Trends in Palliative Care, Hospice Utilization, and Outcomes in Hospitalized Pancreatic Cancer Patients: A Nationwide Analysis

Background and objective The prognosis of pancreatic cancer (PC) is generally poor. PC responds only modestly to chemotherapy and chemoradiation, and surgical resection remains the only curative option. The risk of recurrence is high. PC patients are encountered in the hospital on initial diagnosis and later for surgeries and complications from PC. We analyzed PC hospitalizations in the United States as reported in the National Inpatient Sample (NIS) database from 2005 to 2011 to determine the extent to which aggressive interventions could be avoided, thereby decreasing the cost of hospitalization. We analyzed trends in palliative care utilization and hospice services. Methods The International Classification of Disease 9th Revision (ICD-9) codes were used to identify diagnoses and procedures performed. Weighted analysis was performed using SPSS Statistics 28.0 (IBM Corp., Armonk, NY). Dispositions at discharge were noted. Complications and procedures performed were also documented. Results A total of 574,522 cases with PC were identified. Trends are reported chronologically (2005 to 2011). Over time, inpatient deaths for PC have decreased (11.2%, 11.1%, 9.8%, 9.8%, 9.5%, 8.4%, 8.1%; p<0.001), and hospice discharges (HD) have increased (10.2%, 11.4%, 11.4%, 12,2%, 12.6%, 12.4%, 12.7%; p<0.001). Palliative care utilization has increased (2.9%, 3.9%, 3.8%, 5.6%, 8.8%, 10.2%, 11.9%; p<0.001). Complications including peritonitis, thrombosis, hypovolemia/shock, and acute kidney injury (AKI) have increased mortality rates and HD. Conclusion There is an increasing trend of palliative care and hospice service utilization among hospitalized PC patients. Until better-targeted treatments and screening become available, mortality and morbidity will remain high. The proportion of patients receiving aggressive interventions remains high and is associated with poor outcomes. It is desirable to conduct palliative care evaluation (PCE) early in patients with advanced disease and avoid aggressive interventions.


Introduction
Pancreatic cancer (PC) is a malignancy with high mortality and a five-year survival rate of around 4-6% [1,2]. The incidence of PC has been rising with each passing year [2]. PC is often diagnosed in the late stages of the disease, as it is often asymptomatic in the early stages [3]. The only curative option is surgical resection of resectable tumors [2,4]. PC is not very responsive to chemotherapy and radiation therapy. Thus, advanced unresectable tumors inevitably have a poor prognosis. This is reflected by the fact that the fiveyear mortality of PC closely parallels its incidence [1,5]. Even among patients who are cured by surgery, the rate of recurrence is high, with five-year survival ranging from 15 to 25% [5][6][7]. PC patients are encountered in the hospital on initial presentation with nonspecific symptomatic disease and later for surgeries or due to complications arising from the treatments or disease progression. Frequent hospitalizations and rehospitalizations indicate poor quality of care for patients with advanced cancers [8,9]. Indicators of aggressive care near death include chemotherapy, ED visits, hospitalization (including death during hospitalization), and ICU admissions [10][11][12][13][14][15]. PC is one of the leading cancers in terms of frequency of ED visits and hospitalizations in the last six months and last two weeks of life [9]. Palliative care consultation is associated with less aggressive care near the end-of-life (EOL) in patients with PC [10,16]. Palliative care interventions are associated with improved quality of life and satisfaction near EOL [16][17][18][19][20][21].

Study design
In this cross-sectional retrospective analysis, we analyzed PC hospitalizations reported in the National Inpatient Sample (NIS) database from 2005 to 2011 to find out what aggressive interventions could be avoided in patients, thereby decreasing hospitalization costs and increasing the utilization of palliative care and hospice services.

Primary and secondary outcomes
The study's primary outcome was the disposition of the patient at discharge from the hospital. We classified patients into three groups based on their discharge disposition: died, hospice discharge (HD), and other discharges (which included all other discharges). The secondary outcomes of the study were the complications arising from PC and interventions performed during the hospitalization, as mentioned below under Methods.

Methods
NIS is an administrative database, a part of the Healthcare Cost and Utilization Project (HCUP). NIS is the largest publicly available all-payer inpatient healthcare database designed to generate US regional and national estimates of inpatient utilization, access, cost, quality, and outcomes. Unweighted, it contains data from more than seven million hospital stays each year. Weighted, it estimates more than 35 million hospitalizations nationally. NIS consists of de-identified patient information that is nationally available, which was submitted for the Institutional Review Board (IRB) review. The Thomas Jefferson University Hospital IRB determined that our study did not require IRB approval and permitted us to proceed with the analysis.
Adult cases with the International Classification of Disease 9th Revision (ICD-9) diagnosis codes for PC were identified from the database for the years 2005 to 2011. Furthermore, ICD-9 diagnosis codes were used to identify patients who had metastases of various organs and lymph nodes. ICD-9 procedural codes were used to identify cases with diagnostic procedures such as laparoscopy, biopsies, and other diagnostic procedures of the abdomen, including endoscopic retrograde pancreatography. Treatments and interventions were identified using ICD-9 procedure codes for various forms of pancreatectomy, and chemotherapy administration. Common aggressive interventions including transfusion of red cells and other blood products, intravenous anticoagulation, thrombolytics, vasopressors, an inferior vena cava filter (IVCF) insertion, mechanical ventilation (MV), dialysis, and parenteral nutrition were identified. Common complications arising from PC such as deep vein thrombosis (DVT), pulmonary embolism (PE), peritonitis, and complications relating to the stomach, duodenum, gall bladder, biliary system, and pancreas were identified. Other complications such as cachexia, hypovolemia, shock, acute kidney injury (AKI), and depression were also identified. In addition, cases that received palliative care evaluation were identified. The codes used to identify these conditions can be reviewed in Appendix 1.
Charlson Comorbidity Index (CCI) was calculated using a combination of indicators for chronic conditions as reported in the NIS database and ICD-9 diagnosis codes. CCI is a well-validated measure of comorbidity used to predict one-year mortality in patients [22,23].

Statistical analysis
SPSS Statistics for Windows, Version 28.0 (IBM Corp., Armonk, NY) was used for data analysis. Chi-square tests were utilized to check the association between different categorical variables of interest and outcomes of the hospitalizations. Mann-Whitney U tests were used to test the association between non-normally distributed continuous variables, and median values were reported. Trends were checked using the Cochran-Armitage test of trend. For the values reported in the tables, all p-values were significant (p<0.05) unless otherwise stated. The non-significant values are marked with an asterisk (*). All the results are statistically weighted using DISCWT (weight of discharges in the universe) as provided by HCUP in the NIS database.

Results
A total of 574,522 cases of PC were identified. Of those, 55,095 (9.6%) patients died during the hospital stay, with 68,497 (11.9%) HD; 293,747 (51.1%) of the cases had metastatic disease. Metastatic disease was associated with increased mortality compared to cases without metastasis (11.8% and 7.3%, respectively, p<0.001).
Mortality and HD increased consistently with age, as seen in Table 1. As for sex distribution, 50.5% were male and 49.5% female. Males had higher mortality than females (10.2% vs. 9%, p<0.001), while females had higher HD (12.5% in females and 11.3% in males, p<0.001). The mean age at admission for cases that died was 69.50 years (SD: 12.20), and the mean age for those with HD was 71.18 years (SD: 12.43). The mean CCI was 5.34 (SD: 3.04) for cases that died and 5.59 (SD: 2.97) for HD.  We analyzed the impact of the primary site of PC on in-hospital outcomes. The primary part of the pancreas was not specified for 49.3% of the cases. For the cases in which it was specified, the pancreatic head was the most commonly reported location of primary PC (159,557 cases, 27.8%). The effect of the site of primary PC on death in the hospital and HD can be reviewed in Table 2.  We next looked at the site of metastases and their relationship to in-hospital outcomes, as reported in Table  3. Notable findings are as follows. Metastases were associated with increased HD compared to nonmetastatic PC cases (15.6% with metastases and 8.1% without metastases, p<0.001). The liver was the most common site of metastasis (33.5%). Liver metastases were associated with increased death in the hospital (12.0% with liver metastases and 8.4% without metastases, p<0.001) and increased HD (16.2% with liver metastases and 9.6% without liver metastases, p<0.001). Metastases of other abdominal organs were reported in 13.8% of the cases. Other sites of metastasis with a sizeable frequency were the thoracic cavity [including lungs (7.9% for lungs only), mediastinum, pleura, and other respiratory organs; 8.7% total], and bone/bone marrow (3.4%). Thoracic cavity metastasis was associated with increased death in the hospital (14.6% with thoracic metastases and 9.1% without thoracic metastases, p<0.001) and increased HD (17.9% with thoracic metastases and 11.4% without thoracic metastases, p<0.001). Bone metastasis was also associated with increased death in the hospital (14.6% with bone metastases and 9.4% without bone metastases, p<0.001) and increased HD (18.2% with bone metastases and 11.7% without bone metastases, p<0.001). Brain and spinal cord (CNS) metastases were reported in only 0.6% of the cases but were associated with higher deaths in the hospital (17.2% with CNS metastases and 9.5% without CNS metastasis, p<0.001) and increased HD (21.7% with CNS metastases and 11.9% without CNS metastases, p<0.001).  We also looked into the impact of other medical issues and complications on in-hospital outcomes, which can be reviewed in Table 4. Notable results are as follows. DVT and PE were reported in 5.4% and 4% of cases, respectively. DVT was associated with increased death in the hospital (11.5%) compared to cases without DVT (9.5%, p<0.001), and increased HD (16.4% with DVT and 11.7% without DVT, p<0.001). PE was associated with increased death in the hospital (13.6%) compared to cases without PE (9.4%, p<0.001), and increased HD (16.1% with PE and 11.7% without PE, p<0.001). IVCF insertion was reported in 2% of the total PC cases, with no significant difference in deaths in the hospital (9.7% with IVCF insertion and 9.6% without IVCF insertion, p=0.676). IVCF insertion was associated with higher HD (16.8% with IVCF insertion and 11.8% without IVCF insertion, p<0.001). Oral anticoagulation was reported in 4.2% of cases and was associated with decreased death in the hospital (6.7% with anticoagulation and 9.7% without anticoagulation, p<0.001) and increased HD (12.6% with anticoagulation and 11.9% without anticoagulation, p<0.001). Mechanical ventilation (MV) was reported in 3.1% of cases, with higher deaths in the hospital (51.7% with MV and 8.2% without MV, p<0.001) but low HD (7.8% with MV and 12.1% without MV, p<0.001). The associations with other interventions can be reviewed in Table 5.   AKI (reported in 10.3% of cases) was associated with increased death in the hospital (27.2% with AKI and 7.6% without AKI, p<0.001) and increased HD (17.5% with AKI and 11.3% without AKI, p<0.001). Hypovolemia/shock (4.4% cases) was associated with increased death in the hospital (34.5% with hypovolemia/shock and 8.5% without hypovolemia/shock, p<0.001) and increased HD (12.8% with hypovolemia/shock and 11.9% without hypovolemia/shock, p<0.001). Cachexia (3.3% cases) was associated with increased death in the hospital (16.8% with cachexia and 9.3% without cachexia, p<0.001) and increased HD (21.9% with cachexia and 11.6% without cachexia, p<0.001). Other complications can be reviewed in Table 4.
A total of 48,406 cases (8.4%) were hospitalized for surgical intervention ( Table 5). Surgical intervention was associated with decreased death in the hospital (4.1% with surgery and 10.1% without surgery, p<0.001) and decreased HD (0.5% with surgery and 13.0% without surgery, p<0.001); 7.3% of the total cases received an endoscopic biliary stent placement. Biliary stent placement was associated with decreased death in the hospital (3.5% with the stent and 10.1% without the stent, p<0.001) and decreased HD (9.4% with the stent and 12.1% without the stent, p<0.001). Chemotherapy was reported in 3.3% of cases and was associated with decreased death in the hospital (5.6% with chemotherapy and 9.7% without chemotherapy, p<0.001) and decreased HD (6.1% with chemotherapy and 12.1% without chemotherapy, p<0.001).
Palliative care evaluation (PCE) was reported in 7% of the cases. PCE was associated with increased death in the hospital (36% with PCE and 7.6% without PCE, p<0.001), and increased HD (31.8% with PCE and 10.4% without PCE, p<0.001). The remainder of the numbers for other therapeutic interventions can be seen in Table 5.
The number of PC hospitalizations has been increasing with each passing year (

Discussion
The number of PC hospitalizations has been on the rise with each passing year. However, deaths in hospitals have decreased over the same period. HD gradually increased during the same period. The utilization of palliative care as a resource increased dramatically during the same period. However, during the same time, we did not notice a decrease in the percentage of patients who received aggressive care. Being a female was associated with fewer in-hospital deaths and increased HD.
These findings are consistent with the increasing incidence of PC in general [2]. Previous studies have noted the increased use of hospice over the years. However, it was also noted that an increase in hospice usage did not necessarily offset aggressive care near EOL [24]. Decreased odds of inpatient deaths and aggressive EOL care in females have been noted in a previous study [25].
The proportion of patients with metastatic PC dropped from 2005 to 2011. This is an indirect indicator that the hospitalizations of patients with incurable diseases have decreased over time. Potential explanations for this decrease are multifactorial. Patients with PC were either diagnosed in earlier stages, making them suitable surgical candidates, or an increased number of patients receiving palliative care and hospice utilization in the later years prevented re-admissions of patients with incurable metastatic disease. Given the challenges of diagnosing pancreatic cancer early, the latter explanation is more likely.
The incidence of DVT and PE increased slightly from 2005 to 2011. Oral anticoagulation during hospitalization was associated with decreased mortality and an increase in HD. Despite receiving IVCF, the HD percentage was much higher than the rest of the PC patients. This may suggest that patients who received IVCF insertion had advanced disease with other comorbidities, excluding them from receiving anticoagulation. Given the shortened anticipated survival in patients with advanced cancers, IVCF does not improve survival and may even negatively affect the quality of life [26].
The proportion of cases that received mechanical ventilation (MV) increased slightly from 2005 to 2011. This is an indirect metric to measure aggressive ICU level of care. More than half of these patients died during hospitalization. Not surprisingly, cases that received MV had a decreased HD rate compared to other PC patients (7.8% vs. 12.1%, p<0.001). This could indicate patient/family preference to pursue aggressive care. Previous studies have shown that patients disenrolling from hospice have higher healthcare usage and expenditures with increased ED visits, hospitalizations, and ICU admissions. The mortality rate is exceedingly high (57%) in such patients [27].
Although the data we reported represent an improvement in the utilization of palliative care, HD, and inpatient death numbers, the number of patients receiving aggressive interventions such as MV still remains high. Moreover, these interventions are associated with worse outcomes. Ideally, patients with terminal PC should have PCE earlier in the disease course and establish a reasonable and realistic plan of care. This would enable patients to have a better understanding of their prognosis and help them pursue an appropriate plan of care. This will also likely help decrease the number of hospitalizations and aggressive interventions near EOL. Finally, it should also be mentioned that despite the best efforts from healthcare providers, ultimately the direction of care and the level of aggressiveness in care is determined by patients' personal wishes. This aspect of patient preference can be hard to quantify and account for when analyzing the quality-of-care metrics in cancer patients.

Strengths of the study
The NIS database consists of cases from 48 states and includes patient populations from diverse socioeconomic backgrounds and demographics, making the results obtained more generalizable to the whole population of the United States. It has a large sample size, thereby making it possible to detect statistically significant differences among different variables. Since our study includes data from multiple years, it allowed us to observe trends and changes across time.

Limitations of the study
NIS is an administrative database, and thus clinical information such as histopathology, cancer stage, details of lab values, surgical reports, and imaging studies was not available. Longitudinal follow-up could not be done for the patients after discharge from the hospital. Information about outpatient treatment and care was not available. Patient and family preferences about the level of aggressiveness of care could not be ascertained from this database. We could not determine if the patients disenrolled from hospice before admission. The data used for this study is from more than a decade ago. This is because the method of reporting discharges changed in 2011, making it difficult to distinguish HD (home or facility) from routine discharges and discharge to a nursing home in the data from the later years. Despite utilizing older data, this analysis provides valuable insights into the trajectory of outcomes for PC patients through the years.

Conclusions
Over the years, hospitalizations for PC have increased, but inpatient deaths have decreased, with increased utilization of hospice as palliative care consultation has markedly increased. However, a corresponding decrease in indicators of aggressive care still seems to be lacking. With an increase in median survival times of PC patients, an increase in the incidence of PC-related complications is expected. Pursuing aggressive interventions for these complications is not always associated with better outcomes and, on the contrary, may be associated with worse outcomes in some cases. Thus, early evaluation of PC patients by palliative care is desirable, preferably in the outpatient setting. This would enable discussions about prognosis and goals of care in a less stressful environment, and help the patients come up with a plan of care that aligns with their values and is also medically appropriate. This would help keep PC patients out of the hospital and prevent aggressive interventions near EOL. However, patient preference is a big factor that ultimately determines the plan of care. Until better-targeted treatments and screening options become available for PC, mortality will remain high. The trends that we found in our analysis are encouraging, and surely signify a step in the right direction.

Additional Information Disclosures
Human subjects: Consent was obtained or waived by all participants in this study. Thomas Jefferson University Hospital IRB issued approval N/A. The study consists of de-identified patient information, which is nationally available to researchers for performing retrospective outcomes research. According to the IRB of our institution, "This study does not constitute human subjects research. You may proceed without IRB approval or exemption". Animal subjects: All authors have confirmed that this study did not involve animal subjects or tissue. Conflicts of interest: In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work.