Febrile neutropenia (FN) is a perilous complication seen among patients with cancer treated with chemotherapy or hematopoietic stem cell transplantation and, as such, a common oncologic emergency encountered in emergency departments (EDs). Fever in patients with neutropenia can be life-threatening, with serious bacterial illness being a significant cause of morbidity and mortality for these patients [[1], [2], [3]]. FN occurs in 10–50 % of patients with solid tumors and over 80 % of those with blood malignancies [4]. Mortality rates for patients presenting with neutropenic fever can range from 2.6 to 50.6 % [5].

Empiric administration of antibiotics without real-time knowledge of the cause of infection has lowered the incidence of death and complication in this patient population [[6], [7], [8], [9]]. The selection of appropriate antibiotics is based on the patient's immunocompromised state, the suspected invading organisms, the source of infection, and their susceptibility to antibiotics [10]. During the care of FN patients, diagnostic testing is performed to determine whether a treatable infection is present. The gold standard for this process is collecting and processing two sets of blood cultures from separate venipuncture sites if no central catheter is present; however, blood culture lacks sensitivity and does not provide rapid results [2,11]. Bacteremia is reported in only 10 to 25 % of FN cases, while clinically documented infections are documented only 20 to 30 % of the time [2]. Nonetheless, since infection can rapidly progress and become life-threatening for neutropenic patients, clinical guidelines recommend broad-spectrum antibiotics within 60 min of presentation [7,8,[12], [13], [14], [15]].

There are 361,456 annual US ED visits by cancer patients with FN [16]. Prior work has demonstrated that virtually all are admitted despite recommendations to risk stratify and discharge low-risk patients home on oral antibiotics [2,6,17,18]. This behavior is typical because the available risk stratification tools, such as the Multinational Association for Supportive Care in Cancer (MASCC) score and the Clinical Index of Stable Febrile Neutropenia (CISNE) score, are not only unfamiliar to emergency physicians (EPs) but also insufficiently sensitive for identifying patients requiring care escalation, such as intensive care unit (ICU)-level care [2,7,[19], [20], [21]]. Finding an improved, quick, and accurate method to identify a subgroup of FN patients safe for discharge home on oral antibiotics would protect these immunocompromised and vulnerable patients from the hospital environment, with its attendant cost problems, crowding, nosocomial infections, and medical errors.

Inflammatix Severity classifier version 3b (IMX-SEV-3b) reads the results of expression of host mRNA transcripts and interprets the risk for ICU-level care and 30-day mortality as previously described using a machine-learning algorithm (Inflammatix, Sunnyvale, CA). The assay has been reported to identify patients with risk for clinical deterioration within 7 days and 30 day mortality [[22], [23], [24], [25]]. We studied a sample of FN patients to evaluate the classifier's potential to predict the risk for clinical deterioration and mortality compared to clinical scores. In the future, this line of investigation may help to identify a subgroup that could safely be discharged on oral antibiotics.