Temporal trends and barriers for inpatient palliative care referral in metastatic gynecologic cancer patients receiving specific critical care therapies

Objective Existing evidence suggests that palliative care (PC) is highly underutilized in metastatic gynecologic cancer (mGCa). This study aims to explore temporal trends and predictors for inpatient PC referral in mGCa patients who received specific critical care therapies (CCT). Methods The National Inpatient Sample from 2003 to 2015 was used to identify mGCa patients receiving CCT. Basic characteristics were compared between patients with and without PC. Annual percentage change (APC) was estimated to reflect the temporal trend in the entire cohort and subgroups. Multivariable logistic regression was employed to explore potential predictors of inpatient PC referral. Results In total, 122,981 mGCa patients were identified, of whom 10,380 received CCT. Among these, 1,208 (11.64%) received inpatient PC. Overall, the rate of PC referral increased from 1.81% in 2003 to 26.30% in 2015 (APC: 29.08%). A higher increase in PC usage was found in white patients (APC: 30.81%), medium-sized hospitals (APC: 31.43%), the Midwest region (APC: 33.84%), and among patients with ovarian cancer (APC: 31.35%). Multivariable analysis suggested that medium bedsize, large bedsize, Midwest region, West region, uterine cancer and cervical cancer were related to increased PC use, while metastatic sites from lymph nodes and genital organs were related to lower PC referral. Conclusion Further studies are warranted to better illustrate the barriers for PC and finally improve the delivery of optimal end-of-life care for mGCa patients who receive inpatient CCT, especially for those diagnosed with ovarian cancer or admitted to small scale and Northeast hospitals.


Introduction
Gynecologic cancer is the most common malignancy in women, encompassing ovarian cancer, uterine cancer, and cervical cancer.According to the Cancer Statistics for 2022, it is estimated that there will be approximately 19,880 new cases of ovarian cancer, 65,950 new cases of uterine cancer, and 14,100 new cases of cervical cancer in the United States (US).Meanwhile, the estimated deaths for gynecologic cancer are also less than encouraging (1).Early diagnosis and treatment could improve cancer survival, while a significant number of cases progress rapidly and are diagnosed with metastasis (1).For those admitted to intensive care units, patients are frequently administrated with critical care therapies (CCT) to provide respiratory and nutritional support for life-saving measures (2)(3)(4).These patients are usually experience severe physical, psychological and social suffering (5,6).
Palliative care (PC) is a structured system that provides care to patients with end-stage diseases.It has been reported to improve symptom management, alleviate psychological suffering, and reduce cancer-related mortality (7).The American Society of Clinical Oncology (ASCO) and the Society of Gynecologic Oncology (SGO) have formally endorsed early palliative care for gynecologic cancer patients (8)(9)(10).Multiple studies have demonstrated the beneficial role of early PC in addressing symptoms and managing psychological concerns in patients with gynecologic oncology (11,12).However, studies have reported that PC is highly underutilized in metastatic gynecologic cancer (mGCa) patients, with utilization rates ranging from 5% to 24% (13)(14)(15)(16)(17). mGCa patients receiving CCT have increased cancer-related complications and long-term morbidity, and thus are strong indications for PC referral (18).Increasing awareness and accessibility of PC in this population is clinically significant.Although several publications have examined the utilization pattern of inpatient PC across different cancers in patients receiving life-sustaining treatments (19-21), there is a dearth of data focusing specifically on PC referral in mGCa patients receiving CCT while hospitalized.
The present study aims to investigate the temporal trends, predictors and barriers for inpatient PC referral in mGCa patients who specific CCT from a national perspective using the National Inpatient Sample (NIS) database.

Data source
Data in the study is de-identified and thus exempt from approval by an institutional review board.The NIS database is the largest publically available all-payer healthcare database in the US (22), developed by the Agency for Healthcare Research and Quality (AHRQ), as part of the Healthcare Cost and Utilization project (HCUP), which collected a stratified sample from nearly 1000 hospitals.Each hospitalization contains up to 30 inpatient diagnoses and 15 procedures that could be identified through the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes.

Patient and hospital characteristics
Patient-related, cancer-related and hospital-related characteristics were collected.Patient-related characteristics included age, year of admission, race, insurance type, income category, discharge destination, primary diagnosis and Elixhauser comorbidity score.The last consisted of 29 common comorbidities that could represent the disease burdens (excluded cancer in this study) (24).Cancer-related characteristics encompassed cancer type, metastatic sites, number of metastatic sites and chemotherapy.Lastly, hospital-related characteristics were hospital type, hospital bedsize and hospital region.

Definition of principal diagnosis and inpatient PC use
The principal diagnosis was categorized using the Clinical Classifications Software codes, which collapsed diagnoses and procedures into clinically meaningful categories (22).The primary outcome was temporal trend of inpatient PC referral in mGCa patients who received specific CCT.The secondary outcome included predictors of PC referral in the overall patients and in the subgroup undergoing IMV treatment.PC referral was defined using ICD-9-CM diagnostic code V66.7, which has been validated in metastatic disease with moderate sensitivity and high specificity (25,26).Cases involving patients under 18 years old or admitted to hospitals that did not provide PC service during the study period were excluded from the analysis.

Statistical analysis and covariates
Continuous characteristics between patients with and without PC referral were expressed as mean and compared using t-test, while categorical variables were reported as proportions and compared using chi-square tests.We calculated annual percentage change (APC) in the entire cohort and subgroups by race, hospital region, hospital bedsize, teaching status, cancer type and discharge destination.Sampling stratas, clusters and weights were considered to derive estimates from the national perspective using complex survey methods.Additionally, we preformed multivariable logistic regression analysis to explore the predictors of PC referral in mGCa patients receiving CCT, taking into account patient-related, cancerrelated and hospital-related characteristics.Confidence intervals for the ORs were calculated using the Taylor series method.
A P value ≤ 0.05 was considered statistically significant.All statistical analyses were performed using SAS version 9.4 and R version 3.6.2.

Discussion
Although ASCO and SGO have long recommended early integration of PC to improve end-of-life care, practical evidence shows high underutilization of PC referral in mGCa patients (8,10,13,16).Intensive care therapies are often provided to mGCa patients when severe treatment-related complications occurred or cancer progressed, highlighting the clinical importance and necessity of PC referral in this vulnerable population (2,18).Our analysis suggested that approximately 11.64% of patients received inpatient PC, and the rate of PC referral increased from 1.81% in 2003 to 26.30% in 2015, with an average annual increase of 29.08%.Multivariable analysis suggested that medium bedsize, large bedsize, Midwest region, West region, higher Elixhauser comorbidity score, uterine cancer and cervical cancer were related to increased PC use, while urban non-teaching hospitals, metastatic sites from lymph nodes and genital organs were related to lower PC referral.
Overall, approximately 11.64% of mGCa patients with CCT received inpatient PC, which is more than two times higher than the reported PC rate of 5% in the entire population regardless of CCT,   as reported by Rosenfeld et al. (13).However, this proportion is still far from satisfactory considering that all mGCa patients with CCT are candidates for PC referral.It is worth noting that PC referral consistently increased by 29.08% from 2003 to 2015.This phenomenon might reflect improved adherence of oncological guideline by both physicians and patients.Subgroup analysis indicated that increasing trend of PC referral was more pronounced in White and patients admitted to medium bedsize, urban non-teaching and Midwest hospitals, suggesting a wider acceptance of PC use in these patients.From the trend charts, it is evident that PC rate experienced a sharp increase since 2009, which aligns with the findings of previous publications (13,16).As aggressive measures such as CCT can reduce quality of life in mGCa patients, this unexpected increase may be partly attributed to the landmark ENABLE II trial in 2009 that revealed the effectiveness of PC interventions in improving the quality of life for patients with advanced cancer (27).When considering hospital region, patients hospitalized in Midwest hospitals had the highest PC rate (13.30%), followed by South (11.80%),West (11.46%) and Northeast (10.11%), accompanied by the highest APC (33.84%).Multivariable analysis accounting for potential confounders suggested that Midwest region (OR: 1.37) and West region (OR: 1.30) were associated with increased probability of PC referral compared to the Northeast region.This regional disparities in PC use has been previously reported.Milki et al. enrolled mGCa patients who subsequently died during hospitalization and found that patients in Midwest region (OR: 1.37) and West region (OR: 1.30) had increased PC use (16).Another study focusing on metastatic bladder cancer receiving CCT also described a higher PC rate in the West region (21).Further studies are warranted to understand the undelaying mechanisms for this geographic disparities and to relieve barriers for lower PC utilization in the Northeast region.
When considering hospital size, we observed that both medium bedsize (OR: 1.59) and large bedsize (OR: 1.59) were associated with increased PC use compared to small bedsize.One possible explanation for this finding might be that larger hospitals have more dedicated end-of-life specialists to provide PC services.However, research on this topic has produced conflicting results.For instance, Rosenfeld et al. conducted a study using data from the 2005 to 2011 NIS database, including all mGCa cases, and concluded that bedsize was not a predictor for PC referral (13).Another study by Milki et al. found that large bedsize was a positive predictor of PC referral (OR: 1.36) in mGCa cases who died in hospital (16).We hypothesized that the severity of dying status might result this disparity, as mGCa patients receiving CCT or died in hospital represented more severe conditions with significant symptom burden.Large bedsize hospitals are likely to form wellorganized PC team and well-established relationship between physicians and mGCa patients with more severe conditions.
There has been controversy surrounding the emerging evidence on racial disparities in PC use among mGCa patients (2,(13)(14)(15)(16).  (14).In our subgroup analysis focusing exclusively on metastatic ovarian cancer patients, we did not observe such racial disparities.These discrepancies may be attributed to different population groups and data sources, especially considering that our study specifically involved patients receiving CCT during hospitalization.Therefore, further studies are needed to provide sufficient evidence to better understand the underlying racial differences and to improve equitable provision of PC among mGCa patients, irrespective of race.For cancer types, uterine cancer ranked first in the rate of PC use (14.91%), followed by cervical cancer (14.81%) and ovarian cancer (10.24%).Although ovarian cancer patients has the lowest rate of PC use, the use of PC has dramatically increased over the study period, with the highest APC (31.35%).Previous studies have also reported lower PC use in ovarian cancer (13,14).As we know, ovarian cancer has a higher degree of malignancy and worse survival compared to uterine cancer and cervical cancer (30).Therefore, future efforts are needed to improve and optimize PC referral in metastatic ovarian patients receiving CCT.
The present study utilized a national-level hospitalized database covering long time spans to investigate the temporal trends and predictors for inpatient PC referral in mGCa patients who frequently received CCT, including IMV, TPN, PEG tube, tracheostomy and dialysis for AKF.However, several limitations should also be considered for an accurate interpretation of our results.Firstly, PC use in the NIS database was defined based on the ICD-9-CM diagnostic code V66.7.Being an administrative database, the NIS may not capture all instances of PC discussions, and only those that are documented by physicians are recorded.Therefore, there may be a bias towards underestimating the actual number of PC use cases.However, the code was initially introduced in 1996 and has since been used in several publications, demonstrating moderate sensitivity (66.3% to 83%) and high specificity (95% to 99.1%) (25,26).Secondly, this study focused only on specific CCTs that were frequently used in routine clinical practice.Any external extrapolation (eg, to all critically ill mGCa patients) should be interpreted with adequate caution.Thirdly, race information was unknown for nearly 12.22% of the included patients.Despite these limitations, the present study provides new evidence and insights into the understanding of PC referral in mGCa patients receiving CCT.
This analysis suggests that approximately 11.64% of patients received inpatient PC, which is still considerably below an ideal level.Further studies are necessary to elucidate the barriers to PC and ultimately enhance the provision of optimal end-of-life care for mGCa patients who receive inpatient CCT.This is particularly important for patients with ovarian cancer or those admitted to small-scale and northeast hospitals.

Conclusions
Despite the increase in PC referral over time, the absolute rate has remained low.The rates of PC referral in mGCa patients receiving CCT differ based on various sociodemographic and clinical factors.Thus, further studies are necessary to better understand the barriers to PC in mGCa patients undergoing inpatient CCT.

TABLE 1
Basic characteristics of mGCa patients receiving CCT stratified according to use of inpatient PC.

TABLE 1 Continued
*Small numbers of observations (<10) are at risk of identification of persons according to the HUCP and we replaced the number with an asterisk.FIGURE 2 Inpatient palliative care referral over time, stratified by hospital region, race, hospital bedsize and hospital teaching status.FIGURE 3 Inpatient palliative care referral over time, stratified by cancer type.

TABLE 2
Predictors of PC use in mGCa patients receiving CCT and IMV.

TABLE 2 Continued
(13,29)ial and cultural differences among various racial groups can help personalize palliative care for mGCa patients receiving CCT and improve the delivery of comprehensive cancer care.Studies have reported Studies have reported that racial minority groups, such as Black or Hispanic gynecologic cancer patients, have expressed a desire for more intensive and invasive end-of-life care (2, 28), making them the potential candidates for PC delivery from the perspective of end-of-life decision-making.Consistent with previous publications(13,29), our findings showed that Hispanic patients had the highest rate of PC use (16.83%), followed by Black patients (13.55%) and White patients (11.37%).However, this significant finding disappeared after adjustment for patient-related, cancer-related and hospital-related characteristics.Notably, Islam et al. analyzed data from the 2016 National Cancer Database and found that Hispanic and Black patients were less likely to utilize PC in metastatic ovarian cancer patients CCT, critical care therapies; mGCa, metastatic gynecologic cancer; PC, palliative care; IMV, invasive mechanic ventilation; PEG, percutaneous endoscopic gastrostomy; TPN, total parenteral nutrition; AKF, acute kidney failure; OR, odds ratio; CI, confidence interval.Understanding