The Urgency of Now — Ask your patients about goals of care. Here’s how health systems can do it.

13 min readApr 5, 2020

by: Azalea Kim, MD, MBA, MPA and Jared Lowe, MD

I connected with one of my favorite patients this week by telephone. Our usually monthly appointments have been derailed due to COVID-19 and we are learning what it’s like to check in through telephone visits. We get through talking through blood pressure medicines, blood glucose numbers, and then there is a pause… I use the pause to ask:

“What is your understanding of what’s happening with coronavirus? Have you thought about the kind of care you would want if you got very sick?

While it’s the first time my patient and I have spoken about coronavirus, it’s not the first time I’ve asked questions like this. We’re lucky in that way.

Jared and I trained together in internal medicine residency. We spent a large part of our training working on the design and implementation of a pilot that integrated advance care planning into the workflows of a primary care practice. We did this work well before there was the specter of COVID-19 pandemic. We did this because it’s what our patients and their families wanted and needed. It’s always felt like chicken soup for the clinician soul.

Advance care planning is a necessary innovation to deliver high value care that is concordant with the wishes of our patients. But at its core, it’s about as close you can feel to another human who is in your clinical care — it’s a closeness we can still achieve even though we have to stay physically distanced.

Now in the context of the COVID-19 pandemic, never has advance care planning and serious illness planning felt so necessary and relevant to more people; yet, so elusive in how to actually do this and at scale.

Our pilot study done in 2017-2018 touched the lives of at least 120 patients and their families in the span of 22 weeks through successful engagement in advance care planning conversations. We painfully designed, tested, iterated, and deployed a series of simple tools to help patients, families, and clinicians overcome the inertia of engaging in challenging conversations.

We’ve shared our findings at Society for General Internal Medicine’s Annual Meeting 2018, but then spent the last two years failing to find the right venue to publish what we learned. However, we are both of the opinion that there is an urgency to share this information now with as many people as possible.

What we created here was developed in a pre-COVID-19 world. As we describe the pilot and what we learned, we’ll bold and italicize parts that deserve commentary based on present circumstances. They may take into account the role of virtual care (i.e,, telehealth), the need for social distancing, and the acuity of disease. By sharing what we learned in this process, we hope to equip clinicians and teams waging heroic responses to the COVID-19 pandemic.

What is ACP and why is it hard to do?

Advance care planning (ACP) is a critical component of delivering high value, patient-centered care. There are numerous challenges to engaging a population of primary care patients in ACP and doing so at scale including:

Identification of patients who would most benefit from ACP;
Outreach and engagement of patients and caregivers in a challenging topic;
Design of clinical workflows that enable accurate completion of ACP-associated documentation and accessibility of these data across care settings;
Patient-centered design of educational materials; and
Expansion of clinic provider bandwidth to allow sufficient time to have comprehensive conversations with patients.

Implementing a scalable solution for advance care planning (ACP) delivery across a health system is fraught with challenges. Engagement by patients[1] and providers is low: in 2017, only 2.2% of Medicare beneficiaries had ACP CPT codes billed, with variability of their use based on specialty, region, and care setting. [2,3]This gap points to the barriers to delivery of ACP cited by clinicians, including insufficient time, lack of communication training, complexity of treatment options and decision making, patients’ hesitation to engage, and lack of access to ACP information across care settings.[4,5,6,7] Patients’ trusting relationships with primary care providers (PCPs) suggest these practices may provide the right venue to engage patients in potentially difficult conversations regarding serious illness. [8,9] Yet, many of these barriers exist in primary care.

What did we do?

To address these challenges we used a simple patient identification algorithm that relies on easily available electronic health record (EHR) data, existing EHR functionality to share ACP data across health care settings, and patient-centered design to make a difficult topic more approachable for patients.

Our goal was to design and implement a scalable, patient-centered workflow for ACP in our primary care practice.

In the setting of COVID-19, these workflows will need to be adapted to meet the profound need for ACP, and settings, such as virtual encounters, in which they may be conducted.

Our ACP Pilot from A to Z

Our practice was an urban, academic primary care practice that cares for an underserved population in Durham, NC. As of July 1, 2016, at the time that pilot planning, the clinic estimated 4,043 unique patients attributed to the practice and approximately 20,000 patient visits a year. Our intervention focused on the 2,843 patients attributed to primary care physicians (PCP) who were resident trainees in the internal medicine program affiliated with our health system. The clinic is part of a health system that provides care in both inpatient and outpatient settings and is dedicated to improving patient engagement in ACP but had not yet embarked on a system-wide strategy.

Here is a visual breakdown of the pilot.

Overview of the Advance Care Planning in Primary Care Pilot

Patient Identification and Prioritization

We developed an algorithm to guide prioritization of patient outreach. We sought to identify patients facing serious illness, as well as patients further upstream in their clinical course. We reviewed published literature at the time to identify potentially relevant models and developed a simplified model informed by the variables described by Mathias et al. to distinguish life expectancy of less than 5 years among patients age 50 and older.[10]

Criteria for inclusion in our cohort were: age 70 or greater, with 3 or more of a specified listing of comorbidities as identified in the EHR identified through diagnostic codes, or with 2 or more hospitalizations within our health system in the preceding 365 days. A patient meeting any of these three criteria was included in our cohort. Each criterion counted as one “point.” Patients with higher scores were prioritized for outreach. Patients were excluded if they were younger than 50 years of age or had EHR evidence of having an “ACP Note” filed by a provider.

As there is no “gold standard” to identify patients who should engage with ACP, [11] we developed an exercise to evaluate our model’s ability to identify patients whom PCPs would deem “appropriate” for ACP interventions. The exercise, referred to as our “PCP Concordance Exercise,” was conducted by asking a subset of PCPs at the practice to review the list of their attributed patients that was generated by our model. The PCPs were asked to indicate whether, in their opinion, these patients were “appropriate” or “inappropriate” for an ACP intervention.

This definition of appropriate becomes even more subjective in the setting of COVID-19 and pandemic. One might argue that every single person would benefit from the opportunity to engage in ACP. As we will detail later in this piece, the algorithm is the matters of least importance. What matters the most is finding the patient-centered workflow that you and your colleagues can scale.

The Team

Two Patient Navigators (PNs) were onboarded with the practice to serve as part-time PNs. Both PNs had training in social work but no prior experience working in ACP. Our process included an introduction to ACP [12] and clinic operations by the principal investigators, instruction on completing authorized health care power of attorney (HCPOA) forms compliant with North Carolina state law, and training on the EHR (Epic). PNs also spent time observing a palliative care team and a care management team affiliated with our health system.

To prepare all staff and providers at the practice for program implementation we hosted a series of “eat and learn” sessions where the principal investigators provided education on ACP and reviewed logistics. Many staff members volunteered to share personal experiences with ACP and their care of loved ones facing serious illness during these sessions. The implementation team consisted of the clinic Medical Director, Nurse Manager, a primary care physician, and two internal medicine residents.

COVID-19 has underscored the tremendous value of multi-disciplinary teams and the opportunity to “retrain” team members into new roles. As many primary care and outpatient clinics move to virtual-heavy models, now is the time to consider “up-training” clinical staff to support ACP workflows. There are a wealth of resources available, among them:

This visual guide by Dr. Nathan Gray is an incredible resource for getting started in ACP in the setting of COVID-19.

Patient-Centered Approach to Engaging Patients

A local patient advisory council reviewed and approved all patient-facing materials. The council recommended the use of the term “care planning” to best introduce ACP to patients. We wrote all patient-facing materials at a grade 4 or lower reading level, with some limitation due to required legal terminology.

To engage patients in the program, the PNs reached out to patients by phone using scripts to invite them to in-person “care planning” appointments at the practice. If patients agreed to the appointment, PNs scheduled the visit. Patients were encouraged to bring potential surrogates, such as family, friends, and caregivers, to the appointment. Here is the outreach script:

Sample Outreach Script for Patient Navigators

It would be very reasonable to include mention of COVID-19 in the script above; however, while ACP is very germane in the setting of pandemic, its mention may also trigger anxiety. Additionally, the script above will need to be modified to the extent to which patients can complete advance care planning documents remotely.

Completing ACP Appointments

PNs completed all “care planning” appointments on site at the primary care practice and systematically documented these encounters. PNs prioritized three goals for the appointment:

Identification of surrogate decision makers and completion of a legal designation of a health care power of attorney, including notarization of forms and scanning completed forms into the EMR.
Eliciting and documenting patient-reported values and goals.
Engaging surrogates and family members in ACP.

We developed a suite of tools to support PNs in achieving these goals:

ACP Note Template: We developed and tested a note template in the EMR to guide PNs through a focused ACP encounter.

· Patient Education Cards: PNs provided each patient with a two-sided piece of card stock with educational information outlining the ACP process and covering “Frequently Asked Questions”

· Patient Instructions for After Visit Summary: After the appointment, patients were provided a summary of the ACP encounter and instructions encouraging follow-up conversations with their surrogate decision makers.

· Simplified Legal Form: PNs used a version of an Advance Directive for the State of North Carolina for patients with limited health literacy.[13]

PNs used the dedicated “ACP Activity” in the Epic EHR platform as the central repository for scanned legal forms and ACP notes, thus making our data accessible across all sites in our health system. They were also able to directly message PCPs within the EHR to call attention to patient-identified concerns or questions that might require follow-up.

An important modification to the workflow above will be for virtual encounters. In many ways, virtual encounters may prove to be more comfortable and convenient, especially for frail patients. By leveraging virtual encounters, caregivers may be more readily available to join these ACP conversations to support patients.

What did we find?

Development of the algorithm and validation exercises was conducted April 1, 2016, through March 31, 2017. The pilot was conducted between July 1, 2017, and April 30, 2018, which includes the period of outreach to patients and analysis of results. All care planning appointments occurred between July 31, 2017, and February 1, 2018, a period of 22 weeks.

Validating the Model through PCP Concordance Exercise

In the PCP Concordance Exercise, 278 patients identified by the algorithm were reviewed by 28 resident PCPs. There was strong concordance between the model and PCP judgment as a clear majority of patients, 253 (91%), were marked as “appropriate.” The other 25 (9%) patients were marked as “not appropriate” primarily for perceived logistical barriers to completing a Care Planning appointment.

Algorithm and Description of Cohort

The algorithm was applied to the patient panels of resident PCPs to identify patients using data from April 1, 2016, to March 31, 2017. A total of 479 patients were identified. This totaled 16.8% of the 2,843 patients who were targeted for the intervention.

Patient and Surrogate Engagement

Of the 479 patients identified, PNs attempted outreach to 255 and successfully contacted 209, with 163 patients agreeing to schedule an appointment. Patients were often more willing to schedule an appointment if it could occur immediately before or after an already scheduled primary care visit. A total of 46 patients (22%), declined to schedule a care planning appointment.

Care Planning Appointments

PNs documented care planning visits with 120 unique patients. Of these, 118 (98.3%) resulted in documentation of any discussion of goals, values, and/or care preferences. A total of 110 (91.7%) of the visits resulted in an identified surrogate decision maker, including completion of a legal HCPOA form on-site or documentation of the surrogate’s identity in the ACP note in the EMR.

What were the biggest hurdles?

There is potential for misperception by the patient community of the intent of an intervention like ours. [14]Our engagement with the patient advisory council was a critical early step to gain feedback and insight.

We initially faced difficulty with scheduling patient visits with the PN. Most patients cited limited time and already having too many appointments as barriers. Scheduling rates increased when PNs were given access to the PCP clinic schedules and scheduled “care planning” appointments immediately before or after their PCP appointment.

We found that 10 HCPOA forms notarized on-site and scanned into the EHR were not mapped appropriately to the ACP activity and could not be accessed by providers. We created redundancy to the process by use of a ACP cover sheet and kept copies of the original patient forms securely in clinic medical records until confirmed that a scanned copy was retrievable in the EHR.

The initial algorithm used in the pilot was developed with ICD-9 codes and was later updated using SNOMED CT reference terminology for ease of EHR integration. There are and will continue to be more sophisticated models for patient prognostication.[15] However, at the start of our pilot none could be easily calculated within our EHR and applied readily into a clinical workflow. To promote clinician adoption and acceptability, we prioritized transparency and adaptability when translating our algorithm into an ACP Score in the EHR. The score explains which criteria (age, hospitalizations, and/or co-morbidities) contributed to a patient’s score, is refreshed daily, and available for use across tools including reports for multi-disciplinary rounds, triggers for best practice workflows, and the patient schedule view.

How can you get started?

We recommend first solving for the pain points to implementing a scalable workflow. In our experience, this included asking several questions:

How will you identify patients? It is likely that your organization has several options for identifying patients, including mortality prognostication tools. There are benefits and limitations to all models, and it’s important to find one that matches your use case.
Who has the bandwidth and skillset to engage patients? A PN working alongside the PCP was acceptable to patients, thus enabling ACP encounters that did not solely depend on PCP time.
What content would you cover in a care planning appointment? The scope of ACP can be extensive. We recommend at a minimum the identification of a trusted, legal decision maker and eliciting a patient’s goals and values, rather than trying to answer specific questions about life-sustaining therapies.
How will your ACP documentation be useful across care settings? It is important to understand what EHR functionality are available and to ensure systematic use. We recommend starting first with understanding the clinic workflows and educating all providers and staff.

KEY TAKEAWAYS:

A scalable workflow to meet the ACP needs of a population of patients in the primary care setting is feasible. This can potentially be adopted in the specialty outpatient setting. In the setting of COVID-19 they are amenable to adaptation to virtual encounters.
Integrating a patient advisory council in the design of this program helped make a difficult topic more approachable for patients and their families.
Simple algorithms that rely on readily available EHR data may be a starting point for identifying patients for ACP.
Patient navigators provided the necessary bandwidth for the delivery of ACP in primary care. Skills in communication and team-based care with PCPs are most critical.

Our pilot demonstrated feasibility and scalability for a practice and health system, thus limiting generalizability. Further research is needed to determine optimal methods to identify and engage patients and to evaluate the impact on patient-centered outcomes.

ACP is more important now than ever. We sincerely hope these resources and tools can help clinicians, practices, and health systems design and deploy scalable workflows for ACP.

We grant permission for you to use the resources that we developed for our pilot and to modify them to meet your needs. We also welcome your questions and comments and invite you to contribute your perspective on how to expand access to ACP during this critical time.

We want to acknowledge Lynn Bowlby, MD, Lawrence Greenblatt, MD, Indhira Udofia, MDiv, MSW, and Chenita Jackson for their contributions to the ACP pilot.

You can follow us on twitter @azaleakim and @jaredlowe18