Trends in hospice referral timing and location among individuals dying of ovarian cancer: persistence of missed opportunities

Abstract

Objective.

To evaluate trends, racial disparities, and opportunities to improve the timing and location of hospice referral for women dying of ovarian cancer.

Methods.

This retrospective claims analysis included 4258 Medicare beneficiaries over age 66 diagnosed with ovarian cancer who survived at least six months after diagnosis, died between 2007 and 2016, and enrolled in hospice. We examined trends in timing and clinical location (outpatient, inpatient hospital, nursing/long-term care, other) of hospice referrals and associations with patient race and ethnicity using multivariable multinomial logistic regression.

Results.

In this sample, 56% of hospice enrollees were referred to hospice within a month of death, and referral timing did not vary by patient race. Referrals were most commonly inpatient (41% inpatient, 17% outpatient, 7% nursing/longterm care, 36% other), with a median of 6 inpatient days prior to hospice enrollment. Only 17% of hospice referrals were made in an outpatient clinic, but participants had a median of 1.7 outpatient visits per month in the six months prior to hospice referral. Referral location varied by patient race, with non-Hispanic Black people experiencing the most inpatient referrals (60%). Hospice referral timing and location trends did not change between 2007–2016. Compared to individuals referred to hospice in an outpatient setting, individuals referred from an inpatient hospital setting had more than six times the odds of a referral in the last 3 days of life (Odds Ratio 6.5, 95%CI 4.4–9.8) versus a referral more than 90 days before death.

Conclusion.

Timeliness of hospice referral is not improving over time despite opportunities for earlier referral across multiple clinical settings. Future work delineating how to capitalize upon these opportunities is essential for improving the timeliness of hospice care.

Introduction

Among people with cancer, hospice use is associated with lower pain, symptom burden, and distress; improved quality of life; less aggressive care; higher likelihood of dying where a patient wishes; and improved bereavement.^1–5 For these reasons, hospice is the standard of care for anyone with a prognosis less than six months, and hospice enrollment has significantly increased over time.^6–8 However, hospice enrollment alone is not sufficient for maximizing benefits. Although quality metrics consider hospice referral ≤3 days before death too short, quality of life increases with longer hospice lengths of stay and even small increases in length of stay improved family perceptions of patient comfort, needs being met, and preparedness for death.^2,5,8–10 Earlier referral also increases opportunities to utilize hospice services and avoid intensive hospital-based care with high monetary costs and poor outcomes.^11–13

Across cancer types, people are referred to hospice too late throughout the United States.¹⁴ The often late-stage diagnosis and recurrent nature of ovarian cancer make hospice referral timing and location important facets of high-quality ovarian cancer care. Prior ovarian cancer studies demonstrate increasing hospice enrollment over time, and a persistent trend in entering hospice ≤3 days before death.^6,15 Among people with ovarian cancer, there are known racial disparities in hospice enrollment, but mixed associations between race and hospice referral timing.^6,16 Patterns of hospice referral locations for ovarian cancer patients have not been evaluated since 2007, and to our knowledge, these patterns have never been evaluated by patient race.¹⁵ Changes in utilization patterns over time may create new opportunities for improving timely hospice referral and minimizing racial disparities.

Using the population-based Surveillance, Epidemiology, and End Results-Medicare data set, the purpose of this study was to examine trends in the timing and clinical location of hospice referral for people with ovarian cancer who died between 2007 and 2016, a time of increasing focus on improved palliation and reduced end of life care intensity. We also evaluated racial disparities to understand how referral timing and location may differ by patient race and ethnicity.

METHODS

Data

Because of its comprehensive coverage of care prior to and during hospice referral, and its population-based gold-standard registry data, this analysis used data from the linkage of Surveillance, Epidemiology, and End Results registry data and Medicare claims data from years 2000–2016.¹⁷ The Wayne State University Institutional Review Board determined the study was exempt (IRB # 046917M1X).

Cohort Selection

To build upon prior work, women diagnosed with first and only primary ovarian cancer between 2000 and 2015 who died between 2007 and 2016 were eligible for inclusion.¹⁵ To leverage the rigor of Surveillance, Epidemiology, and End Results cancer diagnosis data, we selected our analytic cohort by ovarian cancer diagnosis restricted to first and only ovarian cancer (with the exception of basal and squamous cell skin carcinomas) because cause of death data is problematic in identifying the underlying cancer cause of death. To measure comorbidity, the sample was restricted to women over age 66 at the time of cancer diagnosis, who had complete case information captured by a Surveillance, Epidemiology, and End Results registry (n=16,661). To ensure complete cancer treatment information, we excluded patients who were not enrolled in both Part A and Part B Medicare, or who were enrolled in a health maintenance organization plan in the 12 months prior to diagnosis, or at any point from the time of diagnosis through the end of study observation (n=5,089).¹⁸ We excluded cases with: unknown month of diagnosis or death, death within 30 days of diagnosis, alive before December 2016, discrepancies between Surveillance, Epidemiology, and End Results and Medicare birth or death dates, diagnosis at death or autopsy, non-invasive disease, no claims after diagnosis, and hospice admission date predating diagnosis (n=3,816). To ensure sufficient utilization data prior to hospice enrollment, we further restricted the sample to women who survived at least six months after ovarian cancer diagnosis, and enrolled in hospice prior to death. In total, 4,258 women with ovarian cancer were included in our analytic sample.

Outcomes

Monthly utilization prior to hospice referral

We evaluated inpatient hospital, outpatient, and emergency department encounters prior to hospice referral to understand potential opportunities for hospice referral. We evaluated inpatient, outpatient and emergency department healthcare encounters in the six months prior to hospice referral. For women only surviving six months, longer duration of hospice enrollment prior to death would result in fewer opportunities for healthcare utilization, so the time denominator was six months - time enrolled in hospice.

Timing and location of hospice referral

We evaluated the timing of a patient’s first hospice referral after ovarian cancer diagnosis. We determined the hospice referral source using a previously developed claims-based timeframe for referral attribution.^15,19 Briefly, hospice enrollment within 2 days of an encounter was attributable to a referral source. Patients who had an emergency department encounter at an acute care hospital were included in inpatient hospital referrals due to small numbers. If a patient did not have an outpatient, inpatient or nursing/residential care encounter in the two days prior to hospice enrollment, they were included in the ―other‖ group. Referral timing was categorized by days before patient death (≤3, 4–30, 31–90, ≥91).

Exposures

Race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, other) and year of death were our exposures of interest. We calculated a modified Charlson Index score for each patient from the 12 months prior to diagnosis.²⁰ A priori-identified confounders included: urban residence, age of death, time between diagnosis and death (months), Charlson score (0,1,2,3+), census tract poverty (>20%), marital status (married/unmarried/unknown), Surveillance, Epidemiology, and End Results summary stage of cancer diagnosis(localized, regional, advanced), grade of cancer (high, low, unknown) and histology (serous/non-serous).²¹

Statistical Analysis

We tabulated patient characteristics by hospice referral time category and bivariate associations were evaluated with chi-squared tests. The association between age of death and hospice referral time category was evaluated with an analysis of variance test. The association of survival time and hospice referral category was evaluated with the Mann-Whitney U test due to the skewed distribution. The associations of patient race and types of utilization prior hospice referral/inpatient days in the hospice referral admission were evaluated with the Mann-Whitey U test due to skewed distributions. To examine trends over time, the average predicted probability of each referral time category (≤3, 4–30, 31–90, ≥91 days) was calculated using multinomial regression, making a prediction for each subject in each year with adjustment for patient characteristics and confounders as detailed previously.⁶ To examine the association between race/ethnicity and with timing and place of hospice referral, we assessed differences in the bivariate distributions using chi-squared tests. We used multivariable-adjusted multinomial regression to estimate the associations between timing of hospice referral, place of referral, and patient race, adjusting for the aforementioned confounders. All analyses were conducted in SAS version 9.4 and a two-tailed p-value of <0.05 was considered statistically significant.

Results

In this sample of 4,258 people with ovarian cancer, most individuals were Non-Hispanic white (87%), lived in urban areas (98%) and census tracks with <20% poverty (84%), had no comorbidities (60%), and were not currently married (53%) (Table 1). Most were diagnosed at advanced stage (78%), with a high or unknown grade tumor (90%), and serous histology (74%). The mean age of death was 80 years (standard deviation 6.9 years). Most women were referred to hospice from an inpatient hospital setting (41%), followed by outpatient (17%), nursing or long-term care (7%), and other (those without an encounter in the two-day attribution period) (36%).

Table 1:

Demographic and clinical characteristics of individuals in the Surveillance Epidemiology and End Results- Medicare database with invasive ovarian cancer who enrolled in hospice and died between 2007 and 2016 (n=4258)

	Patients N (%)
Patient Race and Ethnicity
Non-Hispanic White	3704 (87.0)
Non-Hispanic Black	225 (5.3)
Hispanic	207 (4.9)
Other races	122 (2.9)
Marital Status
Single	2241 (52.6)
Married	1873 (44.0)
Unknown	144 (3.4)
High Poverty
Yes	683 (16.0)
No	3575 (84.0)
Urban
Yes	4173 (98.0)
No	85 (2.0)
Charlson Score
0	2547 (59.8)
1	988 (23.2)
2	393 (9.2)
3+	330 (7.8)
Death Year
2007	466 (10.9)
2008	440 (10.3)
2009	392 (9.2)
2010	477 (11.2)
2011	437 (10.3)
2012	388 (9.1)
2013	425 (10.0)
2014	400 (9.4)
2015	403 (9.5)
2016	430 (10.1)
Stage at Diagnosis
Localized	175 (4.1)
Regional	489 (11.5)
Advanced	3318 (77.9)
Unknown	276 (6.5)
Grade
High	2076 (48.8)
Low	428 (10.1)
Unknown	1754 (41.2)
Histology
Serous	3153 (74.0)
Non-serous	1105 (26.0)
Hospice Referral Source
Nursing/ long-term care	299 (7.0)
Outpatient	703 (16.5)
Other	1525 (35.8)
Inpatient hospital	1731 (40.7)
Hospice Referral Time
≤3 days	556 (13.1)
4–30 days	1851 (43.5)
31–90 days	954 (22.4)
≥91 days	897 (21.1)
Continuous Measures
	Mean (Standard Deviation)
Age of death (years)	80 (6.9)
	Median (Range)
Time between diagnosis and death (months)	29 (6–198)

Timing of hospice referral prior to patient death varied significantly by location of referral (p<0.01) (Figure 1). Across locations, referrals were most common during the month prior to death (57% referred 0–30 days before death, n=2,407 referrals). Both inpatient hospital referral location and nursing/long-term care referral location had about 20% of patients referred in the ≤3 days prior to death, while only 5% of outpatient referrals were within 3 days of death (Figure 1). The distribution of hospice referral timing category did not vary significantly by patient race (p=0.2) or year of patient death (p=0.2) in bivariate analyses (results not shown). Non-Hispanic Black individuals had the largest proportion of inpatient referrals (60%) and nursing/long-term care referrals (9%) (p<0.01) (Figure 2).

Figure 1:

Timing of hospice referral before death by hospice referral location for individuals in Surveillance, Epidemiology, and End Results -Medicare database dying with ovarian cancer between 2007 and 2016 (p<0.01)

Figure 2:

Location of hospice referral for patients dying of ovarian cancer in the Surveillance, Epidemiology, and End Results -Medicare database between 2007 and 2016 by patient race (p <0.01)

Figure 3 shows potential opportunities for hospice referral prior to the hospice referral date (Figure 3, panel A) and during the referral admission for those who were referred from an inpatient hospital (Figure 3, panel B). Utilization opportunities in the six months prior to hospice referral differed significantly by patient race (p<0.01). Non-Hispanic Black individuals had the least outpatient visits and more inpatient days. Non-Hispanic Black and Hispanic individuals had more emergency department visits and inpatient admissions than Non-Hispanic White and other races individuals (Figure 3, panel A). The distribution of inpatient days during the inpatient admission where hospice referral was made was marginally different by patient race (four days for Hispanic patients, six for all others, p=0.05) (Figure 3).

Figure 3:

Opportunities for hospice referral by patient race among hospice enrollees dying of ovarian cancer in the Surveillance, Epidemiology, and End Results -Medicare database between 2007 and 2016 in the (A) Monthly utilization 6 months prior to referral for all individuals and (B) hospital inpatient days prior to hospice referral for patients referred inpatient.

Trends in hospice referral time and location were flat between 2007 and 2016 (Supplemental Figure 1). Compared to individuals referred to hospice in an outpatient setting, individuals referred from an inpatient hospital setting had more than six times the odds of a referral in the last 3 days of life (adjusted Odds Ratio (OR) 6.5 Confidence Interval (CI) 4.4–9.8) versus a referral more than 90 days before death. Compared to individuals referred to hospice in an outpatient setting, individuals referred from a nursing home or long-term care setting had five times the odds of a referral in the last 3 days of life (adjusted OR 5.0 CI 3.0–8.4) versus a referral more than 90 days before death (Figure 4). The adjusted odds of hospice referral timing did not vary significantly by patient race (results not shown).

Figure 4:

Adjusted odds ratios and 95% CIs of the association of hospice referral location and hospice timing before death for individuals in the Surveillance, Epidemiology, and End Results - Medicare database who died with ovarian cancer between 2007 and 2016

Model adjusted for: patient race, poverty, Charlson comorbidity score, marital status, urban residence, age of death, months from diagnosis to death, histology, tumor grade and stage at diagnosis.

Discussion

Summary of Main Results

In this population-based retrospective cohort of people with ovarian cancer who enrolled in hospice, hospice referral close to death was persistent over time across racial and ethnic groups, with 13% of patients undergoing this transition in the last 3 days of life. We also found persistent inpatient referrals, which were associated with higher likelihood of a referral near death. Taken together, these results suggest hospitalization is commonly being used as an impetus for rapid changes in goals of care across racial/ethnic groups, though utilization data suggest there are earlier referral opportunities, even within the hospitalization where the hospice referral is made.

Results in the Context of Published Literature

Considered with continuing trends in aggressive end-of-life care,⁶ these trends in late hospice referrals from inpatient settings likely reflect the use of hospice as a reactive add-on service rather than a proactive choice to prioritize patient quality of life. Wright et al. first noted this pattern from 1997 and 2007, where they found an increasing trend in inpatient hospice referrals.¹⁵ Since that time, palliative care developed as a specialty, hospice enrollment rates have increased, and numerous clinical guidelines have been issued to support early integration of palliative care and timely referral to hospice.^6,7,22,23 However despite this changing context, we found a similar 40% of hospice enrollees with inpatient hospice referrals from 2007–2016 and flat trends across other referral locations. To our knowledge, we are the first to evaluate these patterns nationally by patient race and ethnicity. While we did not find racial ethnic differences in timing of referral, we did see differences in hospice referral locations. Our study only included people who enrolled in hospice, but racial differences in referral location could contribute to the racial ethnic disparities in not enrolling in hospice and warrant further investigation.⁶

Our findings are in line with other studies suggesting goals of care conversations occur too late. Many providers are reluctant to talk about prognosis, which can prevent goals of care conversations, or make the conversation unclear.²⁴ Although prognostic understanding is associated with hospice enrollment, there is widespread prognostic discordance between patients and providers.^25,26 Rapid functional decline leading to hospitalization could spur hospice referral by aligning prognostic expectations, or perhaps prompting a goals of care conversation. A prior study found that 35% of patients dying from ovarian cancer had their first documented goals of care conversation during their final hospital admission in the last month of life.²⁷ Although these patterns of late inpatient referral are seen across diseases,²⁸ waiting for an inpatient admission is a reactive referral approach resulting in late referrals.

Strengths and Weaknesses

A unique strength of this study is identifying potential opportunities for earlier hospice referrals. This dataset also provides comprehensive claims data for end-of-life care delivered across clinical settings. Our study has potential limitations. First, our sample includes Fee-For-Service Medicare enrollees, which limits generalizability to older populations who do not have other Medicare plans. However, the median age of ovarian cancer death is 70, and our sample is population-based with oversampling to improve racial and ethnic diversity. This data provides a rich sample for a rare cancer, and comprehensive hospice enrollment information through Medicare claims. Second, we excluded women who did not survive at least six months after diagnosis. Among the women excluded, mean survival was 3 months. These patients would have different utilization patterns and timely hospice referral would look very different in an abbreviated course of care. Third, given the retrospective claims-based nature of our analysis, we did not have patient preferences, details of hospice conversations, decision-making time or care team composition. We used established methods from prior studies capturing hospice referral from claims.^15,19 It is unlikely that patients who had an encounter within 2 days of hospice enrollment did not have a referral discussion at this encounter. However, not having more specific clinical details may prevent us from attributing the hospice referral to an encounter, which likely contributes to referrals in the “other category.” Finally, reflective of national hospice enrollment patterns, our sample was predominantly non-Hispanic White individuals.²⁹ This highlights a critical need to better understand barriers to hospice referral and enrollment for racially minoritized groups.

Implications for Practice and Future Research

In the six months prior to hospice referral, patients had >1 outpatient encounter each month where a goals of care conversation might have occurred. One way to capitalize upon these visits is to routinely assess which upcoming clinic patients are at high risk of death and proactively have a goals of care conversation with them.³⁰ A single institution study of individuals with gynecologic cancer found that individuals who chose to discontinue chemotherapy proactively in an outpatient setting were less likely to die in the hospital, less likely to die within 30 days of a hospital admission, and more likely to accept hospice. These patients also had more time between choosing to discontinue chemotherapy and time of death.³¹ Even if an outpatient hospice referral is not made, patients who had a hospice discussion in an outpatient visit, may have shorter lengths of hospital stay.³²

Hospital admissions are a second area of opportunity for earlier hospice referral. Among those referred to hospice from an inpatient setting, the median length of inpatient stay was 6 days. For many in our study, shifting those 6 days from inpatient hospital to hospice would triple their hospice length of stay and reduce the likelihood of receiving aggressive end-of-life care.^5,6,12 Having the conversation upon presentation to the emergency department or upon admission could facilitate engaging the patient while they are communicative. A single institution study found that 29% of end of life conversations occurring in an inpatient setting took place with a patient’s family because the patient was unable to communicate with the physician.³³ Studies suggest that systematically identifying patients with serious illness in the emergency department and having a goals of care conversation or referring them to hospice is feasible and can improve hospice length of stay.^34–36

Conclusion

We found persistent trends in late inpatient hospice referrals that suggest hospice referrals are primarily completed in response to an admission resulting from rapidly declining functional status rather than a proactive choice to maximize quality of life. Although outpatient hospice referrals were associated with longer potential lengths of stay, we also found inpatient referral stays had a median of 6 days inpatient that could be targeted for timelier hospice referral. Considering around 10% of patients had a hospice referral within 3 days of death, these additional days could nearly triple their hospice length. Future work is needed to understand setting specific barriers and facilitators to hospice referral, and how these might influence hospice length of stay.

Supplementary Material

Supp1

Supplemental Figure 1: The adjusted yearly proportions of individuals referred to hospice in the Surveillance, Epidemiology, and End Results -Medicare database by (A) referral time prior to death and (B) hospice referral location.

Models are adjusted for patient race, urban residence, age of death, months between diagnosis and death, Charlson score, poverty, marital status, serous histology, tumor grade and stage at diagnosis.

What is already known on this topic –

More benefit is derived from earlier hospice referral, yet referrals are typically close to death. We do not know where there may be opportunities for earlier hospice referral for individuals with ovarian cancer, and how they may vary by race.

What this study adds –

There are a median of 1.7 outpatient visits per month in the last six months of life that could be used for goals of care conversations and earlier hospice referral. Even within the last hospitalization, a median of six inpatient days could be shifted to hospice and substantially increase hospice lengths of stay. Importantly, the location of these opportunities vary by patient race and may influence disparities in hospice enrollment.

How this study might affect research, practice or policy –

These findings highlight patterns of reactive hospice referral and identify opportunities for improving hospice length of stay.