) per 100,000 person-years in 2010 (IQR, 40-53). In the age-sex-race–adjusted analyses, hospitalization rates increased significantly in 9 states and decreased significantly in 5 states between 2000 and 2010 (Fig. 1B Top).
The observed 1-year mortality rate decreased from 82.7% in 2000 to 78.5% in 2009 (P < .001 for trend), a relative decline of 5.1%. The age-sex-race-comorbidity–adjusted annual decrease in the 1-year mortality rate was 4.4% (95% CI, 3.9%-4.9%). This decline was similar across all the major age-sex-race subgroups (Fig. 1A Right). The observed 1-year mortality rate varied by state and over time; in 2000, the 1-year mortality rate ranged from 74.4% (Utah) to 94.3% (Nevada) (IQR, 81.2%-83.9%) and in 2009 it ranged from 65.2% (Montana) to 92.3% (Wyoming) (IQR, 76.1%-80.8%). After risk adjustment, there was no evidence of a significant variation in trends for mortality across states (Fig. 1B Bottom).
From 2000 to 2009, the observed surgical resection rate increased significantly from 14.4% to 22.3% (P < .001 for trend). In contrast, both bypass and stent/biliary drainage rates decreased from 15.2% to 8.1% (P < .001 for trend) and 33.3% to 29.8% (P < .001 for trend), respectively (see online supporting information).
The age-sex-race-comorbidity–adjusted annual increase in the surgical resection rate after initial hospitalization was 6.9% (95% CI, 6.4%-7.5%). For bypass and stent/biliary drainage, the age-sex-race-comorbidity–adjusted annual declines were 7.7% (95% CI, 7.1%-8.3%) and 1.8% (95% CI, 1.4%-2.3%), respectively. The surgical resection rate increased significantly and the bypass rate decreased significantly across all age-sex-race subgroups. The stent/biliary drainage rate also decreased significantly for all age-sex-race subgroups, except for nonwhite male patients (Fig. 3A). There was slight-to-modest spatial variation in procedure rates from 2000 to 2009 (see online supporting information).
The observed 1-year mortality rate was low among patients undergoing surgical procedures and it changed over the study period. Between 2000 and 2009, the 1-year mortality rate declined from 44.9% to 36.7% for surgical resection (P < .001), from 79.6% to 70.7% for bypass (P < .001), and from 83.6% to 84.1% for nonsurgical biliary drainage (P = .5143). The 1-year mortality rate was 91.3% in 2000 versus 92.3% in 2009 (P = .0598) for patients who received none of these procedures.
From 2000 to 2010, the mean LOS decreased from 10.7 days (SD, 9.0 days) to 9.1 days (SD, 7.7 days), and the in-hospital mortality rate declined by a relative 38%, from 12.6% to 7.8%. The rates of discharge to hospice increased significantly from 2.1% to 19.3%, whereas discharge to skilled nursing facilities decreased from 15.0% to 12.8% (all P < .001 for trend) (Fig. 3B). The inflation-adjusted mean Medicare payment for the initial pancreatic cancer hospitalization decreased slightly from $14,118 (SD, $15,381) in 2000 to $13,318 (SD, $17,011) in 2010 (P < .001 for trend). The number of 1-year all-cause rehospitalizations after the initial hospitalization for pancreatic cancer increased from 0.75 per patient in 2000 to 0.82 per patient in 2009 (P < .001 for trend).
In this large observational study of trends in initial hospitalization, mortality, and patterns of care for Medicare fee-for-service beneficiaries with pancreatic cancer, we found increasing hospitalization rates, with substantial variation across age-sex-race subgroups and states. The state-level initial hospitalization rates ranged from 21 to 62 per 100,000 person-years in 2000 and 24 to 67 per 100,000 person-years in 2010. Overall, there was a 0.5% annual increase in the age-sex-race–adjusted hospitalization rate from 2000 to 2010. Given the 45,220 new cases of pancreatic cancer projected for 2013, the 0.5% annual increase in the initial hospitalization rate for the elderly translates to approximately 1,600 additional Medicare patients hospitalized for pancreatic cancer over the next 10 years period. These hospitalizations translate to an increase of at least $21 million in Medicare expenditures, irrespective of post-discharge treatments. Although the initial hospitalization rates increased, the age-sex-race-comorbidity–adjusted 1-year mortality rate had an annual decline of 4.4%. There was minimal variability across states in 1-year mortality.
Significant changes were also observed in the patterns of care for pancreatic cancer over the last decade. Over time, patients were more likely to be discharged to hospice and less likely to have long hospital stays or to die in the hospital, and the cost of initial pancreatic cancer hospitalization decreased by $800 from 2000 to 2010. We observed that the surgical resection rate increased significantly but palliative procedures, including surgical bypass and stent/biliary drainage procedures, declined significantly over the study period. We found that the use of surgical bypass declined much faster than the use of stenting and/or nonsurgical biliary drainage. This finding suggests a potentially important shift in patterns of care for patients with pancreatic cancer, with curative-intent surgical resection substituting for palliative surgical bypass. Nevertheless, other hospital-related factors, such as cancer volume, specificity (eg, cancer center vs noncancer center), and bed size, might impact this shift. This topic requires further examination. The increase in the surgical resection rate, which could reflect secular trends in surgical techniques or treatment approaches, as well as the improvement in the outcome after surgical resection, might have contributed to the decrease noted in the 1-year mortality rate for pancreatic cancer. Nevertheless, the surgical resection rate remains low; in 2009, only 22.2% of patients underwent a surgical resection procedure. The low rate of surgical resection could reflect a lack of a technique to detect pancreatic cancer earlier, but poor access to hospitals performing these complex surgeries may limit the broader uptake of surgical resection among patients with pancreatic cancer. Finks et al estimated that from 1999 to 2008, the number of Medicare patients requiring surgery to treat pancreatic cancer increased by 50%, but the number of hospitals performing the procedures decreased by 25%.
Geographic variations in cancer incidence have been well studied and documented. The results of the current study further demonstrate that the trend in the initial hospitalization rate for patients with pancreatic cancer varied significantly by state. In contrast, trends in treatments and outcomes for patients with pancreatic cancer were homogeneous across states and regions. The increase in resection rates over time was observed in nearly all states and the decrease in the 1-year mortality rate was also noted across all states concurrently. The current study findings illustrate that for trends in pancreatic cancer, the between-state variance in the occurrence of cases is much bigger than the between-state variances in the treatments and outcomes for patients with this cancer. Over the past decade, there were changes in lifestyles in the United States. These changes, such as the decrease in the size of the smoking population, the decline in the consumption of red meat, and the improvement in diabetic control,[21-23] could all contribute to a reduced risk of pancreatic cancer, and these risk factors may vary substantially by geographic and demographic subpopulations. However, a treatment, such as surgical resection, is adapted rapidly once it becomes available.
The findings of the current study complement analyses of Surveillance, Epidemiology, and End Results (SEER) data and present a robust snapshot of the population-level burden of this deadly malignancy nationwide. Our state-level analysis has provided a better picture of spatial variations in the prevalence and patterns of care for patients with pancreatic cancer. The state-/region-specific findings can help providers to understand disease burden and may provide evidence for disease control strategies. The trends we described can also be used to allocate resources and to benchmark progress in the control of this highly lethal malignancy. Our finding of an increase in the initial hospitalization rate for patients with pancreatic cancer is consistent with previous studies,[24-27] although the finding of a substantial decline in the 1-year mortality rate contrasts with results from a recent study that reported little improvement in pancreatic cancer survival over the past 30 years.
Extensive nationwide efforts have gone forward to improve cancer care over the last decade since the Institute of Medicine released its influential report, Ensuring Quality Cancer Care, in 1999. The oncology community in particular has shown significant increases in the development of, adoption of, and support for evidence-based guidelines in clinical practice. The findings of the current study provide a better image of the patterns of pancreatic cancer care in the Medicare fee-for-service population, and inform researchers and policymakers about what is being achieved and what can be done better in the care of patients with pancreatic cancer.[29, 30]
The current study has limitations that merit mention. We were limited to the Medicare fee-for-service population, and therefore the findings may not reflect the changes in the managed care population, a group that tends to be healthier. We accessed only the inpatient data and could not identify individuals with outpatient diagnoses of pancreatic cancer. Nevertheless, the vast majority of patients treated for pancreatic cancer will have at least 1 hospitalization, and a sensitivity analysis using the 2007 SEER-Medicare data found that > 98.5% of Medicare fee-for-service patients were hospitalized after receiving a diagnosis of pancreatic cancer. The lack of cancer registry information in the current study data precluded our ability to account for the stage or histology of the cases of pancreatic cancer in our risk adjustments. It also precluded our ability to assess the changes in the use of other nonsurgical therapies, such as radiotherapy and chemotherapy, which could impact on the mortality rate as well. Our 1-year mortality rate measured survival from the date of the initial hospitalization for pancreatic cancer, a time point that may not correspond to the date of diagnosis. Nevertheless, the difference in magnitude between the 1-year mortality rate from pancreatic cancer admission and from pancreatic cancer diagnosis appears to be small. The 1-year mortality rate for patients aged ≥ 65 years reported from a SEER-Medicare analysis was reported to be 81.4% between 1988 and 2008, whereas we found a 1-year mortality rate of 80.9% from 2000 to 2009. Finally, because we relied on billing codes used by hospitals to obtain reimbursement, we cannot be certain whether the variation we described represents significant differences in disease burden or regional variations in coding practices. The very large size of our data set and the availability of data over multiple years suggest that our findings are reflective of meaningful outcomes.
In conclusion, between 2000 and 2010, the initial hospitalization rate for pancreatic cancer increased considerably within the Medicare fee-for-service population, whereas the 1-year mortality rate declined significantly over the same period. Patients were more likely to undergo curative-intent surgical resection, substituting for palliative surgical bypass. The health and economic burden of pancreatic cancer in the United States is substantial, and additional nationwide efforts are required to fight this disease.
No specific funding was disclosed.
Drs. Dominici and Wang are partially funded by the National Cancer Institute (P01 CA134294; Principal Investigator: Dr. Lin and co-Principal Investigator: Dr. Dominici) and the Agency for Healthcare Research and Quality (K18 HS021991 to Dr. Dominici).