Assessing Goals-of-Care Documentation during the COVID-19 Patient Surge in an Academic Safety-Net Medical Center

Introduction

The novel coronavirus (SARS-CoV-2), known as “COVID-19,” is a new disease with a wide array of clinical outcomes ranging from asymptomatic infection to death from acute respiratory failure.¹ As with any serious life-threatening illness, it is paramount that clinicians assess and document patients' personal priorities to guide goal-concordant medical interventions during the pandemic.^2,3 The electronic medical record (EMR) holds great promise for stewarding patients' values and treatment preferences,⁴ yet frontline clinicians report low confidence in retrieving information about goal-oriented conversations during a crisis.⁵ In response, national experts in palliative care have recommended standardized easily retrievable EMR documentation across settings of care, and in conjunction with clinical education and skills training in serious illness communication.⁶

We offer an example from our division of palliative care at UMass Memorial Medical Center (UMMMC), a 781-bed academic medical center colocated with the University of Massachusetts Medical School. Prepandemic, we developed a “dot phrase,” or structured template within a clinical note, to elevate the documentation quality of goals-of-care (GOC) conversations based on the Serious Illness Conversation Guide from Ariadne Labs.⁷ The dot phrase, funded by an internal grant and launched in 2019, contains an electronic bookmark linking the clinical note to a dedicated module (the “advance care planning navigator” in Epic) for easy retrieval by busy clinicians in any setting of care (Supplementary Appendix SA1). We highlighted the critical need to document patients' GOC, including use of the dot phrase, as part of ongoing training programs in serious illness communication for trainees and staff at our academic medical center.⁸

After the pandemic exploded in New York City, our palliative care leadership deployed multiple strategies to prepare for the patient surge at UMMMC, including membership on its COVID leadership council, which set system-wide clinical practice guidelines. As a result, these guidelines required clinicians to assess COVID patients' GOC within 48 hours of admission (Supplementary Appendix SA2).⁹ We positioned our palliative care team as system leaders in facilitating these conversations and EMR documentation (Supplementary Appendix SA3).⁹

In this retrospective chart review, we sought to measure the impact of our efforts during the pandemic using EMR documentation to discern the numbers, timing, and content of all GOC conversations among a wide array of COVID patients (ranging from good health to end-stage illness) seen at some point by palliative care during their hospitalization. In addition, we examined all clinicians' use of the dot phrase during the peak incidence of COVID-19.

Methods

Inpatients were identified using the clinical system's EMR (Epic) who were admitted with COVID-19 and seen by palliative care between the dates of March 23 and May 31, 2020. Patient demographic data were downloaded from the EMR. Medical records were then reviewed abstracting data for admission and discharge location (including use of hospice); length of stay (LOS); date and content of GOC discussions; timing of palliative care consult orders and completion; completion of a health care proxy (HCP); preadmission chronic medical conditions (e.g., asthma, chronic obstructive pulmonary disease [COPD], diabetes, and dementia); major comorbidities (e.g., cancer, HIV, and cirrhosis); use of high-flow oxygen and/or intubation; and patient status at discharge (alive vs. deceased). GOC discussions were also abstracted from the EMR. Keyword searches included family meeting, GOC, prognosis, values, quality of life (QOL), priorities, code status, treatment decisions, end-of-life, hospice, and comfort care. GOC discussions were also identified using the dot phrase/structured template generated through direct reports from the EMR. Each documented discussion was assigned to one of 12 categories: (1) code status discussion only; (2) code status discussion, prognosis; (3) code status discussion, prognosis, and goals; (4) code status discussion, prognosis, goals, and QOL; (5) discussion of other treatment decisions; (6) discussion of other treatment decisions and prognosis; (7) discussion of other treatment decisions, prognosis, and goals; (8) discussion of other treatment decisions, prognosis, goals, and QOL; (9) general discussion of prognosis and no treatment decision; (10) general discussion of prognosis and goals, and no treatment decision; (11) general discussion of prognosis, goals, and QOL, and no treatment decision; and (12) other comments not otherwise preidentified.

For analytic purposes, these 12 categories were consolidated into three types of GOC conversations: (1) code status decisions (i.e., confirming or changing code status and use of life support throughout the hospital course); (2) other treatment decisions (i.e., gastrostomy tube placement, tracheostomy, dialysis, discharge, artificial nutrition, chest tubes, transitions from the hospital ward to the ICU [or vice versa], implantable cardioverter defibrillator (ICD) deactivation, and hospice and/or transition to comfort-based care); and (3) no treatment decisions but a general discussion of prognosis, QOL, and values. Of note, code status orders alone did not suffice as proxies for GOC conversations. The chart review extended up to a maximum of five GOC discussions (GOC1–5) per patient. Finally, length of time from admission to first GOC discussion, and time from admission to palliative care order and consult completion were measured to estimate palliative care specialists' involvement in initial GOC discussions.

Patients' characteristics, medical history, and location from which they were admitted underscored four distinct disease severity categories: (1) patients with dementia admitted from rehabilitation centers (“rehab”), dementia units or long-term care (LTC) (Category 1); (2) patients with other chronic/serious illness admitted from rehab or LTC (Category 2); (3) patients with chronic/serious illness admitted from home or assisted living facilities (ALF) (Category 3); and (4) patients with minimal (one) or no chronic illness admitted from home (Category 4). The three types of GOC discussions were evaluated by number of conversations, timing, and content for all patients (Fig. 1) and within each patient category (Figs. 2–5).

Demographic data were downloaded from the EMR into a secure Microsoft Excel spreadsheet. The remaining manually abstracted EMR data were collected and managed using a secure web-based software platform, REDCap (Research Electronic Data Capture).^10,11 All data were transferred to a statistical software package (IBM SPSS Statistics for Windows, Version 27.0; IBM Corp., Armonk, NY, USA) and merged for analysis. Univariate statistics (frequencies, percentages, means, and other measures of central tendency) were used to describe all data elements.

The study was reviewed and approved by the University of Massachusetts Medical School Institutional Review Board (H00020463) under expedited review policies and procedures.

Results

Demographics and clinical outcomes

Table 1 displays the demographic and clinical outcomes of the total study population as well as within each of the four patient categories (described earlier). Overall, there was a near equal divide of patients by gender, but the majority of patients were white (74.2%), non-Hispanic (83.3%), had a mean age of 77 (ranging from 43 to 99 years), and had a HCP in place at the time of the palliative care consult (79.2%). Of note, people living at home or in an ALF were more likely to be black or Latinx compared with other groups, had longer hospital LOS, and were more likely to be intubated and for longer periods.

Among all 90 patients, the average LOS was 14.7 days (range: 0–117 days). More than three-quarters (78.9%) of the patients did not receive high-flow oxygen; among those who did, the average length of use was 3.2 days (range: 0–13 days). Approximately two-thirds (64.4%) of patients were not intubated and the average length of mechanical ventilation was 14.2 days (range: 1–39 days). Three-quarters (75.3%) of all patients did not receive hospice services during their illness. Patient survival status at the time of discharge was evenly split between alive (48.9%) or deceased (51.1%). However, nearly all patients who survived (90.9%) were discharged to a facility with or without hospice and only four patients returned home. Unfortunately, many patients who survived to discharge died within three months (per analysis of local newspaper obituaries), resulting in a case fatality rate of 76.7% among all 90 patients as of August 16, 2020.

Among the four patient categories, all patients who were younger, living at home, and without major comorbidities (Category 4) were intubated except for one patient. Two-thirds (66.7%) died in the hospital, and none received hospice services. Two of the surviving four patients from Category 4 were discharged home and all remained alive at three months after chart review. Patients with dementia and/or major health comorbidities (Categories 1–3) were more likely to have a completed HCP on admission (>80% of patients in each category), providing hospital teams with a surrogate decision maker for incapacitated patients. Patients with major health comorbidities living at home or in an ALF (Category 3) were more likely to receive high-flow oxygen and/or intubation and for longer periods compared with patients in Categories 1 and 2. In general, patients with major health comorbidities (without dementia) from Categories 2 and 3 had the highest case fatality rates (inclusive of the postdischarge period), and were more likely to receive hospice services compared with Categories 1 and 4. Among the survivors in Category 3 (who were living at home or ALF before admission), the majority (80.0%) were discharged to a nursing facility.

GOC documentation

Our system-wide guidelines called for assessment of patients' GOC within 48 hours of admission, and 87 of 90 patients (96.7%) had GOC documentation during this time frame (Table 2). Meanwhile, more than one-quarter (28.9%) of palliative care consults were ordered and 26.7% of patients had a consult completed within 48 hours of admission. Nearly one-half (48.9%) of palliative care consults were conducted within two days after the first GOC conversation by another provider. All (100%) palliative consult orders resulted in a completed palliative care consult within 24 hours of referral.

We analyzed patterns in numbers and timing of GOC conversations among various patient categories (Table 2). Of the total 90 patients, 87 of them had an initial GOC conversation, on average, one day after admission (range: 0–20 days). Less than 1 in 10 patients (6.9%) had only one GOC discussion during the admission, whereas one-third (35.6%) had at least five GOC conversations. On average, most patients across categories had three to four conversations during their hospital stay. Among the highest risk patients (Categories 1–3), 96% had a documented GOC discussion within 48 hours, whereas the healthiest patients (i.e., few comorbidities on admission; Category 4) had their first GOC conversation at median hospital day 6.5. Patients with dementia living in nursing facilities (Category 1) had GOC documentation within hours of admission (63.6% on day 0). In general, the five GOC discussions were spaced two to three days apart during the hospital course and were less frequent in number over time.

We analyzed the content of GOC discussions among patient categories (Figs. 1–5). In the full sample, decisions about code status predominated in the first (82.7%) and second (49.4%) GOC discussions, followed by an increasing focus on other treatment decisions in GOC 3–5 (44.1%–57.4%). A significant percentage of GOC discussions 2–5 did not require a treatment decision (18.7%–27.1%), but focused solely on patients' QOL definitions, goals, and values. Figure 2 illustrates that code status was quickly assessed (GOC discussion 1) upon admission and early in the hospital course for dementia patients in Category 1 (84.9%), who had the lowest rates of intubation. Other treatment decisions (including use of high-flow oxygen, artificial nutrition, and hospice care) became prominent in GOC discussions 3–5 (43.4%–66.7%) for Category 1 patients. Patients in Categories 2 and 3 (those with the highest case fatality rates and greatest use of hospice), similarly had earlier conversations about code status (86.4% and 90.0%, respectively) followed by increasing discussions about other treatment decisions in GOC discussions 2–4 (Figs. 3 and 4). Category 2 and 3 patients (and/or their decision makers) had more documentation of QOL definitions, goals, and values compared with patients with dementia (Category 1). Finally, in Category 4, the healthiest patients (i.e., with few comorbidities) admitted from home (all but one of them intubated after COVID infection) had significant numbers of conversations solely about goals and values (33.3%–57.2% in GOC 2–4, respectively) in addition to treatment-related discussions (16.6%–60.0% in GOC 2–5) (Fig. 5).

In terms of dot phrase/structured template use, fewer than half (44.4%) of all first GOC conversations incorporated this tool. Among our study population, the dot phrase was utilized in 69% of (62/90) GOC documentation. Palliative care providers utilized the dot phrase in 60.0% of their GOC conversations. Notably, there was increased use of the dot phrase by all clinicians during the COVID surge study period as compared with before (Fig. 6), based on EMR reports generated by the dot phrase's electronic bookmarking feature (Fig. 6 captures all documentation with dot phrase use at our health system, which includes patients not seen by palliative care).

Discussion

Broadly, we sought to gauge the impact of our efforts in education and systems change to improve documentation of GOC for seriously ill patients at our institution. Our palliative care division has widely promoted the importance of stewarding GOC conversations through EMR documentation, and we created a dot phrase/structured template as the “gold standard” for high-quality notes. This dot phrase became a core element of ongoing training programs for physicians, nurse practitioners, residents, fellows, and medical students in serious illness communication before the pandemic. In preparing for the surge in Massachusetts, the division chief (J.R.) helped define best practices as part of a system-wide COVID leadership council, including required documentation of patients' GOC within 48 hours of admission.⁹

In this retrospective study, almost all patients (seen by palliative care at some point during their hospitalization) had a documented GOC conversation within 48 hours of admission, meeting our institution's self-defined best practice. Ninety-eight percent of high-risk patients (Categories 2–3) had a documented GOC conversation within a median of 48 hours from admission, and 94% of patients with dementia (Category 1) within a median of 24 hours. Notably, palliative care specialists did not lead the majority of these conversations, as only roughly a quarter of all palliative care consults occurred within the first 48 hours after admission. Invariably, these discussions were multidisciplinary including social workers and nurses; however, we examined only physician and APP documentation as it was more accessible for manual chart review.

Regarding structured template use, less than half of initial GOC documentation incorporated the dot phrase, a tool we had promoted among all clinicians for comprehensive, concise, and retrievable EMR documentation. The remaining half of initial GOC documentation required keyword search and laborious chart review, which is neither feasible nor useful in fast-paced health care settings. Beyond our study population, there was increased use of the dot phrase among all providers caring for patients not seen by palliative care during the COVID surge, based on EMR-generated reports (Fig. 6). This was likely due to widespread promotion through educational forums, internal electronic newsletters, bedside coaching of providers, and word of mouth among staff, especially residents. However, the overall low rates of dot phrase use illustrate the need for ongoing marketing and better workflow strategies for adoption.

In terms of timing and pacing of GOC conversations, clinicians had an average of three to four conversations per patient during the hospital course, regardless of patient category and prior health status. We chose to measure up to five GOC discussions and acknowledge that some patients with longer lengths of stay had more. Among the entire cohort, the five GOC discussions were spaced two to three days apart during the hospital course and were less frequent in number over time. However, in Category 4 (the healthiest patients before COVID infection), decisions regarding tracheostomy, dialysis, and other treatments occurred around day 16 (as clinically expected), resulting in a quick series of GOC3, GOC4, and GOC5 with a median of one day in between conversations.

In terms of content, decisions about code status predominated in the first and second GOC discussions, followed by an increasing focus on other treatment decisions in GOC 3–5. A significant percentage of GOC 2–5 did not require a treatment decision but were based solely on patients' QOL definitions, goals, and values. The vast majority of these conversations were with family and surrogate decision makers, as most patients were incapacitated. Theoretically, these “no treatment decision” conversations may have helped providers build rapport with these surrogates and elicit values to tailor future treatment decisions as a patient's condition evolved; not surprisingly, re-exploration of code status and use of life support remained an ongoing topic among the five discussions. In Category 2 (patients with other chronic/serious illness admitted from facilities), there was only a small group of patients with GOC4 and GOC5 conversations, and these patients had long LOS and treatment decisions after the chart review period.

With these categories, we sought to understand and portray the arc of GOC conversations throughout a hospital course for a spectrum of patients. These “snapshots” convey an ongoing dynamic process and may dissuade the notion of GOC conversations as one-time events, which may have been inadvertently portrayed in our institution's guideline.

There were several limitations to our study, which was a retrospective chart review at a single academic health system. Our study population focused solely on COVID patients seen by palliative care (N = 90), as it was not feasible to manually review the charts of all COVID patients (N = 706). In our clinical outcomes, we may have underestimated the number of postdischarge deaths as we reviewed newspaper obituaries and not public death records. We had no baseline rates of timing, content nor quality of GOC documentation in our EMR for comparison. For non-dot phase documentation, we relied on keywords and manual chart review, which is subjective and prone to bias. Indeed, although there is national consensus on the definition of “advance care planning,”¹² there is no universal definition of the term “goals of care” nor the minimum quality standards for GOC documentation to guide research and quality improvement initiatives.¹³

We did not seek to define the minimum standard but assess current state of GOC documentation at our institution. Based on our findings, and our efforts to promote the dot phrase/template as a best practice, we advocate clinicians view this process as intensive “informed consent” tailored to patients' unique goals and medical realities. Therefore, GOC documentation should include the medical acuity/prognosis; patients' priorities and views on QOL; and treatment plans based on risks versus benefits in context of these goals.

We did not have the resources to analyze whether patient outcomes were goal concordant in this study. Our efforts to understand and further promote robust GOC documentation would make future analyses of goal-concordant care more feasible. For now, it is difficult to determine goal-concordant care if the goals are not well articulated in our documentation.

In summary, our study examines the numbers, timing, and content of all GOC documentation in the EMR to illustrate the journeys of COVID patients seen by palliative care at our institution during the initial surge in Massachusetts (Spring of 2020). Our division influenced the health system's official best practices, including required documentation of patients' goals within 48 hours of admission. Although present, timely, and identifiable by manual chart review, the majority of GOC documentation did not meet our standards of comprehensive, concise, and easily retrievable notes to shepherd patients' priorities through use of a structured template. However, the template, in combination with clinician training and extractable reporting on quality metrics, holds great promise for leveraging the EMR as a tool to align patients' goals and values with their medical care, both at our institution and across the country.

Program Structure

Our palliative care team is housed in UMass Memorial Medical Center (UMMMC), a 781-bed academic safety-net health system in central Massachusetts. The team is supported financially by UMMMC and through billing.

Team Staffing

Our team has equivalent to 9.6 total full time equivalent (FTE) as of 2020 (physician 3.7, advanced practice practitioners [APPs] 2.8, social worker 1.5, administrative assistant 1.0, music therapist 0.1, and pharmacist 0.5). All physicians and 2 APPs are board certified in palliative care.

Program Availability

Palliative care team provides inpatient consultative service at two hospitals and four days per week of ambulatory outpatient. We recently have started outreach to select nursing facilities but that was put on hold with pandemic. Our typical coverage is 8 am–5 pm on weekdays, but this was expanded to seven days per week during pandemic surge in March–April 2020. We receive referrals from all areas of the hospital based on patients, families, and provider needs. Our largest number of referrals come from hospital medicine followed by oncology and ICU.

Patient Volume and Interactions

Our penetration rate as of 2019 is 2.6% and we do two to three average visits per patient. Our average length of service or stay on the inpatient palliative care service is 7.9 days.

Funding Information

No funding was received for this article.

Author Disclosure Statement

No competing financial interests exist.

Supplementary Material

Supplementary Appendix SA1

Supplementary Appendix SA2

Supplementary Appendix SA3