Characteristics of Self-Triaged Emergency Department Visits by Adults With Cancer

ABSTRACT

Objectives: Adults with a new diagnosis of cancer frequently visit emergency departments (EDs) for disease- and treatment-related issues, although not exclusively. Many cancer care providers have 24/7 clinician phone triage available, but initial recorded phone messages tend to advise patients to go to the nearest ED if they are “experiencing a medical emergency.” It is unclear how well patients triage themselves to the optimal site of care.

Study Design: Cross-sectional study of tumor registry records (university patients diagnosed 2008-2018 and safety-net patients diagnosed 2012-2018) identifiably linked to electronic health records and a regional health information exchange.

Methods: We geoprocessed addresses to calculate driving time distance from the patient’s home to the ED. We used mixed-effects regression to predict the diagnosis code–based severity for ED visits within 6 months of diagnosis, clustering visits within patients and hospitals.

Results: A total of 39,498 adults made 38,944 ED visits to 67 different hospitals. Patients self-referred for 85.5% of visits and bypassed a median (IQR) of 13 (4-33) closer EDs. Visits closer to home were not significantly more clinically severe; visits were significantly less severe if the patient self-referred (adjusted odds ratio [AOR], 0.89; 95% CI, 0.81-0.97) or they were on weekends (AOR, 0.93; 95% CI, 0.87-0.99). Reanalyzing within each individual health system also showed similar findings.

Conclusions: Adults with cancer infrequently use available clinician advice before visiting the ED and may use factors other than clinical severity to determine their need for emergency care. Future work should explore the challenges that patients face navigating unplanned acute care, including reasons for underusing existing resources.

Am J Manag Care. 2023;29(9):e267-e273. https://doi.org/10.37765/ajmc.2023.89429

Takeaway Points

• Despite available 24/7 nurse triage phone lines and oncology-specific urgent care clinics, a diverse group of patients with cancer frequently did not call for help before an emergency department visit (85.5% of visits).
• Patient self-referred visits, visits to hospitals closer to a patient’s home, and visits on weekends were not for more severe clinical conditions.
• Findings suggest potential targets to help patients manage acute care needs during cancer treatment and are especially relevant to providers who will be participating in the Medicare Enhancing Oncology Model alternative payment pilot.

Adults with a new diagnosis of cancer frequently visit emergency departments (EDs) for disease- and treatment-related symptoms, and many of these visits may be preventable.^1-5 Despite receiving specialized treatment at one cancer treatment hospital, patients often visit other health systems.⁶ Because clinic telephone automated messages often advise that “if this is a medical emergency, please hang up and call 911,” a potential explanation is that patients who self-refer (go to the ED without calling their care team) are sicker and go to the nearest ED.

However, it is unclear whether clinical severity is associated with the proximity of the ED visited. A patient’s decision to go to the ED is rife with uncertainty,⁷ complicated by several factors. First, with little clinical training, patients and caregivers must determine the appropriate site of care by assessing illness severity. Further, appropriately selecting between an outpatient clinic, urgent care, specialized oncology urgent care,⁶ and the ED involves familiarity with the timing of the next available appointment and the clinical capabilities of each setting. Fear, current pain, and prior experiences with the ED or specific facilities^8,9 add further complexity to ED decision-making. Finally, outpatient clinics and cancer treatment centers have 24/7 telephone lines for clinical advice, but these resources may be underutilized.⁶

Inherent limitations in more widely available data sets prevent insights into multiple ED visits by the same patient. Yet cancer diagnosis details and unbilled clinical encounter data, across insurance types, are important.^10-12 We analyzed a comprehensive set of ED visits for patients across insurance types and without insurance who were treated by 2 health systems by linking tumor registries to the respective electronic health record (EHR) and an all-payer regional health information exchange containing identifiably linked hospital visits. To investigate patterns of ED use, we analyzed whether patients (1) traveled to the closest ED, (2) self-referred to the ED, or (3) visited on the weekend when they were most sick.

METHODS

Patient Cohort

Using the tumor registries of the 2 health systems comprising our regional comprehensive cancer center, we assembled a cohort of adults 18 years and older who received a cancer diagnosis between January 1, 2008, and June 30, 2018, at The University of Texas Southwestern Medical Center (hereafter, University) and between January 1, 2012, and December 31, 2018, at Parkland Health (public safety-net health system for the uninsured in Dallas County; hereafter, Safety Net). For patients with synchronous diagnoses, we selected the highest-stage cancer. For patients with metachronous cancer diagnoses, we selected the first diagnosis and excluded subsequent diagnoses. We excluded leukemias and nonmelanoma skin cancers, the former in line with other research¹³ that excluded acute leukemias because of inpatient treatment and hospitalizations due to disease relapse and differential acute care needs in early chronic leukemia.

During the study period, both health systems had 24/7 telephone triage with a clinician—either an oncologist or a nurse with triage protocols—who had access to the EHR and who logged calls into the record. Both health systems created oncology acute care clinics (University in mid-2012 and Safety Net in mid-2015) that had seen relatively little use compared with ED visits,⁶ but we included a variable to adjust for whether the ED visit occurred before or after acute care clinic creation.

The tumor registry provided demographic data including street address at the time of diagnosis, date of diagnosis, cancer type and stage, initial treatment modalities, comorbid diagnoses, and date of death. We dichotomized cancer stage into advanced (stages IIIB and higher for lung cancer, stages III and higher for pancreatic cancer, and stage IV for all others except brain cancer) vs nonadvanced stage.^6,14 We used comorbid diagnoses to generate a Charlson Comorbidity Index score.¹⁵

Dallas–Fort Worth Hospital Council Foundation Linkage

We identifiably linked patients to a database of ED visits in an all-payer regional health information exchange, the Dallas–Fort Worth Hospital Council Foundation (DFWHC Foundation), using a combination of first, middle, and last names; date of birth; zip code; medical record number; and sex. The DFWHC Foundation maintains a longitudinally linked record of approximately 12 million unique patients, including the uninsured, and their 65 million hospital encounters from more than 80 hospitals (98% of nonfederal hospitals, not including freestanding EDs or urgent care clinics) within a 100-mi radius of Dallas, Texas, where both health systems are based. The DFWHC Foundation provided claimslike data including hospital name and address, date of visit, disposition, and diagnosis codes.

Outcome: ED Visit Severity

We used principal diagnosis codes to generate a validated¹⁶ ED visit severity classification (patched¹⁷ and modified Billings severity^18,19: emergent, intermediate, nonemergent, alcohol/mental health/substance abuse, unclassified). We excluded alcohol/mental health/substance abuse and unclassified ED visits from the ordinal severity scale. Data from the DFWHC Foundation contained discharge diagnoses from the ED visit only if the patient was not hospitalized because hospitalizations reported only inpatient discharge diagnoses. Consequently, we added hospitalization as the highest severity in a 4-category ordinal severity outcome: hospitalized, emergent, intermediate, and nonemergent. See eAppendix Table 1 (eAppendix available at ajmc.com) for details on distribution of severities before and after these exclusions.

Geoprocessing

Patient address at the time of diagnosis was cleaned and processed in R version 4.0.2 (R Project for Statistical Computing). Treatment hospital (University or Safety Net), ED, and patient addresses were geocoded, and the distance between residential location and medical home and between residential location and ED was calculated using StreetMap Premium and Network Analyst in ArcGIS Pro version 2.6.1 (Esri). For each ED visit, we counted the number of EDs that were closer to the patient’s residential address by driving time. For our analyses, we categorized distance into quintiles. Visits with patient addresses that geocoded with low precision or to addresses outside Texas were excluded (10.4% of patients) (see eAppendix Table 2 for patient demographic comparison between high- and low-precision address matches).

EHR Linkage

We included ED visits within 180 days (6 months) of diagnosis to create a sample from patients with similar acute cancer care needs and excluded hospital visits overlapping with the diagnosis date. We extracted EHR (Epic) clinical interactions (telephone notes, secure messages, and outpatient visits for all specialties) from the day before an ED visit. ED visits without a preceding clinical contact were defined as self-referred.

Analysis of Primary Predictors

The prespecified primary predictors were patient ED self-referral vs prior clinician contact, distance traveled to the ED (number of closer EDs bypassed), and weekend/holiday vs weekday ED visits.

We used multivariable mixed-effects regression clustering ED visits within patient, and patient within hospital visited, to account for expected correlation between ED visits. Covariates included cancer type, stage, initial treatment modalities, death within 180 days after diagnosis, and Charlson Comorbidity Index score, plus demographic and geocoded variables as described earlier. Finally, we used marginal effects methods to convert adjusted odds ratios (AORs) into adjusted percentages for improved interpretability.

We used SAS 9.4 (SAS Institute) for data management and Stata/MP 15.1 (StataCorp) for statistical analyses. The University of Texas Southwestern Institutional Review Board approved this study (STU 112017-026; 122017-042).

Sensitivity Analyses

We assessed the robustness of our findings by reanalyzing with absolute distance traveled in miles and the number of closer EDs passed using distance in miles and by stratifying analyses by treatment home hospital (University and Safety Net) because these subgroups may have real and perceived differences in ED choice based on insurance status.

RESULTS

Cohort Characteristics

We identified 39,498 adults with a new diagnosis of cancer across both health systems. Half were women (50.0%), half were non-White (53.2%), and one-third (32.3%) were 65 years or older at time of diagnosis. Major cancer types included gastrointestinal cancers, including colorectal, at 16.7%; breast cancer at 14.6%; and lung cancer at 10.5%. One-third (35.1%) of patients were covered by Medicaid or uninsured at the time of diagnosis (18.7% at University and 77.0% at Safety Net).

We excluded ED visits from patients whose address was unable to be geocoded or out of state (5187 visits); when the patient was transferred to another facility (to avoid double counting because the transferred visit was also in the data set) or left without being seen/against medical advice (1644 visits); that had unclassified severity (7429 visits); or that were for alcohol, mental health, or injury (5777 visits). Finally, we excluded ED visits overlapping with the cancer diagnosis date (2551 visits). See eAppendix Table 2 for complete details on exclusions.

Of this cohort (n = 39,498), 40.8% of patients did not make any ED visits in the 180 days after their cancer diagnosis. Half (49.5%) made 1 to 3 ED visits, and 9.8% made 4 or more ED visits, accruing 38,944 total ED visits after the exclusions above. Of the patients who made 1 or more ED visits, 81.1% always self-referred to the ED, 6.5% at least once did not self-refer, and 12.4% did not self-refer at all. See Table 1 for demographics of the final study cohort of patients/visits.

Half of all ED visits were to the cancer treatment hospital (51.2%), but this comprised 29.0% of University patient visits to the University hospital and 74.4% of Safety Net patient visits to the SafetyNet hospital. This reflects wider hospital choice available to University patients (80.6% commercial or Medicare coverage) compared with Safety Net patients (23.0% with more flexible commercial or Medicare coverage; uninsured charity care limits payment for services only at the Safety Net health system).

ED Visit-Level Characteristics

Patients self-referred for 85.5% of ED visits. One-quarter (25.6%) of ED visits occurred on weekends/holidays. Patients lived a median (IQR) of 21.0 (14.1-29.6) minutes from the ED visited, corresponding to a median (IQR) of 13 (4-33) closer EDs bypassed in any direction. More than half (53.7%) of ED visits resulted in hospitalization, and one-quarter of ED visits (25.4%) were classified emergent but not hospitalized. See Table 2^17-19 for complete ED visit-level characteristics and eAppendix Table 3 for the top 5 diagnosis groups across severity.

The Figure displays the mean distance traveled to EDs by patients in each Census tract separately for the 2 subcohorts, as well as a map of the distribution of regional ED facilities and the volume of visits to each ED.

Regression Analysis

Patient self-referral to the ED. Multivariate mixed-effects regression modeling of ED visit severity with hospitalization found significantly lower visit severity when patients self-referred to the ED (AOR, 0.89; 95% CI, 0.81-0.97).

Distance traveled to the ED. There were no significant differences in the distribution of clinical severity of ED visits across any quintile of driving time to the ED.

Weekend ED visits. Weekend/holiday ED visits had significantly lower severity compared with weekday ED visits (AOR, 0.93; 95% CI, 0.87-0.99). See eAppendix Table 4 for complete modeling results and marginally adjusted proportions of ED visit clinical severity.

Sensitivity Analyses

Reanalyzing using alternate ED distance measures (distance in miles, number of EDs passed based on miles) did not substantively change the findings. Reanalyzing after separating the cohorts into their respective cancer treatment hospitals (University and Safety Net) generated no substantive differences in our findings. See eAppendix Table 4 for sensitivity analysis details.

DISCUSSION

Our analysis of the comprehensive ED visits from population-based samples of adults with a new diagnosis of cancer suggests that despite high rates of patient ED self-referral (no clinician contact before ED visit), such visits were significantly less clinically severe. Patients visited an array of different EDs, but visits to EDs closer to home were not associated with higher clinical severity. On weekends when outpatient clinics were closed, ED visits were also significantly less severe. Despite available 24/7 telephone access to a clinician, only 1 in 7 ED visits were preceded by clinician contact. An overarching explanation of our findings is that patients may use other criteria rather than clinical severity to determine their need for ED care.

A smaller analysis of patients with ED visits made to only a single cancer treatment hospital analyzed the details of phone triage advice they received, and it found that less than half of visits to the cancer treatment hospital were preceded by a call and that many clinician recommendations (60%) were to go to the ED.²⁰ The overall high hospitalization rate here is in line with the literature.^3,21 We are not aware of other geospatial investigations of comprehensive ED visits and visit severities among adults with cancer.

Our findings underscore an underrecognized barrier to patient-centered care: Patients and caregivers have little clinical training yet must determine when and where to present for new or worsening symptoms. Given this, many potential nonclinical influences drive ED use. For one, clinical severity does not distinguish the experienced severity of a patient’s pain and does not account for pain exceeding the available home medications.²² Symptoms may begin earlier in the week but unexpectedly escalate over the weekend before outpatient sites reopen. The burden of driving further to a cancer-treating hospital (Figure) rather than to hospitals closer to home also deserves further investigation. Our data do not directly capture patient decisions, but the inherent uncertainty of uninvestigated or worsening symptoms also complicates decision-making.

On the other hand, half of ED visits were made to other hospitals that may have less personalized cancer care information and may be less familiar managing cancer- and treatment-related complications. Because our analysis excluded ED visits for alcohol, mental health, or injury, cancer care details are likely to be relevant to shaping the evaluation and management of these acute presentations. Although we did not exclude other ED visits unrelated to cancer or treatment, the literature finds that only approximately 5% of ED visits made at any time since the cancer diagnosis were clearly unrelated.¹ Differences in the pattern of hospital choice and ED self-referral between the University and Safety Net patients suggest that although the type of insurance strongly influences hospital choice, especially for the uninsured, other unmeasured facility characteristics may also be influential, such as prior good or bad experiences.

Our findings also highlight that without widespread linked EHRs,²³ cancer care teams may not be aware of many of their patients’ ED visits and unplanned acute care needs. Because patients wish to avoid the ED and have more triage support from their clinical team,²² more work should be done to understand why so few patients use existing 24/7 clinician assistance. This finding—consistent across health systems serving disparate populations—underlines broader issues related to navigating care during a complicated time in a patient’s life. Indeed, these findings point to opportunities for care delivery improvement that may be more amenable to change than immutable patient characteristics traditionally noted as risk factors for avoidable emergency care. For both health systems, initial patient calls are routed to clinicians who call patients back with direct triage follow-ups. We did not have data on time to a returned call or number of calls to reconnect, although these may be potential explanations for underuse and important avenues for future research.

Implications for Practice and Policy

Nearly all participants in the Center for Medicare and Medicaid Innovation’s Oncology Care Model (OCM), the first cancer-specific alternative payment model, aimed to reduce ED use,²⁴ but there was no significant reduction in ED visit rates by the end of the 5-year pilot program.²⁵ Both health systems studied here participated in the OCM. Because existing tools for care coordination such as 24/7 telephone triage (a requirement of OCM participation) and promising oncology acute care clinics⁶ may be underutilized, providers should better understand the uncertainties that patients and caregivers face navigating the health system, especially as they move forward to the next iteration of the program, the Enhancing Oncology Model.

Limitations

ED visit severity was based on a validated diagnosis code–based algorithm,^16,17 and although validation included patients with cancer, it is unclear how accurate this multicondition algorithm is in distinguishing severity for this specific population. Classifying clinical severity using a primary diagnosis code has inherent limitations.²⁶ Future work should explore more sophisticated claims-based algorithms, perhaps incorporating procedure codes. Additionally, there is substantial clinician discretion in hospitalization decisions that often incorporates important nonclinical factors.^27,28 However, to maximize the breadth and comprehensiveness of our unique integrated data set, classification using the primary diagnosis code was the strongest available measure of clinical severity.

It is possible that patients had clinical interactions not captured in the EHRs—for instance, if the patient’s primary care provider was outside the health system (especially in subgroups who were insured and lived far from the cancer treatment hospital). However, we expect this undercounting would be minor because patients likely prefer to call their cancer care team in this early phase of care. A prior study detailing the nature of triage advice given has documented regular logging of these unbilled encounters.²⁰

In computing distance to the ED, we used the patient’s home address. However, patients may visit an ED further from their home for several reasons. For example, a further ED might be closer to work or their primary care office or is in network with their health insurance. Patients may also rely on public transportation, presenting different travel times. Alternatively, ambulances may transport patients to a closer ED without much patient input. We were unable to distinguish these factors, which might bias our results, although in an unclear summative direction.

Although we present these findings from the combined population-based sample of an academic referral center and integrated safety-net health system in a large metropolitan area spanning urban, suburban, and rural areas, it is unclear how generalizable these findings are to community cancer providers and other health care markets. However, our findings are from a large, diverse population over many years and across a broad range of insurance types, including the uninsured.

CONCLUSIONS

Despite additional available clinical resources, patients with cancer and their caregivers face challenges navigating the complexity of care, especially when it comes to identifying the appropriate setting for new or worsening symptoms. Although common prerecorded phone messaging advises patients to seek emergency care for emergent conditions, this may be especially unhelpful for patients with cancer who may have a heightened awareness and fear of their symptoms. When patients make less urgent visits to the ED despite facing unpredictable wait times and potential disruptions to cancer treatment, providers should take this as a signal to better address unmet patient needs. Our findings suggest that upstream patient decisions leading to an ED visit may be influenced by nonclinical factors and that there are opportunities to maximize existing clinical resources to optimize acute care coordination. 

Author Affiliations: Department of Internal Medicine (ASH, JWS), Peter J. O’Donnell Jr. School of Public Health (ASH, AH, HF, SZ), and Department of Emergency Medicine (DMC), The University of Texas Southwestern Medical Center, Dallas, TX; Harold C. Simmons Comprehensive Cancer Center, The University of Texas Southwestern Medical Center (ASH, JWS, NS, AB), Dallas, TX; Department of Population Health, University of Kansas School of Medicine (SJCL), Kansas City, KS; Parkland Health (NS), Dallas, TX; Institute for Health, Health Care Policy and Aging Research, Rutgers Biomedical and Health Sciences and RWJBarnabas Health (EAH), New Brunswick, NJ.

Source of Funding: This work was supported by the Texas Health Resources Clinical Scholars Program, by a National Cancer Institute Cancer Center Support Grant (1P30CA142543), and by a Clinician Scientist Development Grant (CSDG-20-023-01-CPHPS) from the American Cancer Society.

Author Disclosures: The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (ASH, AH, SJCL, AB, EAH); acquisition of data (ASH, HF); analysis and interpretation of data (ASH, AH, DMC, HF, SJCL, JWS, NS, SZ, AB, EAH); drafting of the manuscript (ASH, AH, DMC, SJCL, JWS, EAH); critical revision of the manuscript for important intellectual content (ASH, AH, DMC, HF, SJCL, JWS, NS, SZ, AB, EAH); statistical analysis (ASH, HF, SZ); provision of patients or study materials (ASH, NS); obtaining funding (ASH); administrative, technical, or logistic support (ASH, AH, HF, NS, SZ, EAH); and supervision (ASH, DMC, SJCL, JWS).

Address Correspondence to: Arthur S. Hong, MD, MPH, The University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75390-9169. Email: Arthur.Hong@UTSouthwestern.edu.

REFERENCES

1. Mayer DK, Travers D, Wyss A, Leak A, Waller A. Why do patients with cancer visit emergency departments? results of a 2008 population study in North Carolina. J Clin Oncol. 2011;29(19):2683-2688. doi:10.1200/JCO.2010.34.2816

2. Manzano JG, Luo R, Elting LS, George M, Suarez-Almazor ME. Patterns and predictors of unplanned hospitalization in a population-based cohort of elderly patients with GI cancer. J Clin Oncol. 2014;32(31):3527-3533. doi:10.1200/JCO.2014.55.3131

3. Rivera DR, Gallicchio L, Brown J, Liu B, Kyriacou DN, Shelburne N. Trends in adult cancer-related emergency department utilization: an analysis of data from the nationwide emergency department sample. JAMA Oncol. 2017;3(10):e172450. doi:10.1001/jamaoncol.2017.2450

4. Hong AS, Sadeghi N, Harvey V, Lee SC, Halm EA. Characteristics of emergency department visits and select predictors of hospitalization for adults with newly-diagnosed cancer in a safety-net health system. J Oncol Pract. 2019;15(6):e490-e500. doi:10.1200/JOP.18.00614

5. Panattoni L, Fedorenko C, Greenwood-Hickman MA, et al. Characterizing potentially preventable cancer- and chronic disease–related emergency department use in the year after treatment initiation: a regional study. J Oncol Pract. 2018;14(3):e176-e185. doi:10.1200/JOP.2017.028191

6. Hong AS, Froehlich T, Clayton Hobbs S, Lee SJC, Halm EA. Impact of a cancer urgent care clinic on regional emergency department visits. J Oncol Pract. 2019;15(6):e501-e509. doi:10.1200/JOP.18.00743

7. Rising KL, Hudgins A, Reigle M, Hollander JE, Carr BG. “I’m just a patient”: fear and uncertainty as drivers of emergency department use in patients with chronic disease. Ann Emerg Med. 2016;68(5):536-543. doi:10.1016/j.annemergmed.2016.03.053

8. Hudgins A, Rising KL. Fear, vulnerability and sacrifice: drivers of emergency department use and implications for policy. Soc Sci Med. 2016;169:50-57. doi:10.1016/j.socscimed.2016.09.030

9. Uscher-Pines L, Pines J, Kellerman A, Gillen E, Mehrotra A. Deciding to visit the emergency department for non-urgent conditions: a systematic review of the literature. Am J Manag Care. 2013;19(1):47-59.

10. Lau EC, Mowat FS, Kelsh MA, et al. Use of electronic medical records (EMR) for oncology outcomes research: assessing the comparability of EMR information to patient registry and health claims data. Clin Epidemiol. 2011;3:259-272. doi:10.2147/CLEP.S23690

11. Garvin JH, Herget KA, Hashibe M, et al. Linkage between Utah all payers claims database and central cancer registry. Health Serv Res. 2019;54(3):707-713. doi:10.1111/1475-6773.13114

12. Brown J, Grudzen C, Kyriacou DN, et al. The emergency care of patients with cancer: setting the research agenda. Ann Emerg Med. 2016;68(6):706-711. doi:10.1016/j.annemergmed.2016.01.021

13. 2018 Measure Updates and Specifications Report: Admissions and Emergency Department Visits for Patients Receiving Outpatient Chemotherapy, Version 2.0. Mathematica Policy Research; 2018.

14. Brooks GA, Li L, Uno H, Hassett MJ, Landon BE, Schrag D. Acute hospital care is the chief driver of regional spending variation in Medicare patients with advanced cancer. Health Aff (Millwood). 2014;33(10):1793-1800. doi:10.1377/hlthaff.2014.0280

15. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373-383. doi:10.1016/0021-9681(87)90171-8

16. Gandhi SO, Sabik L. Emergency department visit classification using the NYU algorithm. Am J Manag Care. 2014;20(4):315-320.

17. Johnston KJ, Allen L, Melanson TA, Pitts SR. A “patch” to the NYU emergency department visit algorithm. Health Serv Res. 2017;52(4):1264-1276. doi:10.1111/1475-6773.12638

18. Billings J, Parikh N, Mijanovich T. Emergency department use in New York City: a substitute for primary care? Issue Brief (Commonw Fund). 2000;(433):1-5.

19. Ballard DW, Price M, Fung V, et al. Validation of an algorithm for categorizing the severity of hospital emergency department visits. Med Care. 2010;48(1):58-63. doi:10.1097/MLR.0b013e3181bd49ad

20. Hong AS, Chang H, Courtney DM, et al. Patterns and results of triage advice before emergency department visits made by patients with cancer. JCO Oncol Pract. 2021;17(4):e564-e574. doi:10.1200/OP.20.00617

21. Caterino JM, Adler D, Durham DD, et al. Analysis of diagnoses, symptoms, medications, and admissions among patients with cancer presenting to emergency departments. JAMA Netw Open. 2019;2(3):e190979. doi:10.1001/jamanetworkopen.2019.0979

22. Kaufmann TL, Rendle KA, Aakhus E, et al. Views from patients with cancer in the setting of unplanned acute care: informing approaches to reduce health care utilization. JCO Oncol Pract. 2020;16(11):e1291-e1303. doi:10.1200/OP.20.00013

23. Hochman M, Garber J, Robinson EJ. Health information exchange after 10 years: time for a more assertive national approach. Health Affairs Forefront. doi:10.1377/forefront.20190807.475758

24. Evaluation of the Oncology Care Model: performance period one. CMS. Accessed August 15, 2023. https://innovation.cms.gov/ocm-secondannualeval-pp1-fg.pdf

25. Keating NL, Jhatakia S, Brooks GA, et al; Oncology Care Model Evaluation Team. Association of participation in the Oncology Care Model with Medicare payments, utilization, care delivery, and quality outcomes. JAMA. 2021;326(18):1829-1839. doi:10.1001/jama.2021.17642

26. Chen AT, Muralidharan M, Friedman AB. Algorithms identifying low-acuity emergency department visits: a review and validation study. Health Serv Res. 2022;57(4):979-989. doi:10.1111/1475-6773.14011

27. Feng Z, Jung HY, Wright B, Mor V. The origin and disposition of Medicare observation stays. Med Care. 2014;52(9):796-800. doi:10.1097/MLR.0000000000000179

28. Martin GP, Wright B, Ahmed A, Banerjee J, Mason S, Roland D. Use or abuse? a qualitative study of emergency physicians’ views on use of observation stays at three hospitals in the United States and England. Ann Emerg Med. 2017;69(3):284-292.e2. doi:10.1016/j.annemergmed.2016.08.458