Emergency Department Management of Patients With Febrile Neutropenia: Guideline Concordant or Overly Aggressive?

Transcription

1 ORIGINAL CONTRIBUTION Emergency Department Management of Patients With Febrile Neutropenia: Guideline Concordant or Overly Aggressive? Christopher W. Baugh, MD, MBA, Thomas J. Wang, BS, Jeffrey M. Caterino, MD, MPH, Olesya N. Baker, PhD, Gabriel A. Brooks, MD, Audrey C. Reust, PA-C, and Daniel J. Pallin, MD, MPH ABSTRACT Objectives: The Infectious Diseases Society of America and the American Society of Clinical Oncology recommend risk stratification of patients with febrile neutropenia (FN) and discharge with oral antibiotics for low-risk patients. We studied guideline concordance and clinical outcomes of FN management in our emergency department (ED). Methods: Our urban, tertiary care teaching hospital provides all emergency and inpatient services to a large comprehensive cancer center. We performed a structured chart review of all FN patients seen in our ED from January 2010 to December Using electronic medical records, we identified all visits by patients with fever and an absolute neutrophil count of <1000 cells/mm 3 and then included only patients without a clear source of infection. Following national guidelines, we classified patients as low or high risk and assessed guideline concordance in disposition and parenteral versus oral antibiotic therapy by risk category as our main outcome measure. Results: Of 173 qualifying visits, we classified 44 (25%) as low risk and 129 (75%) as high risk. Management was guideline concordant in 121 (70%, 95% confidence interval [CI] = 63% to 77%). Management was guideline discordant in 43 (98%, 95% CI = 88% to 100%) of low-risk patients versus 9 (7%, 95% CI = 3% to 13%) of high-risk patients (relative risk [RR] = 14, 95% CI = 7.5 to 26). Of 52 guideline-discordant cases, 36 (83%, 95% CI = 72% to 93%) involved low-risk cases with treatment that was more aggressive than recommended. Conclusions: Guideline concordance was low among low-risk patients, with management tending to be more aggressive than recommended. Unless data emerge that undermine the guidelines, we believe that many of these hospitalizations and parenteral antibiotic regimens can be avoided, decreasing the risks associated with hospitalization, while improving antibiotic stewardship and patient comfort. A gap in research on oncologic emergencies has been identified, with the formation of the Comprehensive Oncologic Emergencies Research Network (CONCERN). 1 One of the most common oncologic emergencies is febrile neutropenia (FN) associated with chemotherapy. FN occurs in 10% 50% of patients From the Department of Emergency Medicine, Brigham and Women s Hospital (CWB, ONB, ACR, DJP), Boston, MA; Harvard Medical School (TW), Boston, MA; the Gastrointestinal Cancer Center, Dana Farber Cancer Institute (GAB), Boston, MA; and the Department of Emergency Medicine, The Ohio State University Wexner Medical Center (JMC), Columbus, OH. Received May 2, 2016; revision received August 8, 2016; accepted August 15, The abstract for this study was presented at the Northeast Regional Society for Academic Emergency Medicine Regional Meeting, Worcester, MA, March This study was supported by a Milton Fund award at Harvard Medical School. The authors have no potential conflicts to disclose. Authorship contributions: CWB, TW, GAB, ACR, and DJP conceived the study; CWB served as principal investigator; DJP provided statistical advice on study design; ONB analyzed the data; JMC provided assistance with interpretation of the data; CWB, DJP, and TW drafted the manuscript; all authors contributed substantially to its revision; and CWB takes responsibility for the paper as a whole. Supervising Editor: D. Mark Courtney, MD. Address for correspondence: Christopher W. Baugh MD, MBA; cbaugh@partners.org. Reprints will not be available. ACADEMIC EMERGENCY MEDICINE 2017;24: by the Society for Academic Emergency Medicine doi: /acem ISSN PII ISSN

2 84 Baugh et al. FEBRILE NEUTROPENIA RISK STRATIFICATION with solid tumors and > 80% of those with hematologic malignancies. 2,3 These patients are typically hospitalized and receive intravenous (IV) antibiotics, even though clinically documented bacterial infections occur in fewer than 30% of febrile episodes. 2 Thus, 70% are never found to have a bacterial infection and might be spared parenteral antibiotics and hospitalization, if they could be identified. Guidelines from the Infectious Diseases Society of America (IDSA) and American Society of Clinical Oncology (ASCO) recommend use of the Multinational Association for Supportive Care in Cancer (MASCC) score to identify patients safe for outpatient management (Table 1). 2 4 Patients with a MASCC score of 21 are designated low risk and are recommended for oral antibiotic treatment in an outpatient setting, so long as they do not meet additional criteria that dictate inpatient care (see Data Supplements S1 and S2, available as supporting information in the online version of this paper). Emergency physicians are challenged by the requirement to be aware of recommendations from various areas of medicine. They (and their collaborating oncologists, who often give input by phone) may err on the side of more-aggressive care, resulting in the overuse of broad-spectrum IV antibiotics and hospital admission. At our institutions, we have documented a lack of guideline familiarity, especially among emergency physicians, with a bias toward inpatient care and parenteral antibiotics for all FN cases. 5 This patient population is particularly vulnerable to adverse medication reactions, colonization with drug-resistant organisms, antibiotic complications (e.g., Clostridium difficile colitis) and iatrogenic adverse events associated with an avoidable hospitalization. 2,3 Table 1 MASCC Index MASCC* risk-index score Burden of illness (symptom severity) No or mild symptoms 5 Moderate symptoms 3 Severe symptoms or moribund 0 No hypotension (sbp 90 mm Hg) 5 No chronic obstructive pulmonary disease 4 Solid tumor or hematologic malignancy without 4 previous fungal infection No dehydration requiring IV therapy 3 Outpatient status at onset of fever 3 Age < 60 y 2 MASCC = Multinational Association for Supportive Care in Cancer; sbp = systolic blood pressure. *MASCC Score 21 is considered low risk. Our aim was to assess the treatment rendered and outcomes associated with emergency department (ED) visits for cancer patients with FN and to compare guideline concordance among high-risk and low-risk groups. Investigating the connection between management guidelines, actual care delivered, and patient outcomes is an important first step toward improving the quality of care. METHODS Study Setting and Population We conducted a structured chart review of cancer patients with FN presenting to the ED of an urban tertiary care teaching hospital affiliated with a large cancer center, between January 2010 and December A medical student trained to perform the structured chart review collected all data, similar to recent investigations into the care of this patient population and in accordance with previously published best practices for chart review investigations. 6 8 The first author reviewed a random sample of 20 cases (12% of all) to calculate a kappa coefficient around MASCC score agreement, which was j = 0.8 (95% confidence interval [CI] = 0.5 to 1.0). Each case was preassigned with a unique case number, and we used a random-number generator to pull 20 cases from the total pool. We assessed guideline concordance of inpatient/outpatient management and oral/parenteral antibiotics, stratifying by low versus high risk. In addition to MASCC score components, we also reviewed ED notes for the presence of any high-risk IDSA or ASCO criteria, even if the MASCC score was low risk. If any high-risk criteria were present, we took a conservative approach and classified the patient as high risk. We assessed clinical outcomes at 30 days via electronic medical record review. The Dana Farber Cancer Institute is a National Cancer Institute designated comprehensive cancer center. The Brigham and Women s Hospital is an urban, tertiary care teaching hospital; the ED had 60,050 adult visits in 2014 and is the main source of emergency and inpatient care for Dana Farber patients. We queried an electronic health record database to find all patients with active cancer seen in the ED with neutropenia during the study period and then reviewed visit notes to find patients with a diagnosis of FN without a clear source of infection (e.g., lobar

3 ACADEMIC EMERGENCY MEDICINE January 2017, Vol. 24, No Patients with absolute neutrophil count <1000 within 2 days of ED visit 604 patients excluded without neutropenia at ED visit or fever 326 Patients with neutropenia at ED visit and reported or measured fever >100.4 F 153 patients excluded with suspected source of infection at index ED visit: Pulmonary infiltrate with respiratory symptoms Pyuria with urinary tract infection symptoms Indwelling central line with suspected line infection Exam findings suggestive of cellutlitis Other clear exam findings suggestive of infectious source 173 Patients without suspected source of infection 44 Low risk 129 High risk Figure 1. Search strategy and exclusion criteria for study cohort. consolidation on chest x-ray and clinical symptoms of pneumonia; see Figure 1 for details) at the conclusion of the index ED visit. We did not use ICD-9 data for selection of patients. We defined neutropenia as an absolute neutrophil count < 1000 cells/mm 3 and fever as any recorded or reported temperature > 38.0 C. 2 We defined active cancer as any cancer with treatment in the previous 6 months (i.e., chemotherapy, radiation therapy, or surgery). Study Protocol and Measurements We categorized patients as low or high risk according to the recommendations of IDSA and ASCO. Low-risk patients were required to have a MASCC score of 21, but also the absence of another clinical indication for inpatient management as defined in the ASCO and IDSA guidelines (Data Supplement S1 and S2). High-risk patients had either a MASCC score of <21 or at least one clinical indication for inpatient care. The MASCC score was originally created as a tool to be used prospectively to determine the risk of adverse outcomes in cancer patients with FN. In our study, we applied this score retrospectively via chart review. Some of the components of the score are objective and simple to abstract from a chart (e.g., patient age). However, two components are inherently subjective: burden of disease and presence of dehydration requiring IV fluids. Regarding burden of disease, we assigned none or mild if the ED note described the patient as well appearing or in no distress, with the absence of hypotension (systolic pressure < 90 mm Hg), tachypnea (respiratory rate 24 breaths/min), severe tachycardia (heart rate 120 beats/min), or hypoxia (room air oxygen saturation < 90%). We categorized the burden of disease as severe or moribund status if the ED note described the patient as in any degree of distress or ill appearing or if the patient was noted to have any of the vital sign abnormalities described above. We designated the disease burden as

4 86 Baugh et al. FEBRILE NEUTROPENIA RISK STRATIFICATION moderate in the absence of criteria fitting either of the two scenarios above. Regarding dehydration requiring IV fluids, we designated the patient as dehydrated if the ED note documented that the patient appeared dehydrated or had physical examination findings consistent with dehydration (i.e., dry or tacky mucous membranes, poor skin turgor) and also received IV hydration. We also labeled patients as dehydrated if their BUN:creatinine ratio was > 20:1 and their BUN or creatinine was above the upper limit of normal range (>23 and >1.2 mg/dl, respectively). We defined bacteremia or fungemia as a positive blood culture with 1) at least one blood culture positive with no indication of suspected contaminants in the patient notes and 2) a diagnosis of bacteremia/fungemia present in the patient s discharge note We defined sepsis-induced hypotension as systolic pressure < 90 mm Hg during any part of the hospital stay with suspected source of infection. Key Outcome Measures Our main outcome measure was the proportion of patients with guideline-concordant management. We designated care as guideline concordant if both disposition (home vs. inpatient) and route of antibiotics (oral vs. parenteral) corresponded to guideline recommendations. We report results as risk ratios (RRs), because we selected participants based on exposure, not based on outcome, making the RR the appropriate metric. Our secondary outcome measures were the identification of bacteremia or fungemia (i.e., positive blood cultures not suspected to be a contaminant), sepsis-induced hypotension, or death within 30 days of the index ED visit, stratified by risk level. We also examined the use of vancomycin, which is not recommended by the guidelines as empiric treatment without qualifying risk factors, such as history of prior methicillin-resistant Staphylococcus aureus infection or suspected indwelling line infection. 12 Data Analysis For our main outcome, we report the percentage with binary 95% CI. To evaluate predictors of guideline discordance, we constructed a multivariable logistic regression model with guideline concordance as the outcome and the following predictors, chosen a priori: low versus high risk (binary), age (continuous), sex (binary), cancer type (solid or hematologic, dichotomous), and stem cell or bone marrow transplant (binary). These predictors were chosen based on biologic plausibility and clinical relevance. All of these predictors chosen a priori were included in the final model. We included interaction terms for each predictor s interaction with risk level, and upon finding that none was statistically significant, we reverted to the basic model with no interaction terms. All variance inflation factors for the included predictors were < 3, implying lack of problematic collinearity. We performed all statistical analyses with StataMP 13. The study was approved by our institutional review board. Sensitivity Analysis We performed a sensitivity analysis of our results by varying the subjective elements of the MASCC score most vulnerable to disparate valuation: the assessment of disease burden and dehydration requiring IV antibiotics. We tested whether adding/subtracting points in these components of the score would potentially move patients from low risk to high risk or vice versa. First, we tested the impact of misclassifying the disease burden, which is a subjective determination of patient acuity at the time of presentation. In cases where we had determined no or mild symptoms, we upgraded the score to moderate ; in cases determined as moderate, we upgraded the score to severe or moribund. Second, we tested the possibility of dehydration misclassification. We applied more inclusive criteria of either BUN:creatinine ratio of >20:1 or abnormal BUN or creatinine values above the normal cutoff ranges to classify patients as dehydrated. Previously, we required both to satisfy our definition. All other physical examination findings and indication of dehydration in the charts still applied to this reclassification. RESULTS Characteristics of Study Subjects Figure 1 is a participant flow chart. We identified 930 cases of neutropenia that coincided with an ED visit, of which 326 had fever, identified by manual chart review. After excluding FN cases with a known etiology corroborated by physical examination or laboratory findings (Figure 1), and patients without active cancer, we obtained a final sample of 173 ED visits by patients with undifferentiated FN. We classified 44 (25%) as low risk and 129 (75%) as high risk. We

5 ACADEMIC EMERGENCY MEDICINE January 2017, Vol. 24, No present participant characteristics in Table 2. The mean absolute neutrophil count was 500 in low-risk patients and 410 in high-risk patients (difference = 80; 95% CI = 20 to 200). Main Results Figure 2 displays a breakdown of guideline discordance by disposition and by route of antibiotic administration. Overall, 52 patients (30%, 95% CI = 23% to 37%) received guideline-discordant care, including 43 (98%, 95% CI = 88% to 100%) of low-risk patients and 9 (7%, 95% CI = 3.2% to 13%) of highrisk patients. For the low-risk group, discordance was driven by guideline-discordant inpatient admission in nearly all discordant cases, the majority of whom also received IV antibiotics in discordance with guidelines (Figure 2). The median hospital length of stay in this group was 4 days (interquartile range = 2-6 days). For high-risk patients, discordance was very low and similar across both antibiotic route and setting of care, with nearly all admitted for inpatient care and treated with parenteral antibiotics. Oral or no antibiotic treatment represented the majority of discordant care in the high-risk group. The crude RR for guideline-discordant care by risk category was 14, with more discordance in the low-risk group. Table 3 displays the results of the multivariable analysis and shows that being in the low-risk category was a strong predictor of guideline discordant care (RR = 14.5). Hematologic malignancy was the only Table 2 Characteristics of 173 Patients with FN Presenting to the ED Characteristic Low risk High risk % Difference (95% CI) All cases (N = 173) 44 (25%) 129 (75%) Age (y), median (interquartile range) 58 (47 66) 61 (49 67) 1 ( 6 to3) Sex Female (n = 99) 26 (59%) 73 (57%) 2.5 ( 15 to 20) Male (n = 74) 18 (41%) 56 (43%) 2.5 ( 1.9 to 1.5) Underlying malignancy Solid (n = 92) 19 (43%) 73 (57%) 13 ( 31 to 3.8) Hematologic (n = 81) 25 (57%) 56 (43%) 13 ( 3.8 to 31) Stem cell or bone marrow transplant (n = 21) 5 (11%) 16 (12%) 1.0 ( 12 to 10) FN = febrile neutropenia. RR 14 (95% CI ) RR 39 (95% CI ) RR 16 (95%CI ) 100% 98% 91% 98% 80% 60% 40% 20% 0% 30% 25% 30% 7% 2% 6% All Cases Low Risk High Risk All Cases Low Risk High Risk All Cases Low Risk High Risk Either setting of care or antibiotic route Inappropriate setting of care (Inpatient for low-risk; outpatient for high-risk) Inappropriate antibiotic route (Parenteral for low-risk; oral for high-risk; no antibiotics at all for either group) Figure 2. Discordant care between recommended and actual management of ED patients with FN. FN = febrile neutropenia; RR = risk ratio.

6 88 Baugh et al. FEBRILE NEUTROPENIA RISK STRATIFICATION Table 3 Multivariable Analysis of Predictors of Guideline Discordance Predictor RR in Predicting Guideline Discordance 95%CI Low risk (vs. high risk) Age (decade) 1.0 (not significant) Male sex (vs. female) 0.8 (not significant) Hematologic malignancy 0.8 (significant) (vs. solid) Stem cell or bone marrow transplant (vs. not) 0.9 (not significant) RR = risk ratio. other variable that significantly predicted discordant treatment, although significance was borderline. In the high-risk group, 18 patients (32%, 95% CI = 20% to 45%) received vancomycin without a clinical indication. Among all patients who received IV antibiotics (n = 153), vancomycin was used without guideline support in 26 cases, or 17% (95% CI = 11% to 23%). We display 30-day clinical outcomes in Figure 3. No patients were lost to follow-up at 30 days. Comparing low-risk with high-risk patients, positive blood culture rates were significantly higher in the high-risk group and no low-risk patients experienced sepsisinduced hypotension or death, compared with 12 (9.3%) and 7 (5.4%) of high-risk patients, respectively. Positive blood cultures were predominantly due to bacteremia, with fungemia present in only one high-risk patient and one low-risk patient. Our sensitivity analysis yielded only an additional three cases crossing the threshold from low risk to high risk when upgrading the disease burden. The more-inclusive dehydration criterion resulted in seven low-risk patients being reclassified as high risk. A total of 12 low-risk patients were reclassified as high risk when both disease burden and dehydration criterion changed at the same time. With all reclassifications applied, management was guideline discordant in 31 (96.9%, 95% CI = 83.8% to 99.9%) low-risk patients versus 12 (8.5%, 95% CI = 4.5% to 14%) high-risk patients (RR = 11, 95% CI = 6.6 to 20). Comparing low-risk versus high-risk patients, within 30 days, one (3.1%) versus seven (5.0%) had bacteremia or fungemia (RR = 1.6, 95% CI = 0.2 to 13), 0 versus 12 (8.5%) had sepsis-induced hypotension (difference = 8.5, 95% CI = 3.9 to 13), and 0 versus 7 (5.0%) died (difference = 5.0, 95% CI = 1.4 to 8.5). DISCUSSION Our investigation suggests that emergency physicians are indiscriminately using IV antibiotics and hospitalization in cancer patients with low-risk FN. Low-risk patients comprised 25% of all FN cases, and management was more aggressive than guidelines recommend in 98% of these patients with a median hospital length of stay of 4 days. At the same time, there were no episodes of sepsis-induced hypotension or death within 30 days in the low-risk group. This supports the guidelines ability to identify patients who are appropriate candidates for outpatient care and oral antibiotics. 13 First-line therapy for low-risk patients is oral fluoroquinolones, which have excellent bioavailability. 14 Low risk is not synonymous with no risk the 2.2% rate of bacteremia or fungemia in this group highlights the importance of close outpatient followup, both with rapid access to providers and with a reliable patient capable of seeking care if his or her clinical condition worsens. It is important to understand that fluoroquinolones have the same bioavailability by mouth as by the IV route, and thus the rare patient RR 2.4 (95% CI ) Diff=9.3 (95% CI ) Diff=5.4 (95%CI ) 10.0% 9.3% 8.0% 6.0% 5.4% 5.4% 4.0% 2.0% 0.0% 2.2% 0.0% 0.0% Positive blood culture Sepsis-induced hypotension Death Low Risk High Risk Figure 3. Thirty-day clinical outcomes in ED patients with FN. FN = febrile neutropenia; RR = risk ratio.

7 ACADEMIC EMERGENCY MEDICINE January 2017, Vol. 24, No with nonseptic bacteremia would be treated as well at home as in the hospital. 15 Our results show that guideline discordance was primarily driven by overtreatment of low-risk patients, not undertreatment of high-risk patients. As expected, setting of care and use of parenteral versus oral antibiotics were likely to be closely linked to one another in the management plan. We were surprised to find that increased age did not predict risk of guideline discordance in the controlled multivariable analysis. We had expected to observe a tendency to hospitalize patients at higher age despite a low-risk determination. 16,17 Seventeen percent of patients who received IV antibiotics were treated with vancomycin without guideline support, increasing the unnecessary risk of vancomycin-resistant enterococcus bacteremia in a vulnerable patient population. 12,18 This finding reveals an opportunity to improve the quality of care via education and implementation of tools (i.e., electronic health record decision support) to support antibiotic stewardship in a vulnerable population. 2,3,12,18,19 The target of such efforts should include both emergency physicians and oncologists, since it is unlikely that treatment and disposition decisions for these patients are made without input from oncologists. A recent survey found that guideline awareness and use was significantly higher for oncologists, yet still not optimal. 5 The MASCC score has been established for over a decade, but it is neither highly sensitive nor specific for adverse patient outcomes. 20 MASSC also does not meet optimal characteristics of clinical decision rules. 21 A recent alternative scoring method, the CISNE (Clinical Index for Stable Febrile Neutropenia) score is purportedly more accurate in predicting adverse outcomes; we did not, however, include this score in our analysis due to its lack of penetration in existing society recommendations and difficulty in acquiring its scoring factors from retrospective chart reviews. 22 While we found no cases of sepsis-induced hypotension or death in the low-risk population, 2.2% of these patients did have positive blood cultures. We recommend following IDSA and ASCO guidelines, which advise sending these patients home. 2,3 However, if discharge to home after initial ED evaluation is not the best plan (i.e., barriers to close outpatient followup due to weekend/holiday), care in an observation unit may be an acceptable alternative to an inpatient admission. The duration of a typical observation stay will not span the necessary time for a final blood culture result, and the patient is still exposed to the hazards of hospitalization, but it may be a superior alternative to inpatient admission since that exposure is limited to around 15 hours for an observation unit stay, less than half the duration of a typical inpatient admission. 23 Hospitalization carries risk of nosocomial infections, and treatment with broad-spectrum antibiotic confers increased risk of C. difficile colitis and selection for multidrug-resistant bacterial strains. LIMITATIONS Our study has several limitations. First, it was a singlecenter study, which limits generalizability. However, this initial work is an important first step toward recognizing gaps in care in the treatment of oncologic emergencies. We are members of the CONCERN, a newly formed research consortium sponsored by the National Cancer Institute and are planning a large multicenter study that may confirm the results of this single-center study. 1,24 Second, we determined the MASCC score via retrospective chart review; as such, we are limited by the study design to ascertain whether any sepsis-induced hypotension or death was prevented by more aggressive inpatient management of low-risk patients. A prospective cohort study, therefore, may be indicated to further evaluate the efficacy of existing society FN management guidelines. However, our sensitivity analysis tested the most subjective components of the risk stratification criteria, and our study outcomes were not significantly changed. Third, our regression model may lack important unmeasured covariates and should be validated before reaching definitive conclusions about its merit. Fourth, nonadherence to treatment guidelines could be a function either of an awareness gap or of patient factors not captured in the chart review. Additionally, we could not reliably ascertain if the management plan was driven by the emergency physician or outpatient oncologist. Fifth, we did not power this study to detect differences in clinical outcomes between risk groups. CONCLUSIONS Treatment of low-risk cancer patients with febrile neutropenia often did not follow national specialty society guidelines. Low-risk patients had a low rate of adverse outcomes but were hospitalized and treated with parenteral antibiotics frequently. Further research is needed to see if this gap between treatment recommendations and patterns of care is more

8 90 Baugh et al. FEBRILE NEUTROPENIA RISK STRATIFICATION widespread and, if so, investigate strategies to better align them. References 1. Greene J. CONCERN for cancer. New National Institutes of Health Network to Focus on Cancer Patients in the Emergency Department. Ann Emerg Med 2015;66:A13 A Freifeld AG, Bow EJ, Sepkowitz KA, et al. Clinical practice guideline for the use of antimicrobial agents in neutropenic patients with cancer: 2010 update by the Infectious Diseases Society of America. Clin Infect Dis 2011;52:e Flowers CR, Seidenfeld J, Bow EJ, et al. Antimicrobial prophylaxis and outpatient management of fever and neutropenia in adults treated for malignancy: American Society of Clinical Oncology clinical practice guideline. J Clin Oncol 2013;31: Klastersky J, Paesmans M, Rubenstein EB, et al. The Multinational Association for Supportive Care in Cancer risk index: a multinational scoring system for identifying low-risk febrile neutropenic cancer patients. J Clin Oncol 2000;18: Baugh CW, Brooks GA, Reust AC, et al. Provider familiarity with specialty society recommendations for risk stratification and management of patients with febrile neutropenia. New England Regional SAEM Meeting, Worcester, MA, Owolabi DK, Rowland R, King L, et al. A comparison of ED and direct admission care of cancer patients with febrile neutropenia. Am J Emerg Med 2015;33: Gilbert EH, Lowenstein SR, Koziol-McLain J, Barta DC, Steiner J. Chart reviews in emergency medicine research: where are the methods? Ann Emerg Med 1996; 27: Kaji AH, Schriger D, Green S. Looking through the retrospectoscope: reducing bias in emergency medicine chart review studies. Ann Emerg Med 2014;64: Hall KK, Lyman JA. Updated review of blood culture contamination. Clin Microbiol Rev 2006;19: Weinstein MP. Blood culture contamination: persisting problems and partial progress. J Clin Microbiol 2003;41: Mirrett S, Weinstein MP, Reimer LG, Wilson ML, Reller LB. Relevance of the number of positive bottles in determining clinical significance of coagulase-negative staphylococci in blood cultures. J Clin Microbiol 2001;39: Libuit J, Whitman A, Wolfe R, Washington CS. Empiric vancomycin use in febrile neutropenic oncology patients. Open Forum Infect Dis 2014;1:ofu Klastersky J, Paesmans M, Georgala A, et al. Outpatient oral antibiotics for febrile neutropenic cancer patients using a score predictive for complications. J Clin Oncol 2006;24: Sharma PC, Jain A, Jain S. Fluoroquinolone antibacterials: a review on chemistry, microbiology and therapeutic prospects. Acta Pol Pharm 2009;66: Belforti RK, Lagu T, Haessler S, et al. Association between initial route of fluoroquinolone administration and outcomes in patients hospitalized for communityacquired pneumonia, Clin Infect Dis 2016; 63: LaMantia MA, Platts-Mills TF, Biese K, et al. Predicting hospital admission and returns to the emergency department for elderly patients. Acad Emerg Med 2010;17: Wallace E, Stuart E, Vaughan N, Bennett K, Fahey T, Smith SM. Risk prediction models to predict emergency hospital admission in community-dwelling adults: a systematic review. Med Care 2014;52: DiazGranados CA, Jernigan JA. Impact of vancomycin resistance on mortality among patients with neutropenia and enterococcal bloodstream infection. J Infect Dis 2005;191: Davey P, Brown E, Charani E, et al. Interventions to improve antibiotic prescribing practices for hospital inpatients. Cochrane Database Syst Rev 2013; 4: CD Bitar RA. Utility of the Multinational Association for Supportive Care in Cancer (MASCC) Risk Index Score as a Criterion for Nonadmission in Febrile Neutropenic Patients with Solid Tumors. Perm J 2015;19: Green SM, Schriger DL, Yealy DM. Methodologic standards for interpreting clinical decision rules in emergency medicine: 2014 update. Ann Emerg Med 2014;64: Carmona-Bayonas A, Jimenez-Fonseca P, Virizuela Echaburu J, et al. Prediction of serious complications in patients with seemingly stable febrile neutropenia: validation of the Clinical Index of Stable Febrile Neutropenia in a prospective cohort of patients from the FINITE study. J Clin Oncol 2015;33: Mace SE, Graff L, Mikhail M, Ross M. A national survey of observation units in the United States. Am J Emerg Med 2003;21: Brown J, Grudzen C, Kyriacou DN, et al. The emergency care of patients with cancer: setting the research agenda. Ann Emerg Med 2016 [Epub ahead of print]. Supporting Information The following supporting information is available in the online version of this paper: Data Supplement S1. Infectious Diseases Society of America Clinical Practice Guideline: Additional Specific Clinical Criteria that Exclude Oncology Patients with FN from Initial Outpatient Care even with a MASCC Score 21.

9 ACADEMIC EMERGENCY MEDICINE January 2017, Vol. 24, No Data Supplement S2. American Society of Clinical Oncology Clinical Practice Guideline: Additional Specific Clinical Criteria* that Exclude Oncology Patients with FN from Initial Outpatient Care even with a MASCC Score 21.