Study design and setting

This prospective cohort study was conducted between March–August 2018 in one major academic tertiary care hospital (University of Alberta Hospital) and one community hospital (Grey Nuns Community Hospital) located in Edmonton, a city with a population of approximately 1 million people in Alberta, Canada. Both ED sites are staffed by full-time emergency physicians with Royal College of Physicians and Surgeons emergency medicine fellowship training or College of Family Physicians of Canada emergency medicine certification. While the University of Alberta Hospital includes a separate adult and pediatric ED, the study was only conducted in the adult ED. The Grey Nuns Community Hospital includes a single ED that provides care to pediatric and adult patients and has a 20-bed palliative care in-patient unit and palliative care coverage is provided by a city-wide service.

Study participants

Emergency physicians at both study sites were approached by the study team in-person or via email to provide written consent to participate in the study. Between March–August 2018, participating physicians were instructed to screen all ED presentations made by adult patients (17 years and older) during their clinical shifts to identify patients presenting with EOL conditions including cancer, chronic pulmonary disease (COPD), chronic kidney disease (CKD), heart failure (HF), cirrhosis, dementia, and progressive central nervous system (PCNS) disease (e.g., multiple sclerosis, Parkinson’s disease, amyotrophic lateral sclerosis). Clinical judgement was permitted to identify patients with EOL conditions; however, standardized clinical criteria for advanced or end-stage illness for each of the eligible clinical condition were provided to the physicians on the screening form (S1 and S2 Tables) [9, 10]. All presentations for EOL conditions identified by participating physicians during the study period were eligible for inclusion, with the exception of presentations as a result of a hand-over from another participating physician. The hand-over of patients commonly occurs at shift change in over-crowded Canadian urban EDs, and to avoid the risk of screening duplicate presentations, physicians were instructed to avoid screening any patients received in hand-over from another ED physician. Our research staff helped identifying potentially eligible patients using the Emergency Department Information System (EDIS); however, the presence or absence of an EOL condition was confirmed by the patient’s attending physician.

Screening process

For each presentation for an EOL condition that was identified, participating physicians completed a paper-based screening tool to identify the proportion of presentations made by patients with unmet palliative care needs (S1 Table) [9, 10]. The original screening tool, which had content and face validity [9], was reviewed and modified by a panel of clinical experts (in emergency medicine, intensive care, nephrology, palliative care, and internal medicine) as well as methodologists to gather additional information for the purposes of the current study. To obtain a better understanding of the frequency of presentations to the ED by patients with EOL conditions who could benefit from referral to palliative care services, as well as gain a better understanding of the ED management of patients with and without unmet palliative care needs (e.g., including treatments, disposition, and referrals), presentations for septic shock/severe sepsis or other conditions with a high chance of accelerated death (e.g., intracranial bleed not compatible with life, cardiac arrest, etc.) were removed from the list of eligible EOL conditions on the screening tool. In addition, the screening tool was modified so that physicians could specifically identify presentations for advanced dementia separately from other advanced PCNS conditions (e.g., multiple sclerosis, Parkinson’s disease, amyotrophic lateral sclerosis). The additional questions asked physicians to identify whether patients presented with formal GOC and their opinion as to the appropriateness of each patient’s current GOC based on their presenting condition. Finally, physicians were asked to identify any perceived challenges they faced during the screening process.

During the physicians shifts, physicians screened each patient they attended to for an eligible EOL condition as indicated in step one of the modified screening tool (advanced cancer, COPD, CKF, HF, cirrhosis, dementia, or other PCNS conditions) (S1 Table). Physicians were instructed to indicate one or more EOL condition in step one as applicable. Physicians then completed step two of the modified screening tool which included a list of pre-established risk factors for unmet palliative care needs [9] that included: two or more ED/hospital visits in the past 6 months, uncontrolled symptoms, functional decline, uncertainty about goals of care and/or caregiver distress, and a positive response to the surprise question (SQ) (physician would not be surprised if the patient died within 12 months) (S1 Table). Based on their assessment of the patient, physicians selected the relevant risk factors for each patient. As established with the original screening tool [9, 10], patients with two or more risk factors for palliative care needs were identified as having unmet palliative care needs, while patients with only one or no risk factors identified by physicians were classified as not having unmet palliative care needs (met palliative needs).

Data collection

Completed data collection forms were retrieved by research assistants. For each patient presentation with a completed data collection form, trained research assistants collected additional data from medical and EDIS records using standardized chart review methods: 1) patients demographics; 2) mode of arrival; 3) acuity at ED presentation [11]; 4) documented GOC for all ED presentations as well as GOC designations (GOC in the ED); 5) documented GOC upon hospitalization among presentations involving admission to hospital (GOC upon admission); 6) procedures (e.g., invasive or non-invasive ventilation, paracentesis, thoracentesis, intubation), investigations (e.g., laboratory and diagnostic tests including blood work, urinalysis, microbiology, imaging, electrocardiogram) and any medications delivered to the patient in the ED (e.g., fluids/electrolytes, oxygen, narcotics, analgesics, antibiotics, corticosteroids) regardless of the route of administration; 7) consultations requested in ED and post hospital referrals, after inpatient admission; 8) ED times (e.g., length of stay [LOS: triage to discharge time] and physician initial assessment [PIA: triage to doctor assessment time]); 9) patient disposition and ED revisits within 30 days of discharge from the ED, or after hospital discharge when the patient was admitted; 10) ED visits within 6 months prior to their enrollment ED visit. To estimate patient acuity at ED presentation, both participating ED sites utilize the five-level Canadian Triage and Acuity Scale (CTAS) system which is applied by experienced triage nurses at presentation and represents acuity as the following: CTAS 1 –resuscitation; CTAS 2 –emergent; CTAS 3 –urgent; CTAS 4 –less urgent; and CTAS 5 –non-urgent [11]. Formal GOC for each ED presentation and hospitalization was documented in the patients charts by the physician or nurse or in a document referred to as a “green sleeve” which documents the patients formal GOC. In Alberta, the various GOC designation orders are documented based on different levels of medical care including R (resuscitative care), M (full medical care with transfer to acute care), and C (comfort care without a goal to prolong life) with patients being able to specify different levels of care within each level. For example, patients designated as an R GOC could specify R1 (patient accepts any appropriate investigations/interventions, resuscitation and admission to the intensive care unit (ICU)), R2 (patient accepts any appropriate investigations/interventions, resuscitation and ICU care excluding chest compressions), or R3 (patient accepts any appropriate investigations/interventions, resuscitation and ICU care excluding intubation and chest compressions). Additional details on all of the various GOC designations are provided in S3 Table.

Data extraction for the first 10 medical records were completed in duplicate and reviewed by an experienced clinical research nurse to identify and mediate disagreements and ensure a unified data collection methodology. Basic information on the emergency physicians’ socio-demographic characteristics, training, years of experience and participation rate were self-reported using paper-based survey methods. All study data were entered into REDCap, a secure web platform for managing online surveys and databases (Vanderbilt University, Knoxville, TN, USA).

Statistical analysis

All analyses were performed using StataCorp, 2015 (Stata Statistical Software: Release 15. College Station, TX). Descriptive data are reported using proportions for dichotomous variables, means with standard deviations (SD), or medians with interquartile range (IQR) for continuous variables, as appropriate. Bivariate comparisons of dichotomous and continuous variables between patients with unmet and met palliative care needs were completed using chi-squared (χ²) test and using T-test or Mann-Whitney U test, respectively. Physician responses to the screening tool question regarding their recommendations for appropriate GOC and screening challenges were coded based on themes and frequencies. Due to the multiple tests performed, a statistical significance level was set at p < 0.01.

A logistic regression was used to detect trends in relationships between the variables and having unmet palliative care needs and to estimate unadjusted odds ratio (OR) of all variables. Variables with a p < 0.2 were then incorporated in a multivariable logistic regression model and removed from the final model at p > 0.05. The variables in the screening tool (e.g., frequent ED visits/hospitalizations, uncontrolled symptoms, functional decline, uncertainty of GOC/caregiver distress, and SQ) were not included in the adjusted model as these factors were used to directly classify patients as having unmet palliative care needs. Unadjusted and adjusted odds ratios (aOR) with 95% confidence intervals (CIs) for each variable were reported.

Ethics

The Health Research Ethics Boards of the University of Alberta and Covenant Health approved the study protocol, study materials, and was granted access to medical records for the conduct of chart reviews without the need for written informed consent from the individual patients (Reference ID: Pro00078882). The Alberta Research Ethics Community Consensus Initiative (ARECCI) screening tool identified this study as having minimal risk. Emergency physicians provided written informed consent to participate in the study and were provided a $50 gift card for participating in the study and reaching a minimum recruitment target of 20 patients.

Physician recruitment

The majority of emergency physicians (n = 45/60; 75%) across the two study sites agreed to participate in the study. Most were male (n = 32/45; 71%) and the average age of was 41 years (SD: 9.3). The majority of physicians had a Fellowship from the Royal College of Physicians and Surgeons of Canada (n = 21/45; 47%) or a Certification from the College of Family Physicians of Canada (n = 21/45; 47%). A slight majority (n = 24/45; 53%) of participating physicians had over 10 years of experience in emergency medicine practice. The median number of patients enrolled by the participating physicians was 19 (IQR: 8, 20).

Patient enrollment and screening

Over the course of the study period, participating physicians attended a total of 26,328 adult presentations, of which, physicians identified 663 presentations for EOL conditions (see Fig 1). The 663 presentations for EOL conditions consisted of 627 unique patients; the majority of these patients made a single presentation during the study period (95%; n = 594/627). Thirty-three patients presented to the ED multiple times during the study period, with the majority of these patients presenting to the ED twice (91%, n = 30/33). For patients with multiple presentations to the ED, the median number of days between the first and second visits was 30.5 (IQR: 8.5, 46.5). The research staff identified 58 presentations with compatible diagnoses in which the participating physicians did not complete the screening tool (see Fig 1). Across the 663 presentations for EOL needs, 78% of the presentations (n = 518/663) included patients who met the criteria for having unmet palliative care needs (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Identification of 663 presentations to the emergency department by patients with end-of-life conditions during the study period.

https://doi.org/10.1371/journal.pone.0257501.g001

Patient demographics

Of the 627 unique patients, 326 (52%) were female and the median age was 76 years (IQR: 63, 85). There were no statistically significant differences between the presentations among patients with met vs. unmet palliative care needs in terms of CTAS scores, mode of arrival by EMS, or presenting to the ED with documented GOC (as documented in the patient medical records) (see Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics and goals of care documentation in a study of 663 presentations to the emergency department by patients with end-of-life conditions, sub-grouped by met and unmet palliative care needs.

https://doi.org/10.1371/journal.pone.0257501.t001

Unmet vs Met palliative care needs

Table 2 summarizes the information collected through the modified screening tool (overall and according to unmet vs. met palliative care needs) for each presentation with EOL conditions. Briefly, a positive response to the SQ was the most common risk factor identified (80% unmet vs. 25% met), followed by uncontrolled symptoms (64% unmet vs. 22% met), and frequent ED visits or hospitalizations (63% unmet vs. 19% met). The majority of presentations for patients with unmet palliative care needs had two (46%) or three (32%) risk factors documented for unmet palliative care needs. Advanced cancer was the most prevalent condition in both groups (42% unmet vs. 37% met). Patients with unmet palliative care needs were significantly more likely to have presented with two or more EOL conditions (15% unmet vs. 6% met, p = 0.007).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Factors included in a modified screening tool to identify met or unmet palliative care needs among 663 presentations to the emergency department by patients with end-of-life conditions.

https://doi.org/10.1371/journal.pone.0257501.t002

Among all presentations, physicians reported that a similar proportion of patients presented with established GOC across both groups (62% unmet vs. 63% met). There was no significant difference in the proportion of presentations in which a physician felt the GOCs were inappropriate considering the patients current condition (66% unmet vs. 59% met). In cases where the physicians were of the opinion that patients current GOC was inappropriate, physicians were significantly more likely to recommend C designations (comfort care) that focused strictly on providing comfort for patients presenting with EOL conditions (17% unmet vs. 5% met, p = 0.004) (see Table 2). Among the 33 patients who made multiple ED presentations during the study period, 45% (n = 15/33) of patients didn’t have active GOC documented in their charts during the initial visit but did have established GOC in their second visit. Finally, the proportion of presentations in which the physician identified significant challenges during the screening process was similar between the groups (24% unmet vs. 24% met). The presence of a language barrier was the most common challenge during the screening process identified by physicians for patients with unmet palliative care needs (34%), while dementia as the most common challenge for patients with met palliative care needs (40%).

ED management and outcomes

Table 3 summarizes the differences in ED management across the presentations by patients with unmet vs. met palliative care needs. A similar proportion of presentations for EOL conditions with unmet or met palliative care needs received medications, procedures (e.g., paracentesis, thoracentesis, CPAP, intubation) and investigations (e.g., imaging, blood work, urine or microbiology) within the ED (see Table 3). Presentations for patients identified as having unmet palliative care needs were more likely to have an ED-based consultation requested by the attending physician (80% unmet vs. 67% met; p = 0.001); however, the median number of consultations between the groups was similar. There was no difference in the median time from triage to PIA time or the median overall ED LOS between the groups (see Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Emergency department management of 663 presentations by patients with end-of-life conditions compared on unmet vs. met palliative care needs.

https://doi.org/10.1371/journal.pone.0257501.t003

Outcomes regarding differences in patient disposition and post-discharge referral are provided in Table 4. A significantly higher proportion of presentations to the ED by patients with unmet palliative care needs resulted in admission to hospital (69% unmet vs. 51% met; p<0.001). In addition, a significantly higher proportion of patients who died either in the ED or in-hospital during their index visit had unmet palliative care needs (18% unmet vs. 7% met; p = 0.002). A similar proportion of presentations ended with referrals post-discharge between the groups; there were no differences in the targeted specialty.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Disposition and post-discharge referrals of 663 emergency department presentations by patients with end-of-life conditions compared on unmet vs. met palliative care needs.

https://doi.org/10.1371/journal.pone.0257501.t004

No differences in the proportion of patients with return visits to the ED within 30 days or the median number of return visits were found between the groups (see Table 4). Interestingly, patients with unmet palliative care needs were significantly more likely to have visited the ED in the six months prior to the index visit (73% unmet vs. 48% met; p<0.001) (see Table 4).

Adjusted analyses

Variables considered for the logistic regression analysis included: EOL conditions, hospitalization, consultation requested, received medications, mortality, inappropriate GOC, and main diagnosis of cirrhosis. Regardless of their statistical significance in the univariate analysis, some variables (e.g., age, sex) were considered clinically relevant and therefore retained in the final model to correct for possible unidentified age and sex-related confounding (see Table 5). After adjusting for age and sex, having two or more EOL conditions (aOR = 2.41; 95% CI: 1.16, 5.00), requiring hospitalization (aOR = 1.93; 95% CI: 1.30, 2.87), and dying during the index visit (aOR = 2.15; 95% CI: 1.02, 4.53) were strongly associated with having unmet palliative care needs (see Table 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Factors associated with having unmet palliative care needs among 663 presentations to the emergency department by patients with end-of-life conditions.

https://doi.org/10.1371/journal.pone.0257501.t005

Limitations

There are several important limitations with this study that must be considered. First, multiple efforts were made to reduce the risk of missing patients, including enrolling the majority of available emergency physicians (75%) and providing frequent reminders to screen to participating physicians, however, it appears that it is likely that some patients with EOL conditions who presented to the ED during the study period were missed. Our research staff did identify 58 patients who were treated by participating physicians during the study period who were not screened but had compatible ED diagnoses and there was documentation on EDIS or their medical records of end-stage or palliative condition or active GOC focused on control and comfort care. Second, the clinical and administrative outcome data (e.g., ED times, disposition, return visits) of this study were dependent on the information available in the electronic records and paper charts, and unfortunately, some information such as patient GOC, were inconsistently documented. Third, this study did not employ an independent third party to verify the accuracy of the patients EOL status. While it is possible that some patients may have been misclassified as having an EOL condition, the participating physicians had years of experience in emergency medicine, and as such the risk of differential misclassification of patients is unlikely. Fourth, it is possible that the modifications made to the palliative screening tool could have impacted its diagnostic properties. At this point, however, the diagnostic properties of the original tool, along with other available screening tools to identify ED patients with unmet palliative care needs is not well established [4]. The content and face validity of the original tool, however, has been established by a team of palliative care experts using modified Delphi techniques. While there is some evidence that the SQ alone has moderate sensitivity and specificity to correctly identify patients who will die [6, 17, 20], its ability to accurately identify patients with unmet palliative care needs has not been established. While further validation of the screening tool to identify ED patients with unmet palliative care needs is needed, the risk factors utilized in the current and modified version of the screening tool are likely sufficient to identify patients who could benefit from palliative care services. Given the lack of clarity regarding the accuracy of any one of the available screening tool for identifying ED patients with unmet palliative care needs, further studies are needed to assess and compare the diagnostic properties of the available tools employed for screening patients with unmet palliative care needs in the ED. Lastly, differences in healthcare systems in Canada compared to other countries in the world could limit the generalizability of the results of this study.

Conclusions

Notwithstanding the above-mentioned limitations, this study found that a large proportion of patients presenting to the ED with EOL conditions have unmet palliative care needs. Overall, while no differences with respect to medications, procedures, investigations, or diagnostic imaging, patients with unmet palliative care needs were significantly more likely to undergo consultations in the ED, require hospitalizations, and die during the index visit. Based on the results of the adjusted analysis, having multiple EOL or palliative conditions or requiring hospitalization were strongly associated with having unmet palliative care needs, which could be used by emergency physicians to identify patients who may benefit from referral to palliative care services. Despite the availability of palliative care services, referrals to a palliative care consultant in the ED or following discharge was low regardless of the status of their palliative care needs (met or unmet). These findings suggest that steps should be taken to ensure that ED patients who could benefit from palliative services are identified and provided these services which can help improve the EOL trajectory.