Editorial
January 16, 2024
A Tale of 2 Palliative Care Trials: Developing Sustainable and Transferable Models
Ashwin A. Kotwal, Lauren J. Hunt, Alexander K. Smith
JAMA. 2024;331(3):196-198. doi:10.1001/jama.2023.26815
At first glance, the 2 palliative care trials in this issue of JAMA may seem a study in contrasts.1,2 One was an outpatient intervention, the other inpatient. One included 306 participants, the other 24 065. One was an efficacy study, the other a pragmatic effectiveness study. One of the interventions added resources to expand access to palliative care, the other tried to expand access to palliative care with no additional resources. In one study the primary outcome was quality of life (QOL), in the other length of stay (LOS). In one study the primary outcome was positive, in the other it was negative.
And yet these studies, which may seem so disparate at the outset, are unified in addressing palliative care in noncancer serious illness. A 2020 systematic review and meta-analysis of palliative care for persons with noncancer illness published in JAMA found no significant difference in QOL compared with usual care and low representation of advanced lung disease and kidney disease in randomized clinical trials.3,4 The 2 trials in this issue of JAMAare an important advance in addressing palliative care in noncancer illness.
Bekelman et al conducted a single-blind, randomized clinical trial among 306 primarily White, male veterans with chronic obstructive pulmonary disease (COPD), heart failure, and/or interstitial lung disease at 2 Veterans Affairs sites. The intervention involved a nurse and a social worker, each making 6 phone calls focusing on symptoms, psychosocial needs, and health care navigation. The nurse and social worker were not specialty trained palliative care clinicians, although they each received 8 hours of training: the nurse in symptom management, communication, and motivational interviewing and the social worker in psychosocial interventions. The nurse and social worker participated in weekly collaborative team meetings with a primary care clinician, a palliative care specialist, and a cardiologist or pulmonologist as needed.
The intervention led to significant improvements in the Functional Assessment of Cancer Therapy-General (FACT-G) QOL (range, 0-108), which increased by 6 points compared with 1.4 points in the usual care group. This difference met the minimal clinical difference threshold of 4 points and achieved a standardized mean difference of 0.42, which is an effect size at or greater than other resource-intensive interventions.5
In contrast to prior palliative care trials, which often exclude persons with significant mental health needs,6 patients in the ADAPT trial had high baseline use of mental health specialists (62%) and counseling (72%). ADAPT led to clinically meaningful improvements in anxiety (standardized mean difference, −0.51 points), along with improvements in depression, FACT-G subscales (emotional and functional well-being), and disease-specific QOL.
In the second study, Courtright and colleagues report results of a step-wedge pragmatic randomized trial implementing “default orders” for palliative care consultations in 11 hospitals across 8 US states and 15 275 older inpatients with advanced dementia, COPD, or kidney disease. “Default orders” were automatic consults for palliative care placed in the electronic health record, allowing for clinicians to opt-out. The intervention increased consult rates from 16.6% to 43.9% and reduced the time to palliative care consultation by 1 day. Primary teams canceled only 9.6% of default orders, suggesting the orders were acceptable. The primary outcome—median hospital LOS—did not differ between the default order intervention (4.9 days) and usual care (5.0 days) groups. Notably, the default order was associated with an increased odds of a do-not-resuscitate order at discharge (adjusted odds ratio, 1.40 [95% CI, 1.2-1.6]) and being discharged to hospice (adjusted odds ratio, 1.30 [95% CI, 1.07-1.57]), and there was no difference in in-hospital mortality.
These studies raise several key questions for research, implementation, and dissemination of palliative care interventions. First, is the tested model for palliative care in noncancer serious illness ready for dissemination in the real world? The study by Courtright et al of default orders was a pragmatic effectiveness study that could in theory be easily implemented in many health systems. Yet, the intervention was decidedly not “real world.” Default orders were not paired with increased staffing or resources for palliative care teams. Perhaps as a consequence, nearly 46% of hospitalized patients who received default consults were not seen by palliative care teams. A key lesson from dissemination and implementation science is that attention to acceptability and sustainability, including adequate resources and staffing, is critical to testing the full potential of interventions.
Many palliative care clinicians will greet results demonstrating no difference in LOS with a degree of relief. Imagine had the study demonstrated reduced hospital LOS. Hospital administrators might then have approached existing palliative care teams, arguing that without additional support, they could more than triple the number of consults they were seeing for persons with noncancer serious illness and reduce hospital costs without additional resources. Instead, the study demonstrates that simply asking palliative care teams to see more patients, presumably doing less than with typical consultations, does not work. Like most things that are worthwhile, palliative care is not free and requires adequate resources.
The efficacy study by Bekelman and colleagues examined an intervention under ideal circumstances: a highly effective nurse-social worker team, an interdisciplinary collaborative model, and a Veterans Affairs system attuned to holistic needs of veterans with serious illness (including 60%-70% receiving mental health care). However, questions remain about reproducibility and generalizability. For example, the study primarily relied on 1 nurse for all calls at both sites. Might study findings be attributable to an exceptional nurse, or could most nurse-social worker teams attain the same outcomes? Moreover, only 12% of participants meeting inclusion criteria were randomized (of 2427 eligible patients, 306 were randomized, 1007 declined participation, 1283 were unable to be reached, and 37 consented but were not randomized), raising questions about generalizability.
A second consideration highlighted by each study is a common challenge in palliative care trials: there is no perfect primary outcome that will satisfy all interested parties. In the study by Courtright and colleagues, arguably none of the measured outcomes were patient-centered, including hospital LOS.7,8 In our experience, palliative care teams do not approach patient encounters with an agenda to expedite hospital discharge. Instead, their tasks are to address physical, psychosocial, spiritual, and ethical needs and align available treatment options with a patient’s goals of care. These core aspects of consultations were not assessed.
In contrast, the study by Bekelman and colleagues used a multidomain QOL outcome initially developed for patients with cancer. Even this patient-centered outcome had limitations; the FACT-G scale, for example, has no question assessing dyspnea, the most common symptom addressed in the intervention. Furthermore, although QOL is undeniably patient-centered, future uptake of the ADAPT trial intervention is unlikely without evidence of equivalent or reduced costs to the health system.
Third, these studies have implications for expanding the reach and sustainability of palliative care across the outpatient-inpatient continuum. Palliative care is now standard care in the inpatient setting, with 72% of hospitals in the US offering palliative care services (as of 2019).9 The idea of expanding access to inpatient palliative care through default consults may appeal to health systems by leveraging existing inpatient palliative care teams, which are an established, financially sustainable model.10,11 In contrast, in the outpatient and community settings, reimbursement rates are largely insufficient to cover the costs of providing care, leading to wide variability in access nationwide.12 Flexible, phone-based collaborative care models such as in the study by Bekelman et al offer a potential way forward. The Centers for Medicare & Medicaid Services demonstrated a new willingness to fund collaborative care models that support patients and, critically, caregivers (the Guiding an Improved Dementia Experience Model was announced July 2023).13 Rather than funding multiple disease-specific programs, findings from the study by Bekelman et al suggest that the Centers for Medicare & Medicaid Services could consider funding a collaborative geriatric palliative care model for all older adults with serious illness who are not eligible or do not wish to enroll in hospice and their caregivers, as proposed in Senate Bill S.1845, Expanding Access to Palliative Care Act.14
Taken together, the studies by Courtright and colleagues and Bekelman and colleagues, although different in approach and outcome, break new ground in developing sustainable and transferable models to address the needs of persons with noncancer serious illness. Testing models of palliative care delivery in noncancer serious illness remains critically important because palliative care is a limited resource and many clinicians continue to have a poor understanding of when it is appropriate to involve palliative care in noncancer serious illness. Although the findings from each study are promising, more work is needed to adapt these models of care to real-world settings to ensure high-value, cost-effective, and patient-centered palliative care.
