Association of Early Palliative Care Use With Survival and Place of Death Among Patients With Advanced Lung Cancer Receiving Care in the Veterans Health Administration

Key Points

Question  Is early palliative care associated with a survival benefit among patients with advanced lung cancer?

Findings  In this cohort study of 23 154 patients with advanced lung cancer in the Veterans Affairs health care system, palliative care was associated with a survival benefit. The timing of the receipt of palliative care was important; palliative care received 31 to 365 days after diagnosis was associated with increased survival.

Meaning  The findings of this study suggest that palliative care should be considered a complementary approach to disease-modifying therapy in patients with advanced lung cancer.

Importance  Palliative care is a patient-centered approach associated with improvements in quality of life; however, results regarding its association with a survival benefit have been mixed, which may be a factor in its underuse.

Objective  To assess whether early palliative care is associated with a survival benefit among patients with advanced lung cancer.

Design, Setting, and Participants  This retrospective population-based cohort study was conducted among patients with lung cancer who were diagnosed with cancer between January 1, 2007, and December 31, 2013, with follow-up until January 23, 2017. Participants comprised 23 154 patients with advanced lung cancer (stage IIIB and stage IV) who received care in the Veterans Affairs health care system. Data were analyzed from February 15, 2019, to April 28, 2019.

Exposure  Palliative care defined as a specialist-delivered palliative care encounter received after lung cancer diagnosis.

Main Outcomes and Measures  The primary outcome was survival. The association between palliative care and place of death was also examined. Propensity score and time-varying covariate methods were used to calculate Cox proportional hazards and to perform regression modeling.

Results  Of the 23 154 patients enrolled in the study, 57% received palliative care. The mean (SD) age of participants was 68 (9.5) years, and 98% of participants were men. An examination of the timing of palliative care receipt relative to cancer diagnosis found that palliative care received 0 to 30 days after diagnosis was associated with decreases in survival (adjusted hazard ratio [aHR], 2.13; 95% CI, 1.97-2.30), palliative care received 31 to 365 days after diagnosis was associated with increases in survival (aHR, 0.47; 95% CI, 0.45-0.49), and palliative care received more than 365 days after diagnosis was associated with no difference in survival (aHR, 1.00; 95% CI, 0.94-1.07) compared with nonreceipt of palliative care. Receipt of palliative care was also associated with a reduced risk of death in an acute care setting (adjusted odds ratio, 0.57; 95% CI, 0.52-0.64) compared with nonreceipt of palliative care.

Conclusions and Relevance  The results suggest that palliative care was associated with a survival benefit among patients with advanced lung cancer. Palliative care should be considered a complementary approach to disease-modifying therapy in patients with advanced lung cancer.

Palliative care has been reported to reduce patients’ physical symptom burden and improve their mood and quality of life.¹ A meta-analysis has not established an association between palliative care and a survival benefit; however, the design and results of the studies included in the meta-analysis were mixed.¹ A single clinical trial prespecified survival as a primary outcome,² and only 2 others in the pooled analysis, 1 of which had a high risk of bias, reported a survival benefit.^3,4

In addition to the unclear association between palliative care and survival, a significant barrier to the timely receipt of palliative care is the mistaken conflation by patients and clinicians of palliative care with poor disease prognosis at the end of life (EOL). Palliative care is perceived as a modality that should only be recommended when life-prolonging therapies are no longer advised rather than an approach that should be integrated with standard oncology care. Despite its potential association with positive outcomes, palliative care is often underused^5,6 or delivered too late in the disease trajectory to provide meaningful benefit.^7,8

Of particular concern is the underuse of palliative care among patients with advanced lung cancer because these patients experience considerable disease-related suffering,⁹ including physical and psychological symptoms reported to diminish their quality of life.¹⁰ This study examined the real-world use of specialist-delivered palliative care in a national integrated health care system among patients with advanced lung cancer to better understand its associations with survival and place of death. More specifically, we explored the association between the timing of the receipt of palliative care relative to cancer diagnosis and survival. We tested the hypothesis that the early use of palliative care, received soon after diagnosis, is associated with increases in survival if the referral for palliative care does not occur as part of the dying process.

A retrospective cohort study was conducted among 23 154 patients with advanced lung cancer (stage IIIB and stage IV) who received care in the Veterans Affairs (VA) health care system and who were diagnosed with cancer between January 1, 2007, and December 31, 2013. The cohort was identified using the VA Central Cancer Registry, which receives and stores information on cancer diagnoses and treatments; data were compiled by the local cancer registry staff at VA medical centers, adhering to the standards of the Commission on Cancer.¹¹ An external audit reported that the case capture rate of the VA Central Cancer Registry was comparable to that of the National Cancer Institute’s Surveillance, Epidemiology, and End Results program.^12,13 Additional patient characteristics at the time of cancer diagnosis were compiled from the VA’s Corporate Data Warehouse, a clinical database that includes systemwide data from the VA health care system. Patients diagnosed with nonprimary lung cancer of unknown stage or histologic status who were enrolled in hospice before their cancer diagnosis or palliative care encounter were excluded (eMethods 1 in the Supplement). Further cohort details have been reported.¹⁴ This study was approved by the institutional review board of the VA Portland Health Care System with a waiver of informed consent.

Palliative care was defined as a specialist-delivered palliative care encounter received in an inpatient or outpatient setting after a lung cancer diagnosis. Palliative care and dates of service were identified through the VA’s Corporate Data Warehouse files by searching consultation notes and orders, VA-purchased care from external fee files, visit stop codes, bed section codes, Current Procedural Terminology codes, and the International Classification of Diseases, Ninth Revision, Clinical Modification code V66.7.^15-18 Palliative care was differentiated from hospice care using fee basis files and records of care purchased external to the VA, including a search of the Centers for Medicare & Medicaid Services hospice file data set using part A/B insurance claims.¹⁴

Primary outcomes included all-cause mortality and place of death (acute care vs nonacute care setting). All-cause mortality data were obtained from the master VA Vital Status File. Death dates listed in the VA Vital Status File have reported more than 98% sensitivity and more than 97% exact agreement with dates listed in the National Death Index.¹⁹ Place of death, if it occurred in a VA facility, was determined using the medical inpatient data set from the VA Vital Status File. Place of death was categorized as an acute care setting if it occurred during admission to a setting such as a hospital ward or an intensive care unit.

The primary independent variable was the receipt of palliative care. Descriptive statistics were used to characterize patient variables by palliative care. To address potential selection bias, we used propensity score (PS) methods^20,21 to construct a weighted cohort of patients who did not receive palliative care who were similar to patients who received palliative care on baseline characteristics that we could observe in our data using inverse probability of treatment weighting (IPTW).²² This approach minimized selection bias by comparing patients with a similar likelihood of receiving palliative care. The PS of receiving palliative care was estimated using a logistic regression model composed of baseline covariates at cancer diagnosis, including patient characteristics (eg, age, race/ethnicity, Charlson comorbidity index score,²³ and functional comorbidity index score²⁴), tumor characteristics (eg, stage and histologic status), and cancer reporting facility.¹¹ Covariate balance before and after PS weighting was assessed using model-adjusted standardized differences in means and the ratio of variance for each of the covariates (eMethods 2, eFigure, and eTable 1 in the Supplement). The all-cause mortality censoring date for final data extraction was January 23, 2017. Data were analyzed from February 15, 2019, to April 28, 2019.

A Cox proportional hazards model was used to regress time from cancer diagnosis to all-cause mortality based on the receipt of palliative care using IPTW.²⁵ Next, we considered palliative care as a dichotomous time-varying covariate (TVC)²⁶ because this approach more accurately attributed days of survival, without the resultant immortal time bias, compared with a Cox model that would have treated palliative care as a fixed covariate. We categorized the receipt of palliative care a priori from the date of cancer diagnosis to the first palliative care encounter into 8 levels based on frequencies and regular time intervals, from 0 days (cancer diagnosis) to more than 365 days. A time-varying Cox proportional hazards model was used to regress time from cancer diagnosis to all-cause mortality on the TVC version of the receipt of palliative care using IPTW. Model parsimony was also created that collapsed the timing of palliative care into 3 levels (0-30 days, 31-365 days, and >365 days) based on the direction and magnitude of hazard ratios in the Cox model. The final survival analysis was performed through IPTW using the PS. The PS weights were recalculated based on the 3-level TVC to ensure the covariates were balanced, and covariate balance was assessed (eTable 2, eTable 3, and eTable 4 in the Supplement).

Palliative care and place of death (acute care vs nonacute care setting) were analyzed through multivariable logistic regression and IPTW using the PS. Survival time was calculated from the date of cancer diagnosis to the date of death and stratified using the Cox proportional hazards regression model and the 3-level TVC. To illustrate differences in the 3-level TVC between survival over time after cancer diagnosis, we plotted survival estimates from the time-dependent Cox proportional hazards regression model, assuming mean values for model covariates (eg, lung cancer treatments received). Plotted survival estimates excluded time before the receipt of palliative care. For example, among the group that received palliative care more than 365 days after cancer diagnosis, only survival time after the receipt of palliative care or more than 365 days after diagnosis was plotted. In all regression models described herein, lung cancer treatments received (eg, chemotherapy) were added as covariates in the regression model because they were hypothesized to be associated with mortality.

Several sensitivity analyses were completed using (1) PS weights as a covariate in the model instead of weighting the sample, (2) reporting facility as a covariate in both the PS and regression models, (3) reporting facilities with high or low use of palliative care (high users were defined as facilities that used palliative care more often than the total facility mean), (4) hospice enrollment and the setting of the initial palliative care encounter (inpatient or outpatient) as covariates in the regression model, (5) a 4-level TVC (0-14 days, 15-90 days, 91-365 days, and >365 days), and (6) exclusion of patients who died within 14 or 30 days of diagnosis. Data were analyzed using Stata software, version 15 (StataCorp LLC). All modeling used robust SEs (ie, estimators for SEs were efficient for a wide range of error distributions), and statistical testing was 2-sided with a significance threshold of P < .05.

In total, 23 154 patients with advanced lung cancer (stage IIIB or stage IV) were included in the study. The mean (SD) age of participants was 68 (9.5) years and 98% of participants were men (Table 1). Analysis of tumor characteristics found that 89% of tumors were stage IV, and 69% had non–small cell lung cancer histologic status. Participants had received treatment with chemotherapy or radiotherapy for a median of 37 days (interquartile range [IQR], 19-63 days) and 33 days (IQR, 14-62 days) after diagnosis, respectively. Overall, 13 109 patients (57%) received palliative care, and a 41% relative increase was noted in the use of palliative care from January 1, 2007, to December 31, 2013. All sensitivity analyses yielded similar results (eTable 5 in the Supplement).

Considerable facility and regional variations in palliative care use existed, ranging from less than 5% to more than 95% (Figure 1). Differences in the timing and location of palliative care receipt were also noted. Of the initial palliative care encounters received within 30 days of cancer diagnosis, 76% occurred in the inpatient setting. Of those received 31 to 365 days after diagnosis or more than 365 days after diagnosis, 55% and 50% of the encounters occurred in the inpatient setting, respectively. Median survival after cancer diagnosis was 136 days (IQR, 52-320 days), which was similar by year of diagnosis.

Overall, palliative care was associated with decreases in survival (adjusted hazard ratio [aHR], 1.19; 95% CI, 1.15-1.23; Table 2). However, examination of the timing of palliative care receipt relative to lung cancer diagnosis found that palliative care received 0 to 30 days after diagnosis was associated with decreases in survival (aHR, 2.13; 95% CI, 1.97-2.30), and palliative care received 31 to 365 days after diagnosis was associated with increases in survival (aHR, 0.47; 95% CI, 0.45-0.49; Table 3; Figure 2). The effect size among those who received palliative care 31 to 365 days after diagnosis was larger when facilities that were high users of palliative care were solely included (aHR, 0.42; 95% CI, 0.40-0.45). Receipt of palliative care more than 365 days after diagnosis was associated with no difference in survival compared with nonreceipt of palliative care (aHR, 1.00; 95% CI, 0.94-1.07). Secondary analyses examined survival stratified by the setting of the initial palliative care encounter (inpatient vs outpatient; eTable 6 and eTable 7 in the Supplement).

Among the entire cohort, 7710 patients died in VA facilities, with 3455 (45%) of those deaths in acute care settings and 4255 (55%) in nonacute care settings. Receipt of palliative care was associated with a reduced risk of dying in an acute care setting (adjusted odds ratio, 0.57; 95% CI, 0.52-0.64) compared with nonreceipt of palliative care (Table 2).

This study found palliative care to be associated with a survival benefit among patients with advanced lung cancer receiving care in a national health care system. The timing of the receipt of palliative care relative to cancer diagnosis was a component of its association with a survival benefit. Palliative care received within 30 days of diagnosis was associated with decreases in survival, and palliative care received 31 to 365 days after diagnosis was associated with increases in survival. Palliative care was also associated with reductions in the risk of dying in an acute care setting.

In real-world settings, distinct populations of patients with lung cancer likely exist. Patients who receive palliative care soon after diagnosis (ie, 0-30 days) are more likely to be seriously ill; most of the initial palliative care encounters in this study occurred in the inpatient setting, and this supportive approach was intended to ease the dying process rather than increase survival. Patients who were less seriously ill at diagnosis and who received early palliative care were more likely to benefit from its multidisciplinary patient-centered approach, which may have been associated with their increased survival. The results of the study were consistent with our initial hypothesis that early use of palliative care is associated with increases in survival if the referral for palliative care does not occur as part of the dying process. Our results may help to dispel the belief that early palliative care among patients with lung cancer is associated with decreases in survival. Palliative care should be viewed as a complementary rather than a competing approach to disease-modifying therapy for patients with advanced cancer.

The mechanisms through which palliative care may be associated with increases in survival are largely unknown. The health decision model, which uses health beliefs and behavioral decision theory to explain patients’ health decisions and outcomes, has been suggested as a factor in the association between palliative care and survival benefit.²⁷ One example of how this model works can be seen in a patient’s willingness to seek aggressive EOL care, which may be owing to their preexisting perceptions rather than factual data. Palliative care is reported to be associated with improvements in symptom burden, mood, and quality of life^1,4; deficits in these factors are associated with decreases in survival among patients with lung cancer.^28,29 In addition, by enhancing patients’ understanding of their illness and promoting shared decision-making, early palliative care may be associated with the decreased use of overly aggressive EOL care, which may cause more harm than benefit for some patients.³⁰ In a landmark clinical trial that suggested a survival benefit of outpatient palliative care, patients in the intervention group were reported to have reduced receipt of aggressive chemotherapy at the EOL.⁴ More research is needed to explore the association of palliative care and aggressive EOL therapies with survival.

Our results reinforce the importance of the timing of palliative care receipt because there may be associations with a survival benefit. Providing palliative care earlier in the disease trajectory is not a novel concept.³¹ Early integration has been suggested to ensure that the benefits and burdens of treatment are regularly reassessed and the decision-making process regarding the care plan is responsive to changes in the patient’s condition, allowing a gradual transition in goals.³¹ Compared with later palliative care, outpatient palliative care received within 30 to 60 days after an advanced cancer diagnosis was associated with improvement in 1-year survival among patients with advanced cancer.² In addition, it is unreasonable to expect a significant association with outcomes such as survival when a referral occurs in the last few weeks or even days of life. In this scenario, palliative care helps to prepare patients and their families for the dying process by offering bereavement support and an exploration of hospice options, which is consistent with the fundamental elements of palliative care.³¹ This function of palliative care may explain our results among patients who received palliative care soon after diagnosis (ie, 0-30 days). In addition, approximately 5% of our cohort received palliative care up to 1 year after diagnosis, which was not associated with a survival benefit. These patients represented an atypical disease trajectory; therefore, their palliative care needs and survival were likely not representative of patients with advanced lung cancer.

Families reported that they would have preferred earlier palliative care referral.^32,33 The importance of timing in relation to other aspects of EOL care, such as hospice enrollment, has been endorsed as a quality measure for cancer care.^34,35 In a clinical trial that reported a survival benefit of palliative care among patients with advanced lung cancer, patients were enrolled in palliative care within 8 weeks of diagnosis and met with a member of the palliative care team within 3 weeks of enrollment.⁴ The association of a survival benefit with an early 8- to 11-week enrollment window is consistent with our results. Our findings are also consistent with the current American Society of Clinical Oncology guidelines, which recommend that palliative care be received within 8 weeks of diagnosis.³⁶ Unfortunately, critical barriers to early palliative care use remain, including a general lack of knowledge among both clinicians and patients about what palliative care is^37,38 and a perceived equivalence to hospice care.^8,39 Oncologists have also expressed concern that referral to palliative care may destroy their patients’ sense of hope,⁴⁰ potentially decreasing survival. Our results suggest the importance of considering the complementary role of palliative care delivered concurrently with disease-modifying therapies as a key component of comprehensive cancer care.

As the largest integrated health care system in the United States, the VA has developed several initiatives designed to improve the availability and quality of EOL care. In 2002, the VA created a Hospice and Palliative Care Program and mandated that all VA medical centers have an interdisciplinary palliative care consultation team composed of members with specialties that are similar across facilities. Additional VA initiatives have attempted to increase patients’ access to palliative care and improve collaboration with community hospice programs.⁴¹ These systemwide approaches have been reported to increase the use of quality EOL care; however, variability exists.¹⁴ Among Medicare beneficiaries, heterogeneity in EOL care was significantly associated with local factors, such as community leadership.⁴² Other potential factors in variability in palliative care use include the availability of local resources, such as interdisciplinary palliative care staffing. These considerations are important because we found greater survival benefits to be associated with facilities that were high users of palliative care, and the local resource requirements necessary to achieve maximal patient benefit deserve more attention.

Among those who are terminally ill, congruence between preferred and actual place of death is an important component of high-quality EOL care.⁴³ Patients overwhelmingly prefer home palliation⁴⁴ and do not want to die in a hospital.⁴⁵ However, US hospitals remain a major site of EOL care, with 29% of deaths occurring in the hospital.⁴⁶ Among Medicare beneficiaries with cancer, use of the intensive care unit and hospitalizations in the last months of life are increasing, with approximately 1 in 5 patients spending their last days in a hospital.⁴⁷ The most strongly associated factors among patients who died at home were low functional status, an expressed preference for home death, home care, living with relatives, and being able to count on extended family support.⁴⁸ Although we adjusted for functional and marital status in our analyses, patient preferences and other measures of social support were unavailable. Palliative care strategies promote the need to involve patients who are approaching EOL and their families in discussions about the preferred place of death to discern their preferences. Study results, with varying degrees of bias, have been mixed regarding the association of palliative care with at-home death.¹ Our results suggest that within an integrated health care system, palliative care may be associated with improving the quality of EOL care by delivering care that is consistent with most patients’ wishes.

This study has limitations. Our design was a retrospective cohort study that was subject to unmeasured residual confounding. For instance, oncologists who provide earlier referrals for palliative care may also be more likely to focus on quality EOL care, thereby increasing survival in unpredictable ways. We relied on a documented history of a specialty palliative care encounter as evidence of palliative care receipt. Patients may have received primary palliative care from clinicians who were not members of the palliative care consultative teams (eg, primary care clinicians). A standardized definition of a palliative care encounter does not exist; therefore, our exposure was considered a palliative care approach. In addition, patient-level data for patients’ preferences for palliative care were not available; thus, data for patients who were offered but refused palliative care were not captured. Our cohort was limited to patients who received care in the VA health care system, and only 2% were women, which affected the study’s generalizability. Few patients in our cohort were candidates for targeted therapies or immunotherapies that may have increased their survival, which also limited generalizability. Depression likely factors into the association between quality of life and survival among patients with cancer.^29,49,50 Patient depression at baseline was included in the functional comorbidity index score; however, longitudinal assessments were not included. The place of death was limited to VA facilities, and deaths in other settings were not available for measurement.

The results of this study suggest that palliative care was associated with increases in survival among patients with advanced lung cancer when it was received 31 to 365 days after cancer diagnosis. Palliative care was also associated with decreases in the risk of death in an acute care setting. Palliative care should be considered an integral part of comprehensive lung cancer care, with potentially meaningful patient benefits. The results of this study appear to support the timely use of palliative care among patients with advanced lung cancer.

Accepted for Publication: June 5, 2019.

Corresponding Author: Donald R. Sullivan, MD, MA, MCR, Health Services Research and Development, VA Portland Health Care System, 3710 SW US Veterans Hospital Rd. (R&D 66), Portland, OR 97239 (sullivad@ohsu.edu).

Published Online: September 19, 2019. doi:10.1001/jamaoncol.2019.3105

Author Contributions: Drs Sullivan and Chan had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Sullivan, Ganzini, Hansen, Fromme, Golden, Slatore.

Acquisition, analysis, or interpretation of data: Sullivan, Chan, Lapidus, Ganzini, Carney, Fromme, Marino, Golden, Vranas, Slatore.

Drafting of the manuscript: Sullivan, Marino, Slatore.

Critical revision of the manuscript for important intellectual content: Sullivan, Chan, Lapidus, Ganzini, Hansen, Carney, Fromme, Golden, Vranas, Slatore.

Statistical analysis: Sullivan, Chan, Lapidus, Marino.

Obtained funding: Sullivan, Ganzini.

Administrative, technical, or material support: Sullivan, Golden, Slatore.

Supervision: Sullivan, Lapidus, Ganzini, Carney.

Conflict of Interest Disclosures: Dr Hansen reported receiving grants from the National Institute of Nursing Research of the National Institutes of Health and airfare and honorarium for an in-person presentation at the 2018 Association of VA Hematology/Oncology Annual Meeting outside the submitted work. Dr Fromme reported receiving grants from the Gordon and Betty Moore Foundation, The John A. Hartford Foundation, and The Pew Charitable Trusts, and compensation for coauthorship and presentation of a paper entitled “Serious Illness Workforce Training” at a University of California, San Francisco, summit from the Gordon and Betty Moore Foundation outside the submitted work. No other disclosures were reported.

Funding/Support: This research was supported by grants K07CA190706 from the National Cancer Institute of the National Institutes of Health (Dr Sullivan); CIN 13-404 from the VA Health Services Research and Development Service, Center to Improve Veteran Involvement in Care, VA Portland Health Care System; and VA HSR RES 13-457 from the United States Department of Veterans Affairs, VA Informatics and Computing Infrastructure. Support for data from the United States Department of Veterans Affairs and the Centers for Medicare & Medicaid Services was provided by the United States Department of Veterans Affairs, VA Health Services Research and Development Service, VA Information Resource Center (Dr Sullivan).

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The views expressed in this article are those of the authors and do not necessarily represent the views of the United States Department of Veterans Affairs or the United States government.

Additional Contributions: Thomas Meath, MPH, senior data analyst at Oregon Health and Science University, made substantial contributions to the study through data extraction and preparation. Dalia Mobarek, MB, BCh, MSc, assistant professor at the Veterans Administration Medical Center and George Washington University, provided assistance with data collection and management. They received no compensation.