Core Elements of Goals of Care Initiatives Across Eight Health Systems

Abstract

For patients with serious illnesses, goals of care conversations improve quality of life and patient and family satisfaction and may reduce healthcare costs. However, these conversations often happen late in a serious illness or not at all. To better integrate goals of care into routine clinical practice, health systems across the country have implemented initiatives to increase and document these conversations. In this paper, we describe the landscape of goals of care initiatives across eight large health systems in the United States and identify core elements for effective programs: 1) Defining the purpose of the initiative; 2) Identifying the target patient population using patient diagnoses, artificial intelligence algorithms, or length of stay; 3) Engaging key stakeholders, including patient, caregiver, frontline provider, and leadership; 4) Encouraging the conversation through clinician and patient education and electronic health record prompts; 5) Documenting conversations within the electronic health record; 6) Measuring data by building electronic health record and information technology infrastructure; and 7) Planning for sustainability and scalability through leadership and funding support. These core elements can help inform how health systems plan goals of care initiatives, build infrastructure, and garner support to successfully implement these initiatives.

Introduction

Conversations focused on patients’ values, preferences, and goals in the setting of serious illness can facilitate the provision of goal-concordant care.¹ These care planning conversations are especially important for patients with complex, critical, and/or chronic illness and address current preferences and future decisions related to advance care planning.^2,3 Although various terms are used across health systems and literature to describe these conversations, we specifically refer to these as “goals of care” conversations in this paper.^4,5

Health systems across the United States have sought to promote, support, and document these conversations through focused goals of care initiatives.⁶ We define these initiatives as organized efforts to improve the frequency, timing, or quality of these conversations across a health system, beyond a single clinic or inpatient ward. These initiatives have taken many forms, including clinician education, patient identification and outreach, and provider activation.^7–11 Health systems have devised various initiatives to encourage conversations around goals of care. Large payer-based health systems, such as the Veterans Health Administration and Kaiser Permanente, have described a combination of education and changes to the electronic health record (EHR) to promote and document goals of care conversations.^12–14 While several initiatives have been previously described across institutions, one focused only on initiatives within oncology and another on the implementation of the Serious Illness Care Program (SICP).^15–17

In this paper, we summarize and give examples of key aspects and core elements of goals of care initiatives based on the experience of eight large health systems in the United States. These lessons learned can be used to promote and implement goals of care initiatives more broadly in health systems.

Methods

We used purposive sampling to identify health systems with goals of care initiatives across a specialty (i.e., primary care clinics, inpatient medicine wards, or intensive care units). We consulted national leaders to identify health systems engaged in these initiatives and reviewed published literature of goals of care initiatives. Rather than a comprehensive review of initiatives, we aimed to select a diverse group of health systems with varying geographic locations and initiatives to capture a broad range of approaches. We included eight health systems due to the size of the panel and the coordination of experts across health systems. These health systems included: BJC HealthCare, Duke Health, Providence, University of California, Irvine (UCI) Health, University of California, Los Angeles (UCLA) Health, University of California, San Francisco (UCSF) Health, University of Pittsburgh Medical Center (UPMC), and University of Washington (UW) Medicine.

Developed by JEM and GP and disseminated by email, an initial survey was completed by each health system about a key goals of care initiative from their health system (Supplement 1). The initial survey was followed by three 60-minute virtual sessions with a panel consisting of health system champions of goals of care initiatives. All health systems completed the initial survey, except for University of California Irvine (UCI), which was added later as part of the University of California Patient-Centered Outcomes Research Initiative (PCORI) project. Sessions took place between October 2023 to April 2024. Session dates were scheduled to ensure at least one representative from each institution could attend. One session was divided into two meetings to accommodate panel members’ availability. Sessions were facilitated by JEM, field notes were taken by KL and transcribed by Zoom. Summaries of each session were distributed to panel members by email for review and comment. Core elements of goals of care initiatives were identified during panel sessions and finalized after iterative review by panel members. Each health system provided key examples for each core element.

Health System and Panel Characteristics

Health systems included were geographically diverse (i.e., 2 East Coast, 1 Midwest, and 5 West Coast, including 3 from one state health system) and predominantly academic (7/8, 87.5%). Characteristics of eight health systems were summarized in Table 1. Goals of care initiatives spanned inpatient and outpatient settings. The panel included 19 champions with each session including 10 to 17 panel members.

Table 1:

Health System Setting, Size, and Population Characteristics

Health System	Duke Health	Providence	UCI Health	UCLA Health	UCSF Health	UPMC	UW Medicine	BJC HealthCare
Setting	Academic, Community, Urban	Community, Urban, Suburban, Rural, Critical Access	Academic, Urban	Academic, Community, Urban, Suburban	Academic, Urban	Academic, Community, Urban	Academic, Urban	Academic, Community, Urban, Rural
Number of Beds	1,400	11,000	1,300	1,000	1,100	8,700	1,400	3,100
Number of Hospitals	3	52	5	5	2	40	3	14
Geography of Hospitals	North Carolina	Alaska, Washington, Montana, Oregon, California, New Mexico, Texas	California	California	California	Pennsylvania	Washington	Missouri, Illinois
Admissions per Year	67,600	444,500	34,200	41,300	44,500	350,000	46,900	145,000
Outpatient Visits per Year	4.9 million	27.7 million	1.4 million	3.8 million	2.9 million	6.7 million	1.5 million	2.7 million
Electronic Health Record	Epic	Epic	Epic	Epic	Epic	Cerner	Epic	Epic

Considerations for goals of care initiative implementation

Health systems described multiple ongoing goals of care initiatives. The initiatives identified allowed for a rich discussion about key aspects of implementing a health system goals of care initiative. We identified 7 core elements that facilitated the implementation of goals of care initiatives at participating sites. Within each element, health systems shared a multitude of experiences that informed their goals of care initiatives. We summarized core elements utilized at each institution in Table 2. For this paper, each system provided detailed information on one overarching health system initiative rather than an exhaustive list of goals of care efforts at their institution (Table 3). Rather than a formal evaluation of effectiveness between health systems, Table 3 provides descriptive examples to illustrate how each health system implemented a goals of care initiative. Results in Table 3 were defined by each health system.

Table 2:

Core Elements of Goals of Care Initiatives Used by Individual Health Systems

Health System		Duke Health	Providence	UCI Health	UCLA Health	UCSF Health	UPMC	UW Medicine	BJC HealthCare
Defining the Population	Population-based (e.g. patients with a disease or condition, patients with serious illness, patients at high risk of mortality)	X	X	X	X	X		X
	Event-based (e.g. admission to ICU for at least 5 days, patients admitted to the hospital)		X	X	X	X
	Combined population-based and event-based (e.g., patients with advanced cancer and admitted to the hospital)	X	X	X	X	X	X	X	X
Stakeholder Engagement	Leadership engagement	X	X	X	X	X	X	X	X
	Non-specialty palliative care clinician engagement	X	X	X	X	X	X	X	X
	Patient and caregiver engagement (e.g., patient and caregiver advisory board)	X	X	X	X	X		X
	Other (e.g., accountable care organizations, payors, policy makers)		X	X	X				X
Encouraging the conversation	Clinician notifications (e.g., EHR best practice advisory or machine learning mortality model notification)	X	X	X	X	X	X	X	X
	Patient facing goals of care communication (e.g., messages prior to visit in EHR patient portal)	X	X	X	X	X		X	X
	Clinician-targeted education	X	X	X	X	X	X	X	X
	Patient-targeted education		X	X	X	X			X
Documenting Conversations in the EHR	Standardized goals of care and related documentation placement in EHR	X	X	X	X	X	X		X
	Goals of care documentation template	X	X	X	X	X	X	X	X
	Goals of care documentation marker to track notes	X	X	X	X	X	X		X
	Artificial intelligence to identify documentation (e.g., natural language processing)		X					X
Quality Metrics to Track Progress	Clinician documentation of goals of care conversations	X	X	X	X	X	X		X
	Advance directives and/or state portable medical orders		X	X	X	X
	Dashboard for up-to-date goals of care metrics	X	X	X	X	X	X		X
Sustainability and Scalability	Initiative interventions designed to continue long-term (e.g., after the end of the funded project)	X	X	X	X	X	X		X
	Developed interventions that can cross institutions (e.g., intervention can be implemented in another health system)		X		X	X			X
	Long-term funding for goals of care initiative (e.g., health system or payor sponsorship)	X	X	X	X	X	X		X

Table 3:

Detailed Descriptive Comparison of Goals of Care Initiatives

Health System	Duke Health	Providence	University of California Health System (UCLA, UCSF, UCI) PCORI study	UPMC	UW Medicine	BJC HealthCare
Population	Inpatient and outpatient using ICD codes and machine learning mortality risk models	ICU patients with length of stay at least 5 days or more	Primary care patients with serious illnesses (cancer, heart failure, COPD, end-stage liver disease, ESRD, or ALS) and at least 2 primary care visits in the last 12 months	Inpatient; machine learning mortality risk model predicting greater than 30% within 90 days	Inpatient and outpatient patients with chronic illnesses including metastatic cancer, COPD, heart failure, cirrhosis, dementia; Identified by ICD codes	Inpatient and outpatient patients with high risk of mortality based on machine learning mortality risk model
Education	Website with brief video snippets and links; Focused communication training for clinicians	Topics posted to internal learning resource; System-wide available communications skills training; One-page info sheet; Bi-monthly forum with presentations and Q&A with leaders, including executive and palliative care representation	Brief ACP training to primary care clinicians; Created training for healthcare navigators; Patient portal messages for patients on ACP	Mortality prediction model and goals of care documentation	1 page Jumpstart Guide; Clinician staff meetings	Small group, standardized patient training sessions for providers; Videos and informational text for patients.
EHR Tools and Templates	Template and free text notes available to track documentation using Epic dotphrase “bookends”	Templated note and integrated SmartPhrase with custom SmartLists for quality tracking that import to ACP Summary page SmartPhrase has multiple versions for more and less narrative vs structured documentation; Workflows and tools socialization with Hospital Medicine and Primary Care practices across the system	Template and dotphrase to track documentation in “ACP Navigator” in EHR for clinicians; separate note templates for healthcare navigators.	Templated note with surrogate decision maker, prognostic information, values, decisions, and next steps.	Customized SmartText in EHR for clinicians	Templated note and dotphrase in EHR
Measures	Documentation using note template or “bookends” in the last 6 months of life	Documentation using integrated tool with quality standards for the conversation	Advance directive and POLST in the EHR at 12 and 24 months (primary outcome), documentation, and in a research subgroup, self-reported advance care planning engagement, and goal concordant care	Documentation in a templated note	Documentation in the EHR identified using natural language processing with human adjudication	Documentation, mortality index, length of stay, palliative care utilization, hospice utilization
Data Reporting	Up-to-date dashboards on provider and specialty-based goals of care documentation	Dashboards with completion and quality measures; Real time reporting in the EHR	Dashboards for reporting on documentation	Dashboard with report of goals of care conversation completion by patient demographics and severity of illness, clinician, unit, and hospital	Not currently reported outside of research setting	Dashboard with process and outcomes measures
Results *	More than 60% of patients with goals of care documentation in the last 6 months of life	Over 85% met the ICU metric denominator with documentation; Over 99,000 notes written for admitted patients over 18 years old across the system	Among patients without an AD, 13% or more had a documented AD and more than 30% had a documented conversation at 24 months, with higher rates in the navigator group	Over 70% of patients with over 60% risk of 90-day mortality with documentation	Increase in documented goals of care discussions in EHR from 30% in usual care to 34% with intervention (inpatient)	Five-fold increase in documentation, three-fold change in code status, more than 80% adoption by hospitalists and significant decrease in mortality index

ACP: advance care planning; AD: advance directive; ALS: amyotrophic lateral sclerosis; COPD: chronic obstructive pulmonary disease; EHR: electronic health record; ESRD: end stage renal disease; GOC: goals of care; ICD: international classification of diseases; ICU: intensive care unit; POLST: Physician Orders for Life-Sustaining Treatment

^*Results were defined by each health system

Defining the goals of care initiative: “What are we trying to change?”

In general, health systems supported goals of care initiatives because of their potential to improve the quality of care, ensure that patients receive care in line with their goals and values, and affect key metrics, such as patient satisfaction, acute care utilization, inpatient mortality, length of stay, and 30-day readmissions. Examples of the aims of initiatives included: a Duke Health initiative ensuring that all patients seen within the health system had a goals of care conversation documented within the last six months of life, regardless of the setting of care, and a UPMC initiative increasing goals of care conversation documentation for hospitalized patients with high risk of mortality.^8,18,19 In a PCORI funded initiative, three University of California (UC) health systems, UCI, UCLA, and UCSF, collaborated to improve conversations and documentation among seriously ill patients followed by primary care clinicians.^20,21 Providence described an initiative focused on patients with longer intensive care unit stays to have at least one documented goals of care conversation.²²

Lesson: Participating sites gave examples of efforts to improve goals of care conversations for a variety of patient populations and healthcare settings. Areas of focus were often influenced by pain points for the health system or quality metrics implemented by an external entity. For instance, Duke Health was focused on improving inpatient mortality rates and decreasing inpatient length of stay, so the team focused an initiative on those populations that were known to be driving these measures (e.g., patients with advanced cancer and heart failure). Health systems planning goals of care initiatives need to thoughtfully consider how to synergistically align with health system goals.

Defining the population for the initiative: “What is the denominator?”

After defining the initiative, the next step these health systems took was to identify the population, whether by population, event, or a combination of the two.²³ In a population-based approach, University of California health systems created and validated an EHR phenotype to identify primary care patients with serious illness who would benefit from goals of care conversations. This phenotype included a combination of structured data elements such as specific visits and International Classification of Diseases (ICD) code combinations.²⁰ Machine learning mortality models were used to identify high-risk patients, and prognostic thresholds for machine learning models can be designed for specific target populations, shorter prognosis for hospitalized patients (e.g., inpatient, 30-day, or 6-month risk of death), compared to longer prognosis for outpatient populations (e.g., 1 – 2 years). Among the institutions that incorporated machine learning models in their inpatient initiatives, UPMC included patients with a high 90-day mortality, and BJC Healthcare identified patients with a high 30-day mortality.^18,24 In an event-based approach using length of stay, Providence identified patients with at least 5 days in the intensive care unit as a key population that could benefit from goals of care conversations.²² Duke Health incorporated a combination of methods, including machine learning models (high 30-day and 6-month mortality) among hospitalized patients to identify populations of interest.^19,25

Lesson: The denominator of a goals of care initiative should be tailored to the setting, intervention, and budget for the implementation, while also accounting for potential limitations. Length of stay is applicable only for the inpatient populations. ICD codes can identify patients with high-risk diseases, but are also dependent on whether these codes are entered into the EHR by the provider, and may be entered after the inpatient or outpatient encounter.²⁶ Machine learning models may improve accuracy, but these models are often limited to the institution where the algorithm was created and biased towards those who utilize services in the health system.^27,28 Furthermore, resources and expertise within a health system may influence the choice of a specific denominator. For instance, length of stay and ICD-10 diagnoses can be deployed at less cost than more complex models in most health systems. In contrast, machine learning models are available to large health systems with substantial resources, although these technologies have become much more broadly available in the past few years. Health systems should weigh the advantages of more complex strategies with the feasibility and cost of these strategies and ensure validation of the model prior to implementation.

Stakeholder engagement: “Whose buy-in is necessary?”

These 8 health systems all found that support from key stakeholders, such as patients, caregivers, front-line providers, and leadership, can be crucial to the success of an initiative. Each of these stakeholder groups was pivotal for the implementation and dissemination of the initiative across the health system. Patient and caregiver input ensured that the initiative was aligned with important patient and caregiver priorities. Collaboration among key clinicians helped support the initiative from the ground up. Front-line providers provided critical feedback so that the initiative was complementary and fit within the clinical workflow. In a PCORI-funded initiative, UCSF, UCLA, and UCI engaged with primary care clinicians.²⁹ In a separate initiative, UCSF met with hospital medicine clinicians, internal medicine residents, and chief residents, and palliative care leadership to design an inpatient goals of care initiative. Leadership buy-in ensured the initiative’s mission was aligned with health system goals. At Providence, leaders supported the adoption of key metrics on goals of care conversations across the health system.

Primary palliative care involves basic discussions of illness understanding, prognostic awareness, goals, suffering, and code status, and can be led by any provider, and may be compared to a specialty palliative care service that assists in more complex management and decision making between patients, families, and treatment teams.³⁰ These goals of care initiatives all leaned heavily on primary palliative care offered by clinicians not specialized in palliative care. In contrast to the health systems that emphasized primary palliative care to engage patients in these conversations, UPMC’s initiative triggered palliative care consults for hospitalized patients at the highest risk of mortality, if there was not a goals of care conversation documented by the primary medical team after the first 48 hours.⁸

Lesson: Engaging a broad group of stakeholders for the initiative, including health system leaders, clinicians from diverse specialties, patients, caregivers, and health plans, was essential. However, health system leadership support emerged as a main facilitator and barrier to success. Although none of our sites identified leaders who were significant barriers, several reported that it was often challenging to find leaders who could serve as effective champions. This challenge stemmed from the fact that serious illness communication fell outside the scope of most leaders’ responsibilities and is not tied to reimbursable quality metrics. As a result, initiative champions partnered with leaders whose purview included outcomes that could benefit from better communication (e.g., 30-day readmissions, length of stay, mortality).

Specialty palliative care is an important and often leading stakeholder for goals of care initiatives. It was important for health systems to distinguish the role and responsibility of specialty palliative care within the initiative. In the setting of a limited specialty palliative care workforce,^31,32 each health system initiative primarily focused on non-specialty palliative care providers to initiate conversations.

Encouraging the conversation: “How can we facilitate goals of care conversations?”

Health systems used various approaches to encourage conversations with patients in the target group(s). These included technology and education intended for clinicians, patients, and caregivers. Prompts notify clinicians of patients appropriate for goals of care conversations and can take the form of EHR notifications, such as a Best Practice Advisory (BPA) within Epic, or an email or page notification outside of the EHR. Patient and caregiver-facing technology in the Epic patient portal included a goals of care survey at Duke Health and an engagement note on advance care planning in the University of California health systems.^20,33

Prompts helped remind clinicians to have goals of care conversations. The design of the prompts could play a role in how providers respond to the notification. For example, at Duke Health, the prompt was designed into the Epic storyboard as a “soft” alert compared to a traditional “hard stop” BPA. “Hard stop” notifications forced clinicians to respond to the prompt before they could act on another part of the EHR. In comparison, the Epic storyboard “soft” alert simply changes the color of a sidebar to suggest having a goals of care conversation and does not require a clinician to interact with the prompt.³⁴ Actions facilitated by storyboard notifications are less interruptive, but their effect cannot be measured as directly compared to a “hard stop” BPA. Using a different approach, BJC HealthCare sent notifications using the EHR clinician chat message feature.^35,36 UW Medicine developed prompts outside of the EHR, via clinician email and page notification.⁹ Clinician alerts were also prompted by non-clinician navigators in the University of California health systems.

In addition, most health systems incorporated education to enhance patient understanding and motivation and/or clinicians’ skills and confidence in leading effective goals of care conversations. For instance, UC Health (UCLA, UCSF, and UCI) educated patients directly through patient portal messages and links to advance directive paperwork, the PREPAREforYourCare.org website, and healthcare navigator outreach.^29,37 Health systems also focused on clinician communication skill training. UC Health (UCLA, UCSF, and UCI) included a workflow-informed brief clinician training about ACP in the primary care setting. Duke Health, UPMC, and UW Medicine trained between 200 to 1,000 clinicians on communication skills, including VitalTalk, Mapping the Future, and Jumpstart.^9,19,38,39 Education also extended beyond communication skills and included topics such as how to document goals of care conversations within the EHR, how to bill for these conversations, and where to find this documentation in the EHR.¹⁹ Duke Health and Providence developed internal and focused learning resources for clinicians.

Lesson: To promote adoption within the health system, clinician and patient education and nudges should consider clinical workflow and costs. Reminders to patients and clinicians can facilitate conversations by introducing goals of care conversations, particularly for patients unfamiliar with these topics. For clinicians, such prompts can help reinforce the integration of goals of care into routine practice. Furthermore, the budget may impact the choice of the type of approaches to encourage conversations. Live, interactive communication programs may be more effective but also more costly than online, asynchronous education, which cannot be tailored in real time for learners. Additionally, implementing prompts within the EHR may require build time and financial investment. Moreover, prompts need to build on a foundation of existing familiarity with goals of care and incorporate clinician workflows. Otherwise, such initiatives may not meaningfully change clinician behavior on goals of care.

Documenting conversations: “How are these conversations documented in the EHR?”

Health systems considered the design, user experience, and education on goals of care documentation. First, health systems weighed the ease of access to and the location of documentation within the EHR. Several health systems using Epic adopted a ‘tab’, centralized location, or navigator/dashboard with goals of care documentation and other relevant documentation, such as advance directives or portable state out-of-hospital orders (e.g., portable orders for life-sustaining treatment (POLST)). Many health systems added advance care planning to the Epic Health Maintenance Tab for patients and clinicians, such as for vaccination and cancer screening. Duke Health developed a link to the ‘tab’ in the patient’s storyboard. This link changed colors when a patient did not have any goals of care-related documentation. Clinicians could hover over the link to see recent code status orders and goals of care notes.³⁴

Most health systems used templates to guide text entered and track documentation (Table 3). Components included in this documentation varied by health system. One health system found that templates and provider specialty impacted the type of information documented into the EHR.^40,41 Within Epic, “dotphrases” or “SmartPhrases” were implemented to identify notes as goals of care documentation. However, within Duke Health, while documentation was tracked and linked to one central “tab” in the EHR, a number of notes were duplicated or used the “dotphrase” inappropriately with content unrelated to goals of care, e.g., a car seat at discharge from the hospital.⁴¹ Providence and UW Medicine deployed machine learning and natural language processing methods to identify goals of care documentation in the EHR.^22,42 Providence used machine learning analytics to ensure that these notes met the quality metrics of goals of care documentation. Each system described considerable time commitment, paid or unpaid effort, resources, and leadership buy-in necessary to build and test the interface for this documentation within the EHR.

Lesson: As clinicians are pulled in multiple directions to balance clinical care, documentation, and billing, easily identifiable and accessible goals of care notes may help facilitate clinical workflow to align patient goals with healthcare. A common location is pivotal when point-of-care decisions are needed, such as in the emergency room; prior conversations may help clinicians navigate the current conversation.^43–46 One health system emphasized to front-line clinicians the need to document notes as a patient safety issue. Messaging to health system leaders, communication training, and bedside teaching by palliative care emphasized that these conversations already take place and that documenting them in the goals of care tab enabled other clinicians to build on these prior conversations. Several health systems prioritized placing GOC content in a designated place in the EHR, so that clinicians easily find and see goals of care notes in hopes that this constant exposure will further encourage conversations and documentation.

Quality metrics: “How can we track progress?”

In designing these initiatives, health systems needed to figure out how data on goals of care conversations could be collected and measured, and which data would be effective in quality improvement efforts and in strategic alignment with health system priorities. Specifically, health systems focused on structured data elements that could be feasibly counted and compared. These included goals of care notes and note components, advance directives, and portable medical orders (i.e., POLST).

Based on these structured data elements, data dashboards were created to track and provide initiative leaders with feedback on the progress of the initiative. UCLA designed a dashboard to report documentation by provider and clinic. Duke Health’s dashboard displayed the percent of patients with goals of care documentation in the last 6 months of life, the percent of all patients who had a goals of care conversation, and, more recently, the percent of goals-of-care conversations for which a provider billed. Providence created a dashboard to track both the number of documents and quality, where a minimum standard for quality documentation was developed to include the participants, desired medical treatments, and a description of the conversation.²² UPMC’s dashboard tracked the percentage of patients with goals of care documentation for patients with a predicted risk of mortality greater than 30% by hospital location, clinician specialty, patient demographics, including race and ethnicity, and area deprivation index.

Lesson: Leaders of initiatives need to establish metrics of success to iteratively improve the initiative. While meaningful initiative measures differed within each health system, they similarly focused on documenting at least one goals of care conversation for a patient or a conversation within a certain time period. Aligning initiative metrics with health system-wide metrics, such as length of stay, may gain support of health system leadership. Furthermore, establishing clear metrics and systematically tracking data can create an opportunity to assess the success of initiatives. Therefore, special consideration should focus on which data to measure and the potential impact on health system metrics.^47,48

Sustainability and scalability: “How will the initiative keep going?”

Sustainability and scaling an initiative depended on how feasibly these efforts could be integrated into routine clinical workflows and on the financial support needed to continue these initiatives. UC Health Systems specifically built automated patient identification and messaging into the EHR, so that the health systems could continue patient identification and messaging on goals of care after completion of the PCORI grant.²⁰

Notably, several health systems faced difficulty financing these initiatives, especially if goals of care was not an immediate priority for the health system or when grant funding for pilot or research projects ended. Leaders of the initiatives were often supported through institutional resources, external grants, or even donated time to launch the initiative. Such support was typically limited and insufficient for long-term sustainability. Additional funding was necessary to maintain key data reporting, upkeep dashboards, and educate new clinicians. BJC HealthCare continued its initiative by collaborating with the accountable care organization. Specifically, BJC HealthCare advocated that their initiative aligned with health system priorities, and the accountable care organization supported scaling their initiative from inpatient medicine to primary care.

Lesson: Health systems must proactively plan for the sustainability and scalability of their initiatives.⁴⁹ Health system priorities need to be embedded within goals of care initiatives to gain health system leadership buy-in and secure funding. While there is leadership recognition on the importance of goals of care, sustainment will need to consider the overarching health system priorities.⁵⁰ Additional focus is needed on the long-term impact of the goals of care initiative to meet health system leadership objectives. Initiatives must take note of system leadership priorities, which may include utilization metrics, financial measures, or quality indicators as well as the return on investment for health systems. Furthermore, sustainability should consider those responsible for continuing to support the initiative, whether tracking data, upkeep of data dashboards, or promoting the initiative. Lastly, one difficult but high-yield goal for sustainment is institutional culture change, where goals of care conversations become a part of every provider’s responsibility.

Discussion

To increase engagement in serious illness communication, this paper describes the implementation of goals of care initiatives across eight U.S. health systems. While this was not an exhaustive summary of initiatives, we identified seven core components of successful programs, including defining the initiative, identifying the population, engaging stakeholders, promoting, supporting, documenting, and measuring conversations, and lastly, sustaining and scaling the initiative.

Health systems aiming to establish new goals of care initiatives must address several key needs. First and foremost, strong leadership is key to the development, buy-in, and sustainability of initiatives. Champions within health system leadership critically help promote initiatives across the organization. Second, leadership of the initiative must extend beyond specialty palliative care and exist at all levels.⁵¹ Specialty palliative care and non-specialty palliative care roles within the initiative need to be well-defined. Particularly, the included health systems did not depend on specialty palliative care to lead and conduct all the conversations; instead, specialty palliative care proactively engaged and collaborated with non-specialty providers to move the initiative forward. Third, it is important to identify the stakeholders, clinicians, patients, and caregivers, beyond leadership. Engaging a broad group of stakeholders can ensure that the initiative integrates into the routine clinical workflow, which is a key component to the success and sustainability of such initiatives.²⁰ That’s particularly true given the complexity of these initiatives.^15,29,52 Lastly, the initiatives must have clear metrics of success, particularly goals of care-related metrics that align with health system leadership priorities.

Along with leadership support, identifying long-term methods to finance initiatives may be key to continued progress.⁵³ Each health system approached financial support for the initiative differently, from garnering support from leaders using health system metrics or collaborating with accountable care organizations. Beyond the initiative-specific implementation costs, initiatives also should consider non-healthcare or out-of-pocket costs to patients and caregivers, potential cost-savings in health system outcomes, revenue from billing conversations, and how payors may also influence initiatives.^54–57 Beyond financial considerations, further research on how to best measure how goals of care initiatives benefit patients, families, providers, and health systems is needed.

Finally, we found that each health system took very different approaches. For instance, there was no consensus about essential components of goals of care documentation or specific data measures which can hinder the comparisons between initiatives between health systems and identifying best practices. Specifically, research needs to identify which measures can be used to grow programs and potentially even impact state or national policy.⁵⁸ By identifying the necessary and standardizing elements and components, health systems will be able to incorporate best practices when developing and adapting goals of care initiatives.

This preliminary investigation of health systems’ efforts to improve serious illness communication has several limitations. Purposive sampling of health systems certainly missed other health systems engaged in goals of care initiatives. Our findings relied on self-reported information from each health system. Our study was not a comprehensive survey of the population-based goals of care initiatives, but rather an initial effort to highlight the range of approaches and common issues facing health systems trying to implement goals of care initiatives. Due to the variability in each health system initiative’s design and measures, we were unable to assess the impact of individual core elements on initiative outcomes. Although the true range of initiatives is much more diverse at each site, we were nevertheless able to identify several themes and lessons that we believe will be useful to health systems that are contemplating an initiative to improve serious illness communication.

Among the eight health systems included in this paper, the health system goals of care initiatives’ content, structure, and outcomes vary significantly. Despite the variance between each initiative, each health system included multiple steps, from defining the initiative to identifying methods to data measures, to successfully implementing and increasing goals of care conversations. The development of the best evidence-based practices will support health systems aspiring to launch their own goals of care initiatives.

Key Message:

As health systems increasingly seek to integrate goals of care conversations into routine practice for patients with serious illness, implementation guidance is needed. Implementation of goals of care initiatives needs to consider the purpose of the initiative, the target population, key stakeholders, provider prompts, documentation, data measurement, and sustainability. These elements of goals of care initiatives may support effective implementation within health systems.