Palliative care reduces symptom burden and improves quality of life for patients with serious illness, and early palliative care is associated with improved care outcomes for these patients. Although palliative care studies most often focus on patients with cancer, the value of palliative care in the management of patients with other serious illnesses is increasingly being recognized.
William H. Hung, MD, MPH
Three recent studies report important advances in our understanding of palliative care delivery. The first study, an analysis of differences in palliative care delivery patterns across different diagnoses, found that adults with noncancer diagnoses are less likely to receive palliative care early and less likely to receive high-quality palliative care. In the second study, the authors used a machine learning algorithm to predict mortality risk in patients with serious illness and provide behavioral prompts to clinicians to initiate serious illness conversations in those patients with the highest risk. Such interventions can ultimately lead to improved care delivery, including early palliative care initiation for patients with serious illness. The third study was conducted in the intensive care unit (ICU) setting and explored the use of time-limited trials — a shared decision-making strategy in which life-sustaining treatments are implemented for a predefined period to determine how the patient responds — as an approach to delivering care that respects patients' and caregivers' wishes while reducing nonbeneficial treatments in patients with advanced medical illnesses and poor prognoses. These three studies highlight gaps in palliative care delivery, build evidence for strategies that can lead to earlier care planning for patients with serious illness, and suggest a clinical approach for managing critically ill patients with a poor prognosis.
Palliative Care Delivery: Cancer vs Noncancer Diagnoses
A population-based cohort study conducted in Ontario, Canada, included 145,709 adults who died between January 2010 and December 2017 and had palliative care initiated within the last year of their lives. The study population was categorized by diagnosis: those with cancer, those with common illnesses with chronic organ failure (heart failure, chronic obstructive pulmonary disease, end-stage renal disease, cirrhosis, and stroke), and those with dementia. Use of palliative care services was identified through physicians' claims fee codes for delivery of care that was intended to be palliative rather than curative, including symptom management and counseling. Timing of initiation of palliative care, the primary outcome, was classified according to commonly used time frames: 30 days or less, 31-90 days, and more than 90 days. The study found that palliative care was initiated earlier among patients with cancer compared with those with organ failure or dementia: 28.9% vs 15.9% and 15.3%, respectively, had palliative care services initiated more than 90 days prior to death. Those who died of organ failure or dementia were less likely to have palliative care services initiated more than 90 days prior to death (odds ratios: 0.48 and 0.42, respectively) and between 31 and 90 days prior to death (odds ratios: 0.77 and 0.60, respectively), compared with those who died of cancer. In addition, patients with cancer more frequently received palliative care that was delivered by palliative care specialists, rather than generalists (72.9%), compared with those with organ failure or dementia (43.3% and 40.1%, respectively).
This study highlights variations in palliative care delivery across diagnoses, with patients with noncancer diagnoses more likely to have later initiation of palliative care services and less likely to have those services delivered by specialists. Both later initiation of palliative care and delivery of palliative care by generalists are associated with a lower likelihood of dying at home for patients with noncancer terminal diagnoses. To improve palliative care delivery, health systems should explore strategies that facilitate earlier engagement of palliative care specialists or that embed palliative care in specialty clinics for those with noncancer diagnoses.
Machine Learning Prognostication, Behavioral Nudges Promote Serious Illness Conversation
A key step in aligning end-of-life care with patients' preferences is the serious illness conversation, a structured conversation about prognosis, goals of treatment, and patients' care preferences. This study, a randomized, stepped-wedge cluster trial conducted over 20 weeks in nine oncology clinics at a large Pennsylvania healthcare system, examined an intervention that used machine learning mortality predictions followed by behavioral nudges to prompt oncology clinicians to carry out serious illness conversations. The intervention targeted oncology clinicians and included a weekly email with performance feedback (number of serious illness conversations clinician conducted in the previous 4 weeks) and peer benchmarking, along with a link to a dashboard where clinicians could review patients with mortality risk estimates (greater than 10%) who were scheduled for visits in the following week. Mortality risk was generated from a validated machine learning algorithm that used structured data from the electronic health record to predict risk for mortality within 180 days; predictions were updated weekly. Clinicians also received a reminder text message on the morning of a visit with a high-risk patient. A total of 78 clinicians participated in the study, and 4124 patients with high mortality risk had an oncology visit during the study period. The intervention was associated with a higher rate of serious illness conversations (15.2%) when compared with the control period (3.6%).
Machine learning prognostication combined with enhanced communication (behavioral nudges) with clinicians led to a fourfold increase in conversations about treatment goals and preferences among patients with cancer at increased risk for short-term mortality. These serious illness conversations can facilitate earlier initiation of palliative care and delivery of end-of-life care concordant with patients' wishes. This study's findings may be applied in other clinics, perhaps even in noncancer specialty clinics, where the rate of serious illness conversations is low. However, prognostication in other illnesses would require further research to develop the necessary algorithms, and education of clinicians in other specialties would present different challenges.
Time-Limited Trials for ICU Care Planning
Time-limited trials (TLTs) are agreements between the physician, patient, and decision-makers to use certain medical treatments for an agreed upon time period, with subsequent reassessment of the patient's progress according to previously established criteria for improvement or decline. Such agreements help with decision-making in patients with uncertain prognoses, can avoid the use of aggressive, nonbeneficial treatments, and help ensure that patients with poor prognoses receive care consistent with their goals and preferences. This study examines the effect of using TLTs as the default care planning approach for critically ill patients with advanced medical illness and poor prognosis. This study was conducted in three ICUs in California from 2017 to 2019. Patients with advanced medical illness and poor prognosis were identified and categorized using the Society of Critical Care Medicine guidelines for admission and discharge, and by ICU physicians classifying patients on the basis of their likelihood of benefiting from aggressive ICU treatment. Clinicians were trained to use TLTs through focus groups, didactic sessions, protocol reviews, and simulation sessions; a conversation guide was created to help clinicians during family meetings. A total of 113 patients at risk for nonbeneficial ICU treatment in the preintervention period were compared with 96 patients seen in the postintervention period. The rate of family meetings increased from 60% in the preintervention period to 96% in the postintervention period; risks and benefits of ICU treatment were addressed, values and preferences of patients were elicited, and clinical markers of improvement were identified more frequently in the postintervention period as well. Median ICU length of stay decreased from 8.7 days to 7.4 days after the intervention, as did the use of invasive ICU procedures (86% vs 73%), while hospital mortality was similar between both periods. Overall, interventions to incorporate TLTs for care planning for critically ill patients with advanced medical illness with poor prognosis were associated with higher-quality family meetings and a reduction in ICU length of stay.
This study showed that TLTs as a default care planning approach can improve the quality of family meetings for patients with advanced illness and poor prognoses, and may reduce aggressive, nonbeneficial ICU treatment. Limitations of the study are its pre-post design and the lack of a randomly assigned comparison group, which raise concerns about internal validity. Also, although the intervention's effect on family meetings was strong, it had little effect on reducing invasive ICU procedures and length of stay. Furthermore, the study did not explore other important outcomes, such as satisfaction with care when TLTs are used and downstream outcomes of grief among family and caregivers for patients who died. Nonetheless, this study highlights a feasible approach to delivering critical care among patients with poor prognosis for whom palliative approaches to care may be appropriate.
Medscape Family Medicine © 2021 WebMD, LLC
Any views expressed above are the author's own and do not necessarily reflect the views of WebMD or Medscape.
Cite this: Palliative Care Update: Delivery Gaps, Machine Learning, and Time-Limited Trials - Medscape - May 17, 2021.
Comments