Outcomes and disposition of oncology patients with non-neutropenic fever and positive blood cultures

Abstract

Children with cancer and non-neutropenic fever (NNF) episodes are often treated as outpatients if they appear well. However, a small subset have bloodstream infections (BSIs) and must return for further evaluation. These patients may be directly admitted to inpatient units, whereas others are first evaluated in outpatient settings before admission. The best practice for securing care for patients discovered to have outpatient bacteremia are unclear. To determine outcomes and compare time to antibiotics between the two disposition, we retrospectively reviewed all NNF initially treated as outpatients and later had positive blood cultures from 2012 – 2016. Of 845 NNF cases initially treated in outpatient settings, 48 episodes (n = 43 patients) had BSIs. Of those, 77.1% (n = 37) were re-evaluated as outpatients and admitted; 14.6% (n = 7) were direct admissions. The median time to antibiotic did not significantly differ between outpatient re-evaluations (119 minutes) and direct admissions (191 minutes), P = .11. One patient met sepsis criteria upon return and required intensive care unit admission for vasopressor support. No patient died within 1 week of the febrile episode. Most patients with NNF and BSIs initially discharged are stable upon return. Institutions should evaluate their patient flows to ensure that patients receive timely care.

1 |. Introduction

Non-neutropenic fever episodes, defined as a single oral temperature ≥ 38.3°C or ≥ 38.0°C for longer than 1 hour and an absolute neutrophil counts (ANC) ≥ 500 cells/mm³, are common in pediatric patients with cancer.^1,2 Blood stream infections (BSIs) occurred at a pooled rate of 8.6% of non-neutropenic fever episodes in children with cancer.² These patients are at higher risk of BSIs compared to general pediatric population because most have indwelling central venous catheters.^2–7 In contrast with guidelines for fever associated with neutropenia,^8,9 no evidence-based guidelines are available for the management of non-neutropenic fever.² Many pediatric cancer institutions administer ceftriaxone because of the risk of BSIs and observe well-appearing cases as outpatients.^2–6,10 In contrast, some centers do not routinely administer empiric antibiotics and reported no serious long-term complications from sepsis with this approach, despite some patients later developed positive blood cultures.^6,10 Previous studies have reported admission rates after initial episodes of non-neutropenic fevers of 24.4% to 28.1%, indicating that up to 76.1% of patients with cancer and non-neutropenic fever are monitored as outpatients.^3,4,10

Several efforts have identified risk factors associated with positive blood cultures in non-neutropenic fever episodes in pediatric patients with cancer.^2–7,11 However, patients with BSIs can appear well, and positive blood cultures are reported in 1.5% to 6.3% of non-neutropenic fever episodes initially managed in outpatient settings.^3–5,10 After positive blood culture results, patients are typically instructed to return to the hospital for further evaluation and are admitted for intravenous antibiotics, unless a contaminated blood culture specimen is strongly suspected.² Patients with BSIs may progress to severe sepsis without appropriate antibiotic treatment. Because prompt evaluation and antibiotic administration is crucial to optimize sepsis management, ¹² a system that allows immediate notification of positive blood cultures and rapid evaluation by health care providers is recommended.^13,14 Although previous studies have reported no major serious outcomes with outpatient observations,^3–5,10 the clinical presentations and outcomes of outpatients subsequently found to have positive cultures have not been explored in detail.

Historically, the disposition of patients with non-neutropenic fever who later have positive blood cultures after discharge to outpatient settings have varied according to clinician preferences. Some clinicians instruct patients to come back to outpatient settings for evaluation and initiation of intravenous antibiotics, whereas others directly admit patients to inpatient units. It is unclear which practice is the most efficient for minimizing the time from presentation to evaluation and treatment. Controlling variation is known to improve the quality of medical care.¹⁵ Therefore, a quality improvement team was formed at St. Jude Children’s Research Hospital (St. Jude) to standardize the management of these episodes at our institution. Understanding clinical presentation and outcomes upon return will help guide optimal management strategies for this unique patient population. We performed this retrospective review to guide our quality improvement effort.

2 |. Materials and Methods

2.1 |. Patients

All outpatient encounters of non-neutropenic fever with positive blood cultures in pediatric oncology patients treated at St. Jude from January 2012 through December 2016 were captured from our hospital electronic database. These positive blood cultures were obtained in patients presenting with non-neutropenic fever who were subsequently discharged after initial fever evaluation. Fever was defined as a single episode of oral temperature ≥ 38.3°C or ≥ 38.0°C for longer than 1 hour. Non-neutropenic episodes were defined as an absolute neutrophil count (ANC) ≥ 500 cells/mm³.^8,9 Patients that were initially seen or later treated at outside facilities were excluded. Positive blood cultures for common commensals are deemed “probable contaminant” if the organisms did not grow on two or more specimens collected on separate occasions and there were more than one specimens collected prior to antibiotics, and “possible contaminant” for cases that only had one blood culture sample obtained prior to antibiotic administration.¹⁶ Patient demographic data, time of initial blood culture, and positive organisms were abstracted from electronic database.

The retrospective medical record review was performed to determine the disposition upon return in patients with positive blood cultures (treated in outpatient settings before admission vs. directly admitted). In addition, central line type, vital signs upon return, antibiotic choice, time to antibiotics, and admission to the intensive care unit (ICU) within 24 hours were collected. The time to antibiotics was measured from the time point when clinicians were notified of positive blood cultures to the time that the first antibiotics were administered. Hypotension, tachycardia, and tachypnea were defined according to established age-specific values for pediatric vital signs.^17–20 This study was classified by the Institutional Review Board of St. Jude as a quality improvement project, and the requirement for informed consent was waived.

2.2 |. Current process

St. Jude is a free-standing quaternary academic pediatric hematology oncology center that treats approximately 500 new oncology patients annually. The hospital does not have a formal emergency room. At the time of the study, patients with unscheduled visits were seen at their primary clinics during the day, and seen in an after-hours urgent care clinic staffed by pediatric hematology oncology hospitalists or fellows at night and on the weekends. Pediatric patients with cancer and non-neutropenic fever are typically treated in outpatient settings, usually with a single dose of ceftriaxone if they appear well and are not anticipated to become neutropenic from recent chemotherapy. Blood cultures are routinely obtained at initial evaluations. When blood cultures are positive, the microbiology laboratory notifies the primary clinic or on-call clinicians. Clinicians then contact patients with instructions to promptly return for additional evaluation and therapy, including repeat blood cultures, administration of intravenous antibiotics, and hospital admission. The disposition of these patients is dependent on clinician preference. Some clinicians routinely evaluate patients in an outpatient setting before admission, whereas others perform informal screenings of well-being via telephone and directly admit patients to inpatient units, if necessary.

2.3 |. Statistical analysis

Fisher exact tests were used to compare categorical data. Mann-Whitney U tests were used to compare non-normally distributed continuous data. P values < 0.05 were considered statistically significant. All analyses were performed with Prism software, v8.1 (GraphPad Software, San Diego, CA, USA).

3 |. Results

3.1 |. Patients

During 2012 through 2016, 845 non-neutropenic fever episodes in pediatric patients with cancer were treated in outpatient settings, of which 43 patients had 48 episodes (5.7%) of possible BSIs. Five patients initially treated or later seen at outside facilities were excluded. Table 1 depicts the patient demographic characteristics.

TABLE 1.

Demographic and clinical characteristics

Characteristics	Results (n = 48)
Age (years)	3.78 (0.73–23.64)
Sex (%)
Male	30 (62.5)
Female	18 (37.5)
Diagnosis (%)
Acute lymphoblastic leukemia	9 (18.8)
Acute myeloblastic leukemia/myelodysplastic syndrome	4 (8.3)
Lymphoma	3 (6.3)
Sarcoma	9 (18.8)
Neuroblastoma	9 (18.8)
Retinoblastoma	5 (10.4)
Wilms tumor	1 (2.0)
Central nervous system tumors	3 (6.3)
Langerhans cell histiocytosis	3 (6.3)
Hemophagocytic lymphohistiocytosis	2 (4.2)
Type of central line (%)
Subcutaneous port	16 (33.3)
External catheter (tunneled and peripherally inserted central catheter)	28 (58.3)
No central line	4 (8.3)
HSCT, < 100 days from allogeneic HSCT (%)	4 (8.3)
ANC (cells/mm3)	3950 (500–22 000)
ANC, 500–1000 cells/mm3 (%)	6 (12.5)
APC (cells/mm3)	4461.5 (550–23 661)

Abbreviations: ANC, absolute neutrophil count); APC, absolute phagocyte count; HSCT, hematopoietic stem cell transplantation; TMax, maximum temperature.

Note. Continuous variables are reported in median and range. Categorical variables are reported in frequency and percentage.

3.2 |. Non-neutropenic fever episodes

The median highest temperature at initial non-neutropenic fever evaluation was 38.9°C (range, 38.1°C –39.7°C). The median ANC and absolute phagocyte count (APC) were 3950 cells/mm³ (range, 500–22,000 cells/mm³) and 4461.5 cells/mm3 (range, 550–23,661 cells/mm³), respectively (Table 1). A total of 52 isolates were obtained from the 48 episodes, and three episodes had multiple organisms isolated (Table 2). Of the 52 isolates, 38 were gram-positive bacteria (73.1%); 13 were gram-negative bacteria (25%); and one was a yeast (1.9%). Twelve isolates (23.1%) were deemed likely contaminant, all of which were gram-positive bacteria. Of those, 4 isolates (7.7%) were possible contaminants, and 8 (15.4%) were probable contaminants. The median time to detection was 12.79 hours (Interquartile range (IQR), 10.46 – 17.36 hours; range, 7.53–188.3 hours) for the whole cohort. The median time to detection was longer in cases that were deemed likely contaminant (23.2 hours; IQR, 19.1 – 29.07 hours; range, 15.89 – 188.3 hours) compared to cases that are proven BSIs (11.47 hours; IQR 9.85 – 13.4 hours; range, 7.53 – 46.19 hours). This difference was statistically significant (P < 0.001).Of the 49 isolates tested for ceftriaxone sensitivity, 19 (38.7%) were sensitive. For levofloxacin, 24 of 36 isolates tested (66.7%) were sensitive.

TABLE 2.

List of organisms isolated from blood cultures

Organisms (number of isolates)	Proven BSI n (%)	Sensitive to ceftriaxone n (%)	Sensitive to fluoroquinolones n (%)
Gram positive organisms
Coagulase-negative Staphylococcus (1)	0	0	1 (100%)
Enterococcus faecalis (1)	1 (100%)	0	0
Enterococcus faecium (1)	1 (100%)	0	Not tested
Gram-positive bacilli, suggestive of diptheroids (1)	0	Not tested	Not tested
Large spore-forming, aerobic Gram-positive bacilli (1)	0	0	Not tested
Micrococcus luteus (3)	0	1 (33.3%)	Not tested
Staphylococcus aureus* (2)	2 (100%)	0	2 (100%)
Staphylococcus epidermidis† (17)	13 (76.5%)	2 (11.8%)	10 (58.8%)
Staphylococcus haemolyticus‡ (2)	1 (50%)	1 (50%)	2 (100%)
Streptococcus agalactiae* (1)	1(100%)	1 (100%)	1 (100%)
Streptococcus pneumoniae (5)	5 (100%)	5 (100%)	Not tested
Streptococcus viridans group (3)	2 (66.7%)	0	Not tested
Gram negative organisms
Brevundimonas species (1)	1 (100%)	0	1 (100%)
Citrobacter freundii‡ (1)	1 (100%)	1 (100%)	1 (100%)
Enterobacter cloacae (2)	2 (100%)	1 (50%)	2 (100%)
Escherichia coli (2)	2 (100%)	0	0
Haemophilus influenzae (2)	2 (100%)	2 (100%)	Not tested
Klebsiella pneumoniae (1)	1 (100%)	1 (100%)	1 (100%)
Pantoea species (1)	1 (100%)	1 (100%)	1 (100%)
Proteus mirabilis‡ (1)	1 (100%)	1 (100%)	1 (100%)
Providencia stuartii (1)	1 (100%)	1 (100%)	0
Serratia marcescens (1)	1 (100%)	1 (100%)	1 (100%)
Fungi
Candida albicans† (1)	1 (100%)	0	0

Abbreviations: BSI, bloodstream infection

^*Multiple organisms from the same non-neutropenic fever episode.

^†Multiple organisms from the same non-neutropenic fever episode.

^‡Multiple organisms from the same non-neutropenic fever episode.

3.3 |. Management of positive blood cultures

Upon notification of positive blood culture results, 37 patients were evaluated as outpatients before admission (77.1%); seven were directly admitted (14.6%); three received antibiotics as outpatients (6.3%); and one was observed as an outpatient without antibiotics (2.1%). We found no difference between age, type of catheter, history of hematopoietic stem cell transplant, presenting temperature, ANC, APC, or time to detection between patients who were treated as outpatients before admission and those who were directly admitted to inpatient units (Table 3). Patients (n = 4, 8.3%) who were treated exclusively as outpatients were infrequent and had unique clinical circumstances associated with the decision to treat them as outpatients (Supplemental Table S1, Supplemental Digital Content 1, http://links.lww.com/JPHO/A388). Of the 12 patients whose positive blood cultures were likely contaminants, 8 (66.7%) were treated as presumed bacteremia upon return and admitted for intravenous antibiotics. The antibiotics given upon return were determined according to clinical presentation, past history of BSIs or colonization, and initial gram stain results. Provider’s antibiotic choices were most commonly consisted of a combination of cefepime and vancomycin (n = 15, 31.3%), followed by ceftriaxone and vancomycin (n = 13, 27.1%). One patient whose blood culture was initially positive for budding yeast received amphotericin B in addition to ceftriaxone and vancomycin (Supplemental Table S2, Supplemental Digital Content 2, http://links.lww.com/JPHO/A389).

TABLE 3.

Comparison of clinical characteristics between two disposition groups

Clinical characteristics	Evaluated as outpatients (n = 37)	Directly admitted (n = 7)	P-value
Median age (years)	3.680	3.959	0.6317
External central catheter	22	6	0.393
HSCT (< 100 days from allogeneic HSCT)	3	1	0.5135
Median TMax (°C)	38.80	39.00	0.6195
Median ANC (cells/mm3)	4,400	3.400	0.8938
Median APC (cells/mm3)	6090	4002	0.95
Median time to detection (hours) Median (IQR)	12.14 (10.07 - 15.89)	13.04 (11.47 – 20.19)	0.4496
Time to antibiotics (minutes) Median (IQR)	119 (76-162)	191 (120-267)	0.11

Abbreviations: ANC, absolute neutrophil count; APC, absolute phagocyte count; HSCT, hematopoietic stem cell transplantation; IQR, interquartile range; TMax, maximum temperature

3.4 |. Presentation upon return

From our medical records review, we accurately determined patient vital signs upon return and the time to antibiotic treatment for 38 episodes. Tachycardia was noted in 16 (42.1%) episodes, with 5 episodes of severe tachycardia (13.2%). Two patients had tachypnea (5.3%). Hypotension was noted in one (2.1%) patient. For the whole cohort of 48 episodes, 1 (2.1%) patient required fluid resuscitation, followed by ICU admission for vasopressors, and recovered without long-term complications. No deaths occurred within 7 days. The median time to antibiotic treatment was 119 minutes in the group initially re-evaluated as outpatients and 191 minutes in the group directly admitted (Table 3). This difference was not statistically significant (P = .11). All patients seen at outside facilities were well upon return to St Jude.

The patient that required ICU admission was an 11 year old male with Ewing sarcoma who had non-neutropenic fever associated with subclavian thrombus (diagnosed later), initially received ceftriaxone, and discharged to outpatient settings. Blood culture became positive for Staphylococcus aureus at 9 hours, and vancomycin was given 50 minutes after notification of positive blood culture. Patient developed fluid-resistant hypotension after vancomycin was started and required vasopressors for less than 24 hours and was discharged from ICU on hospital day 2.

4 |. Discussion

Currently, guidelines for the management of non-neutropenic fever are not established for pediatric patients with cancer.² Although these patients are at high risk of BSIs, ^{2–7, 10, 11, 21} well-appearing patients are often observed as outpatients, sometimes without antibiotics.^3–7,10 Esbenshade et al. developed a prediction model for BSIs in patients with non-neutropenic fever, incorporating 12 clinical characteristics and laboratory values.⁴ This model has undergone multisite validation with retrospective data; however, it cannot be reliably used for general clinical practice yet without further prospective validation.¹¹ As a part of a quality improvement project, we explored the outcomes of non-neutropenic fever episodes with positive blood cultures in pediatric patients with cancer who were initially observed as outpatients at our institution. We further studied presentation upon return and the time to antibiotic treatment to understand the current local process and determine the best care option for patients.

The incidence of positive blood cultures in non-neutropenic fever episodes in pediatric patients with cancer who were initially treated as outpatients in this study was similar to previously reported incidences.^3–5,10 Time to positive cultures were longer in cases that were deemed likely contaminated. However, this is not clinically relevant as there was some overlapping between likely contaminated and true BSIs cases. Most patients who were called back to the hospital for positive blood cultures did not require immediate resuscitation upon return. Only one patient with BSI related to subclavian thrombus experienced septic shock requiring prompt resuscitation and ICU admission. The median time to antibiotic treatment was shorter for patients who were initially re-evaluated as outpatients than for patients who were directly admitted to inpatient units, but this difference was not statistically significant. These findings are similar to those of previous reports, with comparable care for outpatient and direct admission settings.²² Because our center does not have an emergency room and most patients reside in hospital-provided housing located close to campus, the findings from this study may not be applicable to other centers with different settings.

Patients with BSIs are at risk of sepsis; therefore, prompt evaluation and antibiotic administration are crucial for these patients. We recommend that centers evaluate time to antibiotic administration in patients with positive blood cultures to evaluate best practices. In our population, the median time to antibiotics were longer in the direct admission group, however the difference was not statistically different. Since most patients did not require a higher level of care, and direct admission has been shown to improve patient flow²³, the quality improvement team plan to optimize the local direct admission process by developing triage and direct admission protocols to ensure safe and timely patient care. Further research is needed to identify optimal management strategies for this at-risk group of children.

The 2018 Surviving Sepsis Campaign initiated by the Society of Critical Care Medicine recommends that patients with sepsis receive antibiotics within 1 hour.¹² In our study, the median time to antibiotic treatment from when clinicians were notified of positive blood cultures was longer than 1 hour for both patients initially treated as outpatients and those directly admitted to inpatient units. Although it is unclear what an ideal time to antibiotic treatment is for patients with BSIs and without apparent sepsis, our findings still identify opportunities for improving the process of contacting patients and families to return for evaluation after notification of positive blood cultures.

The number of ceftriaxone-sensitive isolates in our cohort was comparable to that recently reported by Esbenshade et al.⁴ We also explored fluoroquinolone sensitivity because this class of antibiotics is commonly used for antimicrobial prophylaxis or to treat bacterial infections in immunocompromised hosts, especially when oral administration is required.²⁴ Most of the isolates tested were fluoroquinolone sensitive, indicating that fluoroquinolone is a reasonable alternative when evaluating stable oncology patients with concerns for BSIs.

Our study has certain limitations because of the retrospective nature of the study. First, all records of non-neutropenic fever episodes were obtained electronically by predefined temperature and ANC values. Patients that reported fever at home who were afebrile upon presentation were not captured. Furthermore, vital signs upon return could not be accurately determined in patients who already had clinic visits for other reasons at the time of notification of positive blood culture. The duration for contacting patients and families and for patients to return to our facility was also not documented in medical records. This is an important variable that may affect the time to antibiotic treatment and should be explored further. In addition, the small sample size, particularly for the direct admission group, may affect the power to detect a difference in time to antibiotics between the two groups. Moreover, external factors, such as the relatively short distance of patients from our hospital or other unknown variables, may have biased the decision to send patients for direct admission versus outpatient management. Centers may wish to capture data prospectively as they seek to remove practice variability in this population. Finally, this retrospective review did not compare clinical characteristics between patients that developed BSIs and the larger cohort of patients (N =845) with fever who remained outpatient during the review period. Such an analysis may help guide clinicians seeking to determine the disposition at initial presentation of fever. We plan to perform these comparisons as a part of our ongoing work around treatment guidelines for non-neutropenic fever using a larger data set.

In conclusion, although most pediatric cancer patients with non-neutropenic fever and positive blood cultures who were initially treated as outpatients experienced generally favorable outcomes, some required immediate resuscitation upon return. Our quality improvement team will further standardize the direct admission process at St. Jude by creating a direct admission protocol and a tool for screening patients with sepsis. We recommend that other institutions also prospectively evaluate and standardize their practices. Furthermore, this study also highlights the opportunity for process improvement at our institution, which includes redesigning our direct admission workflow and improving communication with patients and families upon the notification of positive blood cultures.