Analysis of Diagnoses, Symptoms, Medications, and Admissions Among Patients With Cancer Presenting to Emergency Departments

Key Points español 中文 (chinese)

Question  What are the characteristics of patients with active cancer presenting to US emergency departments?

Findings  In this multicenter cohort study of 1075 adult patients with active cancer in the Comprehensive Oncologic Emergencies Research Network (CONCERN), patients commonly presented with symptoms such as pain (62.1%) and nausea (31.3%), were frequently treated for potential infection (26.5%), and were admitted (57.2%; 25.0% for <2 days) or placed in observation (7.6%).

Meaning  Opportunities for improving emergency department care for patients with cancer include establishing protocols and processes for prompt and appropriate symptom control, creating improved risk stratification tools, and improving outpatient management to prevent ED visits.

Importance  Better understanding of the emergency care needs of patients with cancer will inform outpatient and emergency department (ED) management.

Objective  To provide a benchmark description of patients who present to the ED with active cancer.

Design, Setting, and Participants  This multicenter prospective cohort study included 18 EDs affiliated with the Comprehensive Oncologic Emergencies Research Network (CONCERN). Of 1564 eligible patients, 1075 adults with active cancer were included from February 1, 2016, through January 30, 2017. Data were analyzed from February 1 through August 1, 2018.

Main Outcomes and Measures  The proportion of patients reporting symptoms (eg, pain, nausea) before and during the ED visit, ED and outpatient medications, most common diagnoses, and suspected infection as indicated by ED antibiotic administration. The proportions observed, admitted, and with a hospital length of stay (LOS) of no more than 2 days were identified.

Results  Of 1075 participants, mean (SD) age was 62 (14) years, and 51.8% were female. Seven hundred ninety-four participants (73.9%; 95% CI, 71.1%-76.5%) had undergone cancer treatment in the preceding 30 days; 674 (62.7%; 95% CI, 59.7%-65.6%) had advanced or metastatic cancer; and 505 (47.0%; 95% CI, 43.9%-50.0%) were 65 years or older. The 5 most common ED diagnoses were symptom related. Of all participants, 82 (7.6%; 95% CI, 6.1%-9.4%) were placed in observation and 615 (57.2%; 95% CI, 54.2%-60.2%) were admitted; 154 of 615 admissions (25.0%; 95% CI, 21.7%-28.7%) had an LOS of 2 days or less (median, 3 days; interquartile range, 2-6 days). Pain during the ED visit was present in 668 patients (62.1%; 95% CI, 59.2%-65.0%; mean [SD] pain score, 6.4 [2.6] of 10.0) and in 776 (72.2%) during the prior week. Opioids were administered in the ED to 228 of 386 patients (59.1%; 95% CI, 18.8%-23.8%) with moderate to severe ED pain. Outpatient opioids were prescribed to 368 patients (47.4%; 95% CI, 3.14%-37.2%) of those with pre-ED pain, including 244 of 428 (57.0%; 95% CI, 52.2%-61.8%) who reported quite a bit or very much pain. Nausea in the ED was present in 336 (31.3%; 95% CI, 28.5%-34.1%); of these, 160 (47.6%; 95% CI, 12.8%-17.1%) received antiemetics in the ED. Antibiotics were administered in the ED to 285 patients (26.5%; 95% CI, 23.9%-29.2%). Of these, 209 patients (73.3%; 95% CI, 17.1%-21.9%) were admitted compared with 427 of 790 (54.1%; 95% CI, 50.5%-57.6%) not receiving antibiotics.

Conclusions and Relevance  This initial prospective, multicenter study profiling patients with cancer who were treated in the ED identifies common characteristics in this patient population and suggests opportunities to optimize care before, during, and after the ED visit. Improvement requires collaboration between specialists and emergency physicians optimizing ED use, improving symptom control, avoiding unnecessary hospitalizations, and appropriately stratifying risk to ensure safe ED treatment and disposition of patients with cancer.

Patients with cancer account for more than 4.5 million US emergency department (ED) visits annually.^1,2 This estimate, derived from ED diagnostic codes, likely underestimates the true annual incidence of cancer-related ED visits among the 15.5 million US residents with cancer.³ Approximately two-thirds of ED visits by patients with cancer result in hospital admission,^4,5 affirming the ED as a critical portal addressing acute illness issues and the continuum of cancer management.

Oncologic emergency medicine has traditionally focused on the diagnosis and management of the adverse effects of cancer treatment (eg, febrile neutropenia, thrombosis, tumor lysis syndrome, or hypercalcemia) and patient-centered care issues (eg, symptom control and quality of life).⁶ Advances in oncology treatments present challenges of new adverse effects and an expanding spectrum of cancer-treatment-related emergencies.⁷ Symptom-driven issues are the most common complaints in patients with cancer visiting EDs, including pain and respiratory and gastrointestinal tract issues.^2,4 However, a knowledge gap exists for providing optimal cancer care in the ED.⁸

Retrospective studies have described characteristics of patients with cancer who present to the ED.^2,4,9-12 Currently available data, gleaned from retrospective registries or surveys, lack granularity regarding ED course, functional status, symptom inventories, and other variables critical to understanding ED visits by patients with cancer.⁴ To address this knowledge deficiency, the National Cancer Institute sponsors the Comprehensive Oncologic Emergencies Research Network (CONCERN), a multicenter research consortium including representatives from oncology and emergency medicine.^13,14 The National Cancer Institute has expressed interest in clarifying cancer-related use of emergency care, risk stratification, and outcome improvement.¹⁵ CONCERN aspires to accelerate knowledge generation as well as design, execution, and translation of oncologic emergency medicine research. CONCERN’s focus areas include but are not limited to epidemiology, febrile neutropenia and other infections, acute symptom control, venous thromboembolism, and palliative care.^8,14

This study addresses these essential high-priority goals and interests. The primary objective of this study is to establish a benchmark description of the population of patients with active cancer presenting to CONCERN EDs.

We conducted a multicenter prospective observational cohort study in 18 CONCERN-affiliated EDs from March 1, 2016, through January 30, 2017, with institutional review board approval at each site. All participants provided written informed consent, and no incentive was provided. This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.¹⁶ The study sites were academic EDs with a mean annual volume of 71 886 patients and a mean admission rate of 30%. The median proportion of ED visits of patients with active cancer was 7.5% (interquartile range [IQR], 4.0%-15.0%). Seventeen sites were urban, and 13 were affiliated with National Cancer Institute–designated comprehensive cancer centers (a complete list appears in eTable 1 in the Supplement).

Study participants consisted of a nonconsecutive sample of adults (aged ≥18 years) with active cancer who presented in the ED when research staff were available. Availability varied by site, but generally consisted of weekdays 7 am until 8 to 11 pm with some weekend and overnight (11 pm-7 am) availability. Active cancer was defined by (1) ongoing (or within 12 months) antineoplastic therapy (radiotherapy, chemotherapy, cancer-related surgery, or other), (2) previously identified or ED physician–identified cancer recurrence or metastasis, or (3) cancer-related symptoms.¹⁷ Exclusion criteria included pregnancy, incarceration, psychiatric chief complaint, primary evaluation as a trauma response, non–English speaking, previous enrollment, or too ill or otherwise unable to participate in survey administration.

Research staff administered a patient questionnaire in the ED. Although the patient was the primary responder, family and caregivers could assist with survey completion. After a minimum of 1 hour of training, study staff collected data from medical record review 30 days after enrollment using standardized electronic forms and a data dictionary.¹⁸ Records reviewed included the ED and inpatient records and subsequent outpatient notes during the 30 days.

Survey variables included demographics, cancer type, cancer therapy within the preceding 30 days (chemotherapy, targeted therapy, systemic corticosteroids, radiotherapy, or cancer-related surgery), and outpatient medications by pharmacologic class (eg, pain medication classes included nonsteroidal anti-inflammatory drugs, acetaminophen, tramadol hydrochloride, short- or long-acting opioids, and/or other). We defined advanced cancer based on the protocol of Kandarian et al¹⁹ as metastatic, recurrent, unresectable, locally advanced, stage III, or stage IV disease. Advanced directives were reported as not present, do not resuscitate, do not intubate, comfort care only, or other. We recorded receipt of palliative care or hospice services within the last 30 days. We collected data for Eastern Cooperative Oncology Group performance status²⁰ and the Functional Assessment of Cancer Therapy–General.²¹ Participants reported their highest measured temperature in the prior 24 hours and symptoms in the ED, including any pain, chest pain, shortness of breath, nausea, abdominal pain, or urinary tract symptoms.

Medical record review data included emergency severity index²²; ED vital signs; comorbidities sufficient to calculate the Charlson comorbidity index²³; primary cancer type; cancer grade and stage; recent cancer-related therapies; outpatient- and ED-administered pain medications, nausea medications, and antibiotics; ED diagnoses and disposition; inpatient diagnoses; additional ED visits and/or hospitalizations within 30 days; 30-day mortality; ED laboratory and imaging results; hospital length of stay (LOS); and presence of advance directives. We recorded patients’ initial ED pain scores on a 10-point scale categorized as none (0), mild (1-4), moderate (5-6), or severe (7-10).²⁴ We recorded as many as 4 ED diagnoses using the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision.

We anticipated enrolling as many as 70 participants per site (ie, a maximum of 1260 participants). To allow for potential underrecruitment, we calculated that a minimum of 840 patients (two-thirds of enrollment target) would provide sufficiently precise estimates such that proportions would have exact 95% CIs with a 2-sided width of less than 5%, which would allow precise estimates to clarify salient characteristics of the population presenting to an ED with active cancer.

Data were analyzed from February 1 through August 1, 2018, using SAS software (version 9.4; SAS Institute, Inc) and included descriptive statistics. We calculated means with SDs, medians with IQR, and proportions with 95% CIs. Normality was tested using the Shapiro Wilk test with a 2-sided P < .05 considered significant. We measured interrater reliability of the medical record review using the Cohen κ statistic through review of a random sample of 10% of medical records at each site by a second abstractor.

For enrollment, 1562 of 2337 screened patients were eligible, and 1075 of these (68.8%; 95% CI, 66%-71%) consented to participate (eFigure in the Supplement). Nonenrollees did not differ from enrollees by sex, age, or day of week screened. Study population characteristics are summarized in Table 1. Mean (SD) age was 62 (14) years, with 505 (47.0%; 95% CI, 43.9%-50.0%) 65 years or older and 99 (9.2%; 95% CI, 7.5%-11.1%) 80 years or older (557 female [51.8%; 95% CI, 48.8%-54.8%] and 518 male [48.2%; 95% CI, 45.2%-51.2%]). Four hundred thirty patients (40.0%; 95% CI, 37.0%-43.0%) had an emergency severity index of 2; 542 (50.4%; 95% CI, 47.4%-53.4%) had an emergency severity index of 3. Mean (SD) Charlson comorbidity score was 4.2 (3.1). The 5 top ED diagnoses were symptom related, including abdominal pain (100 [9.3%; 95% CI, 7.6%-11.2%]), fever (87 [8.1%; 95% CI, 6.5%-9.9%]), breathing abnormalities (77 [7.2%; 95% CI, 5.7%-8.9%]), nausea and vomiting (60 [5.6%; 95% CI, 4.3%-7.1%]), and throat and chest pain (51 [4.7%; 95% CI, 3.6%-6.2%]) (Table 2). Other symptom- and pain-related diagnoses were common.

Emergency department disposition included admission for 615 participants (57.2%; 95% CI, 54.2%-60.2%) and observation status for 82 (7.6%; 95% CI, 6.1%-9.4%) (Table 1). Of those admitted, mean (SD) LOS was 5.0 (5.6) days and median LOS was 3 days (IQR, 2-6 days); 154 of 615 admissions (25.0%; 95% CI, 21.7%-28.7%) had an LOS of 2 days or less, including 91 participants (14.8%; 95% CI, 12.1%-17.8%) with an LOS of 0 to 1 day and 63 (10.0%; 95% CI, 8.0%-12.9%) with an LOS of 2 days. Sixty-two participants (5.8%) died within 30 days of first presentation to the ED. Thirty-day mortality included 62 participants (5.8%; 95% CI, 4.4%-7.3%). Within 30 days of the enrollment visit, 286 participants (26.6%; 95% CI, 24.0%-29.45) had an ED revisit and 274 (25.5%; 95% CI, 22.9%-28.2%) had a separate hospital admission, including a second hospital admission in 180 of the 635 admitted during the initial enrollment visit. Forty-four participants (4.1%; 95% CI, 3.0%-49.7%) had more than 1 readmission.

The most common cancer types were gastrointestinal tract (220 [20.5%; 95% CI, 18.1%-23.0%]), lung (139 [12.9%; 95% CI, 11.0%-15.1%]), hematologic (128 [11.9%; 95% CI, 10.0%-14.0%]), and breast (118 [11.0%; 95% CI, 11.0%-15.1%]) (Table 3). Most participants had received treatment for cancer in the last 30 days (794 [73.9%; 71.1%-76.4%]). Although advanced or metastatic cancer was present in 674 (62.7%; 95% CI, 59.7%-65.6%), advance directives were present in 502 (46.7%; 95% CI, 43.7%-49.7%) including 193 (18.0%; 95% CI, 15.7%-20.4%) with a full-code directive. Twenty participants (1.9%; 95% CI, 1.1%-2.8%) were receiving hospice care and 86 (8.0%; 95% CI, 6.4%-9.8%) were receiving palliative care.

In the ED, pain was present in 668 patients (62.1%; 95% CI, 59.2%-65.0%), with shortness of breath (370 [34.4%; 95% CI, 31.6%-37.3%]), nausea (336 [31.3%; 95% CI, 28.5%-34.1%]), and abdominal pain (343 [31.9%; 95% CI, 29.1%-34.8%]) the next most common symptoms (Table 4); 249 patients (23.2%; 95% CI, 20.7%-25.8%) had no pain assessment documented in the ED medical record. Mean (SD) pain score among those with pain was 6.4 (2.6). More than one-third (386 [35.9%; 95% CI, 33.0%-38.8%]) reported moderate or severe pain in the ED. Opioids were administered to 381 participants (35.4%; 95% CI, 32.6%-38.4%), including 55 of 156 (35.2%; 95% CI, 27.8%-43.3%) with mild, 53 of 117 (45.3%; 95% CI, 36.1%-54.8%) with moderate, and 195 of 269 (72.5%; 95% CI, 66.7%-77.7%) with severe pain. Of the 386 participants with moderate or severe pain, 228 (59.1%; 95% CI, 18.8%-23.8%) received an opioid in the ED. Antiemetics were administered in 260 (24.2%; 95% CI, 21.3%-26.9%). Of those with nausea, 160 (47.6%; 95% CI, 12.8%-17.1%) received antiemetics in the ED.

In the week before the ED visit, 302 participants (28.1%; 95% CI, 25.4%-30.9%) had no performance deficit based on Eastern Cooperative Oncology Group score (Table 1). For the Functional Assessment of Cancer Therapy–General reporting prior week symptoms, 776 participants (72.2%; 95% CI, 69.4%-74.8%) reported pain and 368 (47.4%; 95% CI, 31.4%-37.2%) of these had outpatient opioid prescriptions. This included 428 (39.8%; 95% CI, 36.9%-42.8%) participants with quite a bit or very much pain, 244 (57.0%; 95% CI, 52.2%-61.8%) of whom had outpatient opioid prescriptions. Half of all participants (541 [50.3%; 95% CI, 47.3%-53.4%]) reported nausea, with 359 of these having home nausea medications (66.4%; 95% CI, 62.2%-70.3%). Also, 214 participants (19.9%; 95% CI, 17.6%-22.4%) reported quite a bit or very much nausea, with 161 (75.2%; 95% CI, 68.9%-80.9%) of these having taken nausea medications at home. Similar rates of quite a bit or very much symptoms were due to lack of energy (584 [54.3%; 95% CI, 51.3%-57.3%]), worry that the condition will worsen (360 [33.5%; 95% CI, 30.1%-36.4%]), poor sleep (408 [38.0%; 95% CI, 35.0%-40.9%]), decreased enjoyment of life (469 [43.6%; 95% CI, 40.6%-46.6%]), and decreased quality of life (409 [38.0%; 95% CI, 35.1%-41.0%]).

Table 5 shows suspected infection. Fever immediately before presentation or in the ED was present in 155 (14.4%; 95% CI, 12.4%-16.7%) and neutropenia in 26 (2.4%). Antibiotics were administered to 285 participants (26.5%; 95% CI, 23.9%-29.2%). Blood and urine cultures were obtained in 255 patients (23.7%; 95% CI, 21.2%-26.4%), and blood cultures yielded positive findings in 27 (2.5%; 95% CI, 1.7%-3.6%) and urine cultures in 54 (5.0%; 95% CI, 3.8%-6.5%). Participants with suspected infection as identified by ED antibiotic administration were admitted at higher rates than those not receiving antibiotics (209 of 285 [73.3%; 95% CI, 17.1%-21.9%] vs 427 of 790 [54.1%; 95% CI, 50.5%-57.6%]), with 19.1% (40 of 209) having an LOS of no more than 2 days. Although some antibiotic recipients probably did not have acute bacterial infections, the paradigm of timely antibiotic administration in at-risk populations without a priori evidence of acute bacterial infection accounts for the antibiotic use observed in this population where 465 participants (43.3%; 95% CI, 40.3%-46.3%) recently received chemotherapy.

Missing values are noted in Table 1 and Table 3. Interrater reliability was measured in 115 medical records (eTable 2 in the Supplement). Pain scores and worst vital signs had κ ≥ 0.80. Presence of advanced cancer, recent cancer therapy, and comorbidities had κ ≥ 0.60 (with most κ ≥ 0.80), except stroke (κ = 0.56) and recent corticosteroid administration (κ = 0.50).

The growing ED population of patients with active cancer^2,8 present while receiving cancer therapy and with high acuity, high symptom burden, and frequent need for admission. This analysis provides the epidemiologic foundation to pursue care improvement opportunities, notably risk stratification, symptom management, and disposition in this population. These opportunities extend beyond the ED to include a need for timely specialty consultation, adequate outpatient follow-up to facilitate ED discharges, consistently applied processes targeting commonly encountered issues, and improved outpatient management to prevent ED visits.

We have identified a need for improved symptom control before and during the ED visit. Pain is a driver of ED presentation among patients with cancer,^4,12 and appropriately managing pain is a significant contributor to improving quality of life in these patients.^25,26 Pain in the ED was present in 62.1% of study participants, whereas 249 (23.2%; 95% CI, 20.7%-25.8%) had no pain assessment documented in the ED medical record. Nevertheless, a significant percentage of participants with moderate or severe pain did not receive opioids in the ED. This finding may reflect unexpected efficacy of nonnarcotic pain medications or an undesirable effect of current efforts to decrease overall ED opioid administration. Moreover, a significant portion of study patients were receiving outpatient opioid regimens, likely necessitating larger doses of opioids to effectively treat acute pain because opioid-tolerant ED patients with cancer frequently receive inadequate initial opioid doses.²⁷ The pervasiveness of pain among patients presenting to the ED with active cancer underscores the need to recognize potential opioid tolerance and to explore less conventional ED pain control modalities (eg, subdissociative-dose ketamine and intravenous lidocaine).²⁸

Poorly controlled pain was also frequently present in the week before the ED visit, with almost half of those reporting quite a bit or very much pain not receiving outpatient opioids. A significant opportunity therefore exists to improve pain control in the ED and outpatient settings, which could improve quality of life and avoid ED visits and hospital admissions. Improvements could include pragmatically designed clinical pathways that are primarily developed by emergency medicine physicians, in collaboration with oncology and palliative care colleagues, to improve symptom control and ensure adequate administration of analgesia. Given the low proportion of patients receiving palliative care services at time of ED visit (8.0%), the opportunity for additional intervention in these patients before prior to the ED visit is substantial.

Dyspnea and nausea were frequently present. Dyspnea is a particularly worrisome and challenging symptom because it may represent many causes of life-threatening pathologic conditions (eg, acute coronary syndrome, pulmonary embolism, pneumonia, cardiac tamponade, pleural effusion, pneumonitis) that, after disease progression, are leading causes of death in patients with cancer.²⁹ The frequent presence of dyspnea as well as the broad differential diagnosis of life-threatening conditions likely contributes to high testing and admission rates. For nausea, ED and outpatient treatment appear to be suboptimal. This finding provides additional opportunities to improve quality of life and outpatient and ED management and develop care plan processes facilitating outpatient rather than inpatient care.

Suspicion of bacterial infection was also common. Although only 14.4% of participants had a recent or an ED-measured fever, blood and urine cultures were obtained in 255 (23.7%; 95% CI, 21.2%-26.4%), and antibiotics were administered in 26.5%. Participants receiving antibiotics were admitted at a higher rate than those not receiving antibiotics. Although some antibiotic recipients probably did not have acute bacterial infections, the paradigm of timely antibiotic administration in at-risk populations without a priori evidence of acute bacterial infection accounts for the antibiotic use observed in this population where 465 participants (43.3%; 95% CI, 40.3%-46.3%) recently received chemotherapy. Although only a small percentage (2.4%) of our cohort were found to have febrile neutropenia (a rate similar to that in prior ED studies),¹² concern for this condition and the urgency to minimize time to antibiotics for these participants may have increased empirical antibiotic administration rates in febrile individuals who were ultimately found not to have neutropenia.

The large proportion of participants with suspected infection and subsequent admission identifies a population that could benefit from improved processes of care to ensure appropriate resource use and antibiotic administration in patients with and without neutropenia who have cancer and suspected infection. Risk stratification tools for infection focus on febrile neutropenia with conflicting results when applied in the ED setting.^30-33 However, these tools are rarely used by oncologists or emergency physicians.^34,35 Moreover, most of the participants with fever and/or suspected infection did not have febrile neutropenia. Little evidence regarding evaluation and risk stratification of these patients is available, and this area requires further investigation.

Use of palliative services is essential to outpatient care of patients with active cancer but was rare in our cohort, with only 8.0% receiving palliative care. Fewer than half reported having an advance directive. These low proportions, juxtaposed with the illness severity and substantive symptom burden of this ED population in the ED and in the week before the ED visit, justify exploring the ED’s potential for facilitating linkage to palliative care. Such linkage could, in turn, decrease subsequent ED use and improve quality of life.^36-39 In addition, emergency physicians may communicate the need for advanced directives to patients with cancer and their outpatient physicians. Opportunities may exist for hospice referral, given the 5.8% 30-day mortality of the cohort, with only 1.9% receiving hospice care at the time of ED visit.⁴⁰

Among our cohort, the admission rate was 57.2%, similar rates in prior research in patients with cancer,^2,4,5 and much higher than the overall 30% rate in the study EDs and the national 9.0% rate.¹ Given the risks of cancer-related complications along with the high symptom and comorbidity burdens of the ED population with active cancer, this observation is not surprising. Population-based research has identified disposition of patients with cancer who present to the ED to be associated with cancer type, comorbidities, age, race, and insurance type.⁴¹ Still, for those patients admitted purely for symptom control, for unavailability of specialty consultation, or for minimal interventions (eg, intravenous fluids or waiting for culture results only), admission may be overused and constitute a risk (hospital-acquired infection, deep venous thrombosis, fall) that outweighs the benefit of inpatient management. The potential to affect admission decisions may be greatest among the 19.1% of admitted patients with lengths of stay of no longer than 2 days.

The Centers for Medicare & Medicaid Services⁴² established a quality metric in 2016 that identifies diagnoses associated with potentially preventable ED use and hospital admission in patients with cancer, including pain, fever, dehydration, nausea, and emesis, among others. Such diagnoses may be present in half of patients with cancer who present to the ED.¹² However, because patients with these symptoms may have serious life-threatening conditions, additional work will be needed to develop risk stratification tools and modifications to ED and outpatient oncology processes that could provide an opportunity to safely decrease ED presentation and admission rates for patients with cancer. An additional area requiring study is the use of ED observation units to safely and efficiently treat this population. In addition, cancer center models, including infusion centers and the Oncology Medical Home Model,⁴³ may provide similar benefit.

These identified areas represent areas of opportunity to better support patients through their cancer treatment. Use of the ED appears to frequently be a marker for poorly controlled symptoms. Ultimately, identifying those patients at risk of ED use for potentially preventable conditions would allow intervention in the outpatient setting before symptoms become poorly controlled, improving overall care of the patient with cancer as well as achieving better ED resource use. Reducing ED use will also require novel approaches by oncologists to improve the support, access, and coordination of care for patients with cancer. These approaches could include development of systems to allow for more aggressive and preemptive management of expected complications. Such support will have to be easily accessible and available after hours and on weekends. The support should include scheduled follow-up checks as well as provisions for unscheduled in-person and remote care. Solving the acute access problem for those who need contact with the medical system but may not require ED care will be a key challenge for oncologists.

In addition, given the frequency and predictability of these symptoms, the opportunity is available to assist with and improve ED care, particularly when the primary oncology team is unavailable to treat the patient. Creating diagnostic, risk stratification, and treatment algorithms in partnership among emergency medicine, oncology, palliative care, and other stakeholders is necessary. A critical factor to improving such systems will be ensuring appropriate and rapid communication and follow-up for the patient, because lack of patient information and concern about the post-ED course is one of the primary drivers for ED admission decisions.⁴⁴

Our study has several limitations. First, this convenience sample included patients from large, primarily urban, academic medical centers, many affiliated with comprehensive cancer centers. Community hospital ED cancer care may differ in patient characteristics, resource availability, and outcomes.^45,46 As a result, our findings should be considered representative of large academic cancer centers. However, as we note below, in administrative national and statewide data sets, admission rates were similar between our study and others.^2,4

We also may have underestimated severity within the participating academic institutions themselves. Although only 135 participants (12.6%; 95% CI, 10.6%-14.7%) arrived on a weekend compared with 27% to 28% in administrative data sets, we enrolled 94 patients (8.7%; 95% CI, 7.1%-10.6%) arriving at night (11 pm-7 am) compared with approximately 13% in administrative data sets.^2,4 Also, 162 of 2337 approached patients (6.9%) were too ill or otherwise able to participate. One hundred forty non–English-speaking patients were assessed for eligibility (6.0%) and constituted 18.1% of the 773 approached who were deemed ineligible. Because patients with cancer and limited English proficiency report inferior treatment outcomes, their ineligibility may have affected results.⁴⁷ Any potential underestimated illness severity allows for 2 possible effects. First, our identified symptom severity and frequency admission rate, hospital length of stay, and other variables related to severity potentially represent the lower bound of severity. Second, underestimation of severity could result in a ceiling effect because it is more challenging to improve outcomes in a healthier compared with less healthy populations. Also, ED revisit and hospital readmission rates may be underestimated because we could not identify use of health care services in other hospitals systems.

Despite these concerns, findings in administrative data sets suggest that our study population was reasonably representative. Rivera et al² examined ED visits among adults with cancer using the Nationwide Emergency Department Sample, which includes academic and community EDs, and Mayer et al⁴ examined administrative data, including all ED visits in North Carolina. Both studies found similar admission rates (59.7% and 63%, respectively) as our study and a similar distribution of cancer types, which should diminish concerns regarding underestimation of disease severity.

This study represents, to our knowledge, the first prospective, multicenter investigation describing ED use by patients with cancer. By prospectively enrolling patients, we were able to identify in much greater detail several patient factors than in previous work. Most of the patients presenting to the ED with active cancer were ill on presentation and had poorly controlled symptoms. Most were admitted or placed in observation, many for short hospital stays. Our data suggest opportunities to improve care for patients before, during, and after an ED stay. Success will require goal-oriented collaboration among oncology, palliative care, and emergency medicine but will allow for improved ED use, improved symptom control and risk stratification in the ED, and establishment of safe disposition decisions for patients with cancer.

Accepted for Publication: February 2, 2019.

Published: March 22, 2019. doi:10.1001/jamanetworkopen.2019.0979

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2019 Caterino JM et al. JAMA Network Open.

Corresponding Author: Jeffrey M. Caterino, MD, MPH, Departments of Emergency Medicine and Internal Medicine, The Ohio State University Wexner Medical Center, 376 W 10th Ave, 751 Prior Hall, Columbus, OH 43035 (jeffrey.caterino@osumc.edu).

Author Contributions: Dr Caterino had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Caterino, Baugh, Grudzen, Madsen, Pallin, Reyes-Gibby, Shapiro, Lyman.

Acquisition, analysis, or interpretation of data: Caterino, Adler, Durham, Yeung, Hudson, Bastani, Bernstein, Baugh, Coyne, Grudzen, Henning, Klotz, Madsen, Pallin, Reyes-Gibby, Rico, Ryan, Shapiro, Swor, Venkat, Wilson, Thomas, Bischof.

Drafting of the manuscript: Caterino, Adler, Hudson, Madsen.

Critical revision of the manuscript for important intellectual content: Adler, Durham, Yeung, Hudson, Bastani, Bernstein, Baugh, Coyne, Grudzen, Henning, Klotz, Madsen, Pallin, Reyes-Gibby, Rico, Ryan, Shapiro, Swor, Venkat, Wilson, Thomas, Bischof, Lyman.

Statistical analysis: Caterino, Durham.

Obtained funding: Klotz.

Administrative, technical, or material support: Hudson, Bernstein, Baugh, Coyne, Grudzen, Henning, Klotz, Madsen, Pallin, Reyes-Gibby, Rico, Swor, Venkat, Wilson.

Supervision: Caterino, Baugh, Coyne, Pallin, Reyes-Gibby, Wilson.

Conflict of Interest Disclosures: Dr Caterino reported grants from AstraZeneca and personal fees from Wellstat Therapeutics outside the submitted work. Dr Yeung reported grants from DepoMed and Bristol-Myer Squibb and personal fees from Celgene outside the submitted work. Dr Bernstein reported nonfinancial support from National Cancer Institute during the conduct of the study. Dr Henning reported grants from Roche Diagnostics and Baxter outside the submitted work. Dr Ryan reported personal fees from EMCREG International outside the submitted work. Dr Wilson reported grants from Gilead and grants and personal fees from Janssen and Pfizer outside the submitted work. No other disclosures were reported.

Additional Contributions: The Comprehensive Oncologic Emergencies Research Network (CONCERN) Network supported this study with scientific review and suggestions on the study design and conduct. Nonniekaye Shelburne, MS, CRNP, AOCN, and the Epidemiology and Genomics Research Program within the National Cancer Institute’s Division of Cancer Control and Population Science, Bethesda, Maryland, assisted with study design and data analysis and ongoing support of the research network. Sarah Fagan, MPH, Epidemiology and Genomics Research Program, assisted with database cleaning and data analysis. Demetrios N. Kyriacou, MD, Department of Emergency Medicine, Northwestern University, was a leader in establishing CONCERN to allow the completion of this trial. None of them were compensated for this work.