a b s t r a c t

Context: Febrile neutropenic immunocompromised children are at a high risk of serious bacterial infections (SBI). Objective: This systematic review and meta-analysis report the prevalence of SBI in healthy children with febrile neutropenia.

Data source: PubMed, EMBASE, and Web of Science from their inception to August 2020.

Study selection: Patients with an Absolute Neutrophil count (ANC) <1000 cells/mm³ up to 18 years of age pre- senting to the ED with a chief complaint of fever (temperature > 38?C) and who had a workup for SBI as defined by each study.

Data abstraction: Data from individual studies was abstracted by a subset of the authors and checked indepen- dently by the senior author. Any discrepancies were adjudicated by the joint agreement of all the authors. We calculated the prevalence of SBI by using the number of SBI’s as the numerator and the total number of febrile events in patients as the denominator. Bias in our studies was quantified by the Newcastle Ottawa Scale.

Results: We identified 2066 citations of which five studies (1693 patients) our inclusion criteria. None of our reviewed studies consistently tested every included patient for SBI. Spectrum bias in every study resulted in a wide range of the SBI prevalence of 1.9% (<0.01% - 11%) similar to non-neutropenic children.

Limitations: All of our studies were retrospective and many did not consistently screen all subjects for SBI. Conclusion: If the clinical suspicion is low, the risk for SBI is similar between febrile healthy neutropenic and non- neutropenic children.

(C) 2021

1. Introduction

Febrile neutropenia in a child who is immunocompromised is a medical emergency and management includes initiating empiric antibi- otics upon the onset of fever [1]. Children with a depressed immune sys- tem are at higher risk of serious bacterial infection [2]. In healthy children with isolated febrile neutropenia, studies have shown that the prevalence of SBI is between 2 and 8% [3-6] [7]. Although the data

Abbreviations: SBI, Serious bacterial infections; ANC, absolute neutrophil count; IQR, Interquartile Range; ED, Emergency Department; PRISMA, Preferred Reporting Items for Systematic Review and Meta-analyses; ICC, The intraclass correlation coefficient; CBC, Complete Blood Count; CXR, Chest X-ray; CSF, Cerebral Spinal Fluid; UTI, Urinary Tract Infection.

* Corresponding author at: Department of Emergency Medicine, 450 Clarkson Avenue, Brooklyn, NY 11203, United States of America.

E-mail address: [email protected] (R. Hao).

are limited, viral infections have been considered the most frequent eti- ology in isolated neutropenia in previously healthy children [7]. The lack of a unified consensus for management based on strong clinical evi- dence creates a challenge for the clinician treating these patients.

Identifying the prevalence of SBI and risk factors among healthy chil- dren with isolated febrile neutropenia would allow physicians to adopt a less aggressive approach to management. In healthy febrile children with isolated neutropenia, Empiric treatment with antibiotics is poten- tially harmful [8]. Patients are at risk for Iatrogenic complications which include but are not limited to Allergic reactions, IV infiltrations, risk of exposure to other infectious agents in the hospital, an increase in resis- tance pattern, and an overall rise in Healthcare costs [8,9].

The objective of this systematic review and meta-analysis is to deter- mine the prevalence of SBI in children who are otherwise healthy pre- senting with isolated febrile neutropenia to further guide clinical practice.

https://doi.org/10.1016/j.ajem.2021.02.017

0735-6757/(C) 2021

1. Methods
   1. Study design

We conducted a systematic review of studies that reported the prev- alence of serious bacterial infections in febrile neutropenic otherwise healthy children without other risk factors. The systematic review was conducted using the Preferred Reporting Items for Systematic Review and Meta-analyses (PRISMA) guidelines [10] The protocol for this sys- tematic review can be accessed in PROSPERO with the registration num- ber: CRD42020186862 [11].

• 1. Search strategy

The design and manuscript structure of this systematic review con- forms to the recommendations from the Meta-analysis of Observational Studies in Epidemiology (MOOSE) statement [12]. In conjunction with a medical librarian, we searched the medical literature up to June 2020 in PUBMED, EMBASE, and Web of Science using the search terms “epide- miology, prevalence, bacteremia, neutropenia” (Appendix-1). We also searched the following online databases to avoid missing relevant un- published articles and abstracts: Clinicaltrials.gov, open grey, OpenDOAR, BASE, WorldWideScience.org, Mednar, and HSO. We also completed a hand search of references of included studies. We limited studies to the English language only.

Data from individual studies were abstracted by RH, MS, TL, NM, IG, CA and checked independently by RS. Any discrepancies were adjudi- cated by the joint agreement of all the authors.

• 1. Patients

We included studies that examined patients with moderate and se- vere neutropenia(ANC <1000 cells/m³)up to 18 years of age presenting to the ED with a chief complaint of fever (temperature > 38.0C) and those who had a workup for Serious Bacterial Infection as defined by each study.

• 1. Laboratory test

Laboratory tests for bacteremia, meningitis, urinary tract infection included cultures of blood, CSF, and urine, which were collected before starting antibiotics as empiric treatment. We included studies with all types of cultures with no restriction with respect to the time the culture results were reported.

• 1. Outcomes

Serious Bacterial Infections were defined by each study, which meant the growth of a specific organism after a defined period of time.

• 1. Quality assessment

Table 2 indicates the results of each individual study quality assess- ment according to the Newcastle-Ottawa scale [13]. The included stud- ies were not cohort-studies in design and therefore not every category and/or item included in the scale is applicable, but we have summarized the results of the applicable categories of Selection and Outcome. Two reviewers (TL, CA) rated the study quality according to the checklist and their responses were analyzed to quantify interrater reliability. A third reviewer was available in case there was a discrepancy in the se- lection of the articles. intraclass correlation coefficient .

• 1. Data analysis

Data were reported as percentages with 95% Confidence Intervals (95% CI) and medians with InterQuartile Range (IQR 25, 75%). We

calculated the prevalence of SBI by using the number of SBI’s as the numerator and the total number of febrile events in patients with neu- tropenia as the denominator. Heterogeneity between prevalence esti- mates was assessed using the I² statistic, which describes the percentage of variation not because of sampling error across studies. An I² value above 75% indicates high heterogeneity [15].

We conducted the meta-analysis with prevalence estimates that had been transformed using the double arcsine method [16]. The final pooled result and 95% CIs were back-transformed for ease of interpreta- tion. Meta-analysis was undertaken using a random-effects model (to account for heterogeneity) conducted using the MetaXL (www. epigear.com) add-in for Microsoft Excel. A pooled prevalence figure was calculated with 95% CI. Statistical package: IBM Corp. Released in 2011. IBM SPSS Statistics for Windows, Version 20.0. Armonk, NY: IBM Corp.

1. Results
   1. Selection of the included studies

The PUBMED, EMBASE, and Web of Science searches identified 2066 citations (Fig. 1). 9 articles were evaluated for full-text review. The pa- pers written by Husain, Alexandrapoulou, and Karavanaki were ex- cluded after the whole text review because it included patients with mild neutropenia (ANC <1500, ^{17, 18, 19]}. The paper written by Serwint was excluded because some patients in the study were diagnosed with an oncologic etiology for the neutropenia [20]. We included in our review 5 articles [3-5] [6,7]. All 5 reviewed studies used a retrospec- tive design with data extracted from medical records. In addition, Barg et al. conducted a case-control study with a matched group of non- neutropenic children; we did not review the control study, just their data from their cases.

Fig. 1. Study Selection process.

• 1. Description of included studies

The five studies [3,4] [5-7] in our review included a total of 1693 individual patients (Table 1). Inclusion criteria of having a fever (>38C) was required for four studies [3,4,6,7] and the study by Melendez et al. did not explicitly require a fever for subject inclusion. As stated in the paper by Melendez et al., patients were selected if a complete blood count was sent from their ED with an ANC

<1000 cells/mm³. It does not appear that all subjects in the study by Melendez et al. had a fever. Between 1995 and 2000, many but not all of Melendez’s subjects had CBC’s as part of a febrile workup of chil- dren <36 months. After 2000, a CBC was recommended in Melendez’s cohort only for febrile young children if they did not receive three doses of conjugate pneumococcal vaccine. Melendez et al. does not provide the exact number of febrile as opposed to non-febrile neutro- penic children in their sample.

This difference in sampling by Melendez et al. may explain why their study had a significantly larger size (n = 1317) compared to the range (n = 52, ^3] to (n = 190, ^7] of sample sizes in the other four studies which required a fever for study inclusion. Though the paper says 1888 pa- tients were evaluated, only 1317 had at least one culture sent [5].

Age as an inclusion criterion varied widely among the studies from an upper limit of 21 years [5] to only 36 months [3]. Lower age limits were from birth [5,7] to 1.5 months [3]. Age as inclusion was not speci- fied by Bonadio et al. [4].

Neutropenia was defined similarly by four studies [4-7] as

ANC < 1000 cells/mm³. Wittman et al. used a lower limit of ANC < 500 cells /mm³ as their inclusion criteria.

All of our studies generally excluded patients because of comorbid immunosuppressive diseases. Three of our studies also excluded pa- tients recently on antibiotics [3,4,6].

Finally, the definitions of Serious Bacterial Infections (SBI) varied widely across our studies. Besides the study by Bonadio et al., which

did not clearly define SBI, the other four studies included bacteremia, meningitis, and urinary tract infections in their definition of SBI. In addi- tion, Barg et al. and Wittman et al. included Septic arthritis, mastoiditis, bacterial gastroenteritis, and osteomyelitis in their definitions of SBI.

• 1. Quality assessment

Table 2 indicates the results of each individual study quality assess- ment according to the Newcastle-Ottawa scale. Of the included studies, all were not cohort-studies in design except Melendez et al., and there- fore not every category and/or item included in the scale is applicable, but we have summarized the results of the applicable categories of Se- lection and Outcome. Two reviewers (TL, CA) rated the study quality ac- cording to the checklist and their responses were analyzed to quantify interrater reliability. The Intraclass Correlation Coefficient (ICC) was 1.0.

All our reviewed studies were retrospective chart reviews. Gilbert

et al. defined eight criteria for retrospective chart reviews to improve accuracy and minimize inconsistencies in data acquisition: 1) training,

2) case selection, 3) definition of variables, 4) abstraction forms,

5) meetings, 6) monitoring, 7) blinding, and 8) testing of Interrater agreement. None of our studies utilized these precautions to prevent in- accuracies and potential bias in their reviews of their subjects’ medical records [14].

• 1. Prevalence of serious bacterial infections

In reviewing the prevalence of SBI across our reviewed papers, we discovered as a consequence of their retrospective design and lack of pre-study testing protocols, none of our studies had all their subjects simultaneously cultured for bacteremia, meningitis, urinary tract in- fection (UTI), or had chest X-rays (CXR) to evaluate for pneumonia. The wide variety of testing in febrile neutropenic children in our

Table 1

Description of reviewed studies.

Study Design Population Definition of SBI

Bonadio

et al.1989

Melendez et al.2010

Barg

et al.2015

Pascual

et al.2016

Wittman

et al.2017

Retrospective chart review

Retrospective Cohort

Retrospective Case-control

RetrospectiveChart review

RetrospectiveChart review

Inclusion:Fever >38?CAge: Not specifiedNeutropenia

• ANC < 1000 cells / microliter

Exclusion:Antibiotic use within 72 h of evaluation with negative culturesUnderlying malignancy or chronic systemic disease.Sample Size:N = 63Median Age: N/A

Inclusion:Fever: Not requiredAge: < 21 years oldNeutropenia

• ANC < 1000 cells / microliter

Exclusion:Patients with increased risk of SBI:Known underlying immunosuppressionchemotherapy-induced neutropeniaNewly diagnosed, suspected, or known malignanciesBone marrow failure of any etiology.CVC or any implantable deviceCongenital heart condition or genitalurinary abnormalities with increased

risk for bacterial infectionsSample Size:N = 1317Median Age:8.4 months

Inclusion:Fever >38?CAge: 3 months-18 years oldNeutropenia

• ANC < 1000 cells/ul

Well-appearingCBC and blood cultureExclusion:Known underlying immunosuppressive conditionAntibiotic use within 48 h.Known history of neutropeniaPancytopeniaSample Size:N = 71Median

Age:14 months

Inclusion:Fever >38?CAge: <18 years oldNeutropenia

• ANC < 1000 cells / microliter

Exclusion:Known underlying immunosuppressive conditionSample Size:N = 190Median Age:8.5 months

Inclusion:Fever >38?CAge: 1.5 months - 36 monthsNeutropenia

• ANC < 500 cells / microliter

Exclusion:Chronic condition with increased bacterial Infection riskAntibiotics within 48 hSevere anemia or thrombocytopeniaSample Size:N = 52Median Age:8 months

Not defined

BacteremiaUrinary Tract InfectionMeningitis

BacteremiaUrinary Tract InfectionPneumoniaMeningitisSeptic ArthritisBacterial GastroenteritisMastoiditisOsteomyelitis

BacteremiaUrinary tract infectionPneumoniaMeningitisBacterial SynovitisBacterial Pleural Effusion

BacteremiaUrinary tract infectionPneumoniaMeningitisSeptic ArthritisSepticemiaAbscessBacterial GastroenteritisAcute MastoiditisLymphadenitisOsteomyeolitis

Abbreviations: SBI (serious bacterial infection), ?C (degrees Celsius), ANC , N/A (not available), CVC (central venous catheter).

Table 2

Newcastle-ottawa rating scale.

Selection

Comparability

Outcome

Studies

Representativeness of exposed cohort

Selection of non-exposed cohort

Ascertainment of exposure

Outcome of interest not present at the start

Comparability of cohorts

Assessment of outcome

Follow-up

Adequacy of follow up of cohorts

Bonadio et al.

A*

N/A

A*

A*

N/A

A*

A*

A*

1989 Melendez et al. 2010

A*

N/A

A*

A*

N/A

A*

A*

A*

Barg et al. 2015

A*

N/A

A*

A*

N/A

A*

A*

A*

Pascual et al.

A*

N/A

A*

A*

N/A

A*

A*

A*

2016

Wittmann et al. 2017

Selection:

A* N/A A* A* N/A A* A* A*

1. Representativeness of the exposed cohort-A = truly representative of the average number of febrile children in the community. B = somewhat representative of the average number of febrile patients in the community. C = selected group of users eg nurses, volunteers D = no description of the derivation of the cohort.
2. Selection of the non-exposed cohort A = drawn from the same community as the exposed cohort. B = drawn from a different source. C = no description of the derivation of the non- exposed cohort.
3. Ascertainment of exposure A = secure record (eg surgical records) B = structured interview C = written self report D = no description.
4. Demonstration that outcome of interest was not present at start of study A = yes* B = no. Comparability.
5. Comparability of cohorts on the basis of the design or analysis a = age B = study controls for any additional factor (geographic area). Outcome.
6. Assessment of outcome a) independent blind assessment* B = record linkage* C = self report D = no description.
7. Was follow-up long enough for outcomes to occur A = yes (select an adequate follow up period for outcome of interest) B = no.
8. Adequacy of follow up of cohorts a) complete follow up - all subjects accounted for b) Small number of subjects lost to follow up unlikely to introduce bias c) No description of subjects lost d) No statement on loss to follow up.

studies reflects individual hospital management patterns at the time of their study.

From Table 3, we have documented the actual number of cultures obtained from blood, cerebrospinal fluid (CSF), urine, and the number of CXRs for each study. Blood cultures were the most consistently cul- tured source and were obtained in 100% of subjects for Bonadio et al., Barg et al., and Wittman et al., as compared to only 76% of Pascual’s sub- jects and 95% of Melendez’s subjects. urine cultures were also obtained in all our studies ranging from 18% [6] to 61% [7]. CSF cultures were only obtained by Bonadio et al. (48%), Melendez et al. (25%), and Pascual et al. (10%). Chest X-rays were the least commonly ordered test only by Wittman et al. (42%) and Barg et al. (55%).

We reproduced in Table 3, the prevalence of SBI as reported by all of our studies in the fourth column, which varied from 1.9% [3] to 8.5% [6]. The reported SBI prevalence in each study represents the number of pos- itive cultures or cases of pneumonia on CXR divided by the total sample size of each study. Since not all patients received the full complement of cultures or CXR to rule-out the source of infection, the true SBI prevalence in each study is suspect. To give a more granular view of each study’s prevalence of individual elements of their definition of SBI we recalculated, in column 5 of Table 3, the culture results of each study’s blood, CSF, and Urine tests, and CXR. Since the denominator of each cul- ture result is generally smaller than the study’s sample size the confidence intervals are mostly much larger than the stated SBI prevalence for each

Table 3

Comparison of prevalence of serious bacterial infection by study and by cultured specimens.

Study	Sample size	Cultured specimens	SBI prevalence by study (n, %, 95%CI)	SBI prevalence by sensitivity analysis (n, %, 95%CI)
Bonadio et al. 1989	n = 63	BC (n = 63)	5, 8% (3% -18%)	2, 3% (0.2% - 11.5%)
		CSF (n = 30)		3, 10% (3% - 26%)
Melendez et al. 2010	n = 1317	U C&S (n = 33) CXR (n =?) BC (n = 1254)	31, 2.4% (1.6% - 3.3%)	0 ? 8, 0.6% (0.3% - 1.3%)
		CSF (n = 325)		4, 1.2% (0.3% - 3.1%)
Barg et al. 2015	n = 71	U C&S (n = 676) CXR (n =?) BC (n = 71)	6, 8.5% (4% - 18%)	23, 3.4% (2.1% - 5.1%) ? 0
		CSF (n =?) U C&S (n = 13) CXR (n = 39)		? 1, 8% (<0.01% - 35%) 5,13% (5% - 27%)
Pascual et al. 2016	n = 190	BC (n = 145)	4, 2.1% (0.6% - 5.5%)	0
Wittman et al. 2017	n = 52	CSF (n = 19) U C&S (n = 116) CXR (n =?) BC (n = 52) CSF (n =?) U C&S (n = 19) CXR (n = 22)	1, 1.9% (<0.01% - 11%)	0 2, 1.2% (<0.01% - 6.5%) 2, 1.2% (<0.01% - 6.5%) 0 ? 0 1, 5% (<0.01% - 24%)

Abbreviations: Serious bacterial infection , BC (blood culture), CSF (cerebrospinal fluid), U C&S (urine culture & urinalysis), CXR (chest XRay), CI (confidence interval). Melendez:

• positive urine culture >=1000 Colony Forming Units suprapubic, >10,000 catheterized, >50,000 clean void. No mention of urinalysis results
• 31 patients with SBI, total of 35 positive cultures

study. We had questions about the biases in determining which subjects were tested for SBI in each study so we decided to forgo our planned meta-analysis of the prevalence of SBI across our studies.

1. Discussion

Although we planned to do a meta-analysis for a pooled estimate of the prevalence of SBI in healthy appearing febrile neutropenic children, our literature search failed to find any article(s) that screened all their subjects with a consistent set of test modalities for SBI (blood cultures, CSF cultures, CXRs, and urine cultures). When we recalculated these stud- ies’ prevalence using only those patients with definite cultures or CXRs, we found their data were incalculable or that the confidence intervals for SBI were much larger than was stated in their respective studies [3-7]. We can infer that in all these retrospective studies clinical judgment must have been used by the subjects’ primary caregivers to choose which test modality was most likely to diagnose their source of fever. This we feel explains why blood culture, CSF culture, urine culture, and CXR were ordered as the febrile workup only in select patients. The number of patients who received blood cultures were 66% [5], 75% [7], 92% [4], and 100% [3,6]. Lumbar punctures for CSF cultures

were obtained in 0% [3,6] 17% [5], 19% [7], and 44% [4] of patients.

Urine cultures were ordered in 18% [6], 36% [3,5], 49% [4], and 61% [7]

of patients. CXR were obtained in 0% [4,5,7], 42% [3], and 55% [6] of pa- tients. This represents a significant risk of spectrum bias in each of our 5 studies, which falsely increases the sensitivity and specificity of these tests [21,22]. Spectrum bias was also responsible for the wide SBI prev- alence which varied from 1.9% [3] to 8.5% [6] among our studies [3-7].

Still, blood cultures and urine cultures were commonly ordered in these previously healthy neutropenic patients, and the prevalence of these cultures is similar to that of febrile immunocompetent children [23,24]. The reported prevalence of occult bacteremia in febrile, previ- ously immunocompetent healthy children is 1.9% [23]. As can be seen in Table 3, a similar prevalence of bacteremia is seen in febrile neutropenic patients in our reviewed studies, ranging from 0% [3,6,7], 0.6% [5], to 3% [4]. The prevalence of UTI in both populations is also similar; in immuno- competent children, it is 7% [24]. In our reviewed studies, UTI prevalence ranged from 0% [3,4], 1.2% [7], 3.4% [5], to 8% [6]. However, there again is the spectrum bias inherent in retrospective reviews. Tests were ordered when the clinician had a higher suspicion for bacteremia or UTI, yet prev- alence is the same as that of the general population. Though Bonadio et al., had a prevalence of 3% for bacteremia in their neutropenic patients versus 1.9% in the general pediatric population [23], our studies comment on the clinical presentations of the two infants with positive blood cultures [4]. Both were less than 9 months old, had apnea, and signs of shock, which were clinically concerning for sepsis [4].

When we compared the prevalence of Bacterial pneumonia and meningitis in immunocompetent versus incidental neutropenic pa- tients that had CXR and CSF obtained, both SBIs are more common in neutropenic patients. The annual incidence of pneumonia of general pe- diatric children in resource-rich countries is estimated to be between 0.1%-0.3% [25]. CXRs were not obtained in Bonadio and Melendez but were positive in 1.2% [7], 5% [3], and 13% [6] of patients tested [3-7].

The incidence of Bacterial meningitis in the general pediatric popu- lation is <1% [27]. CSF was not obtained in Barg and Wittman but was positive in 0% [7], 1.2% [5], and 10% [4] of patients tested. However, the Bonadio paper describes the three patients out of 30 who had posi- tive CSF culture-all three were ill-appearing and lethargic [4]. This high- lights spectrum bias-CSF culture and CXRs had a higher prevalence, sensitivity, and specificity because the clinicians used their judgment of pre-test probability of SBI to order the tests which preferentially ex- plained their fever [3-7].

We found three prospective studies that reviewed the prevalence of SBI in patients with mild to severe neutropenia (ANC <1500, ^{17, 18, 19]}. We rejected these studies from our systematic review because the ANC was above our inclusion criteria of ANC <1000 and did not

comment on the degree of neutropenia in relation to positive Bacterial cultures. Sixty-one percent [19] and 94% [17] of children had ANC

<1000, and in Alexandropoulou et al. 2013, 24% had ANC <500 and 76% had ANC between 501 and 1500. These prospective studies ob- tained blood and urine culture on all patients. The prevalence of bacter- emia among these three studies ranged from 0% [17], 1.5% [19], to 4.4% [18], which mirrored the range of bacteremia (0% [3,6,7]- 3% [4]) in our reviewed studies [3,4] [5] [6,7]. Urinary tract infections varied among these three studies from 6.6% [18], 7% [17], and 10% [19]. This, again, is very similar to our reviewed neutropenic studies (0%- 8%) [3-7].

When we reviewed studies that also obtained viral testing in addi- tion to blood and urine culture, we saw that the majority of positive tests were viral (47% [18]- 50% [17]). For those patients who tested bac- terial culture-negative and viral test negative, one can assume their in- fection was from a virus that was not detected. Thus, a large percentage of these neutropenic patients will be either viral positive or bacterial culture-negative (79% [18]- 98% [19]).

There are limitations to these prospective studies. These studies were all done in other countries (Greece [18,19] and Kuwait [17]. None of these studies describe the clinical picture of each child. In Alexandropoulou et al., there is a higher prevalence of bacteremia (4.4%) than the immunocompetent population quoted, 1.9% [18,23]. However, three of the positive blood cultures in the Alexandropoulou paper were rare bacterial pathogens in the US (Rickettsia and Brucella) [18]. The other bacteria growing from blood cultures were Pneumococ- cus (2) and Pseudomonas (1) making the prevalence of bacteremia 2.2% [18]. This is similar to the USA bacteremia prevalence in immuno- competent children of 1.9% [23].

With routine pneumococcal vaccination, incidence of PEdiatric bac- teremia, meningitis, and pneumonia have decreased [23,25,27]. Empiric antibiotic treatment carries risk of adverse effects and development of antibiotic resistance [26]. Our study suggests that a risk stratification strategy could be developed to approach febrile, immunocompetent children with moderate to severe neutropenia. A potential strategy could be based on history and clinical appearance. Our analysis suggests that prevalences of UTI and bacteremia are similar in immunocompe- tent children with moderate to severe neutropenia to those of the gen- eral pediatric population, and that most causes of incidental neutropenia seem to be viral. Thus, if a previously healthy child presents with fever and is found to be incidentally moderately to severely neu- tropenic, the clinician should approach this patient similarly as if the pa- tient were not neutropenic, as long as there are no concerning features of bacterial meningitis or pneumonia.

• 1. Limitations

Limitations of our analysis include that all studies reviewed were retrospective chart reviews. Each study was dependent on documenta- tion by healthcare professionals and recorded laboratory data, which is subject to error. Not all of the studies did each test or Imaging study on every patient to screen for SBI. The true prevalence of SBI of each study could not be ascertained.

1. Conclusion

This systematic review could not conduct a meta-analysis given the limitations of our retrospective studies. However, there is spectrum bias inherent in retrospective reviews; clinicians clearly used their judgment in choosing which tests to confirm their diagnosis. It would be unethical and unrealistic to conduct a prospective review where every neutropenic child receives a lumbar puncture and CXR if clinical suspicion for bacterial meningitis or pneumonia is low, given the harms and risks of the tests.

When examining each of the tests individually, the prevalence of bac- teremia and urinary tract infections was similar in incidentally neutrope- nic patients compared to immunocompetent patients. Prevalence of bacterial meningitis and pneumonia was higher in neutropenic patients

in our reviewed studies, but the spectrum bias we identified falsely in- creases the prevalence, sensitivity, and specificity of these tests.

If a febrile child is found to be incidentally moderately to severely neu- tropenic, and clinical suspicion for bacterial pneumonia or meningitis is low, their risk for SBI is similar to that of immunocompetent patients. It seems most causes of incidental neutropenia are viral related, and most of these patients are bacterial culture-negative and/or viral test positive. Antibiotics can cause allergic reactions, reduce the growth of normal in- testinal flora, and increase Antimicrobial resistance [26]. It is worthwhile to risk-stratify moderate to severely neutropenic pediatric patients pre- senting with fever who were previously healthy; empiric antibiotic treat- ment can cause harm and is oftentimes unnecessary in these patients.

Role of Funder/Sponsor

Not Applicable. Registration: PROSPERO: CRD42020186862.

Article summary: Our systematic review of five studies (1693 pa- tients) found a similar prevalence (1.9%) of serious bacterial infections between healthy febrile neutropenic and non-neutropenic children.

Contributors’ statement page

Rosy Hao MD, Mona Saleh MD, Tian Liang MD, Neh Molyneaux MD, Isaac Gordon MD, and Chiemelie Anyachebelu MD conceptualized and designed the study, drafted the initial manuscript, and reviewed and re- vised the manuscript.

Richard Sinert, DO conceptualized and designed the study, coordi- nated and supervised data collections, and critically reviewed the man- uscript for important intellectual content.

Funding

No sources of funding, data was extracted from public databases.

Declaration of Competing Interest

The authors have no conflicts of interest relevant to this article to disclose.

Appendix A PubMed

(“neutropaenia”[All Fields] OR “neutropenia”[MeSH Terms] OR “neutropenia”[All Fields]) AND (“bacteraemia”[All Fields] OR “bacteremia”[MeSH Terms] OR “bacteremia”[All Fields] OR “bacterial infections”[MeSH Terms] OR (“bacterial”[All Fields] AND “infections”[All Fields]) OR “bacterial infections”[All Fields] OR (“bacterial”[All Fields] AND “infection”[All Fields]) OR “bacterial infection”[All Fields]) AND (“risk”[MeSH Terms] OR “risk”[All Fields] OR “Risk Assessment”[Mesh]) AND (“infant”[MeSH Terms] OR “child”[MeSH Terms] OR “adolescent”[MeSH Terms] OR “infant”[All Fields] OR “child”[All Fields] OR “children”[All Fields] OR “adolescent”[All Fields])

Appendix B. EMBASE

(‘neutropenia’/exp. OR neutropenia OR neutropaenia) AND (‘bacter- emia’/exp. OR bacteremia OR bacteraemia OR ‘bacterial infection’/exp. OR ‘bacterial infection’) AND (‘risk’/exp. or risk) AND ([infant]/lim OR [child]/lim OR [adolescent]/lim)

Appendix C. Web of science

(neutropenia OR neutropaenia) AND (bacteremia OR bacteraemia OR bacterial infection) AND risk AND (infant OR child OR adolescent).

Funding support

None.

References

1. Freifeld AG, Bow EJ, Sepkowitz KA, et al. Clinical practice guideline for the use of an- timicrobial agents in neutropenic patients with cancer: 2010 update by the Infec- tious Diseases Society of America. Clin Infect Dis. 2011;52:427-31.
2. Alario AJ, O’Shea JS. Risk of infectious complications in well-appearing children with transient neutropenia. Am J Dis Child. 1989;143:973-6.
3. Wittmann O, Rimon A, Scolnik D, Glatstein M. Outcomes of immunocompetent chil- dren presenting with fever and neutropenia. J Emerg Med. 2018;54:315-9.
4. Bonadio WA, Stremski E, Shallow K. Clinical characteristics of children with fever and transient neutropenia who experience serious bacterial infections. Pediatr Emerg Care. 1989;5:163-5.
5. Melendez E, Harper MB. Risk of serious bacterial infection in isolated and unsus- pected neutropenia. Acad Emerg Med. 2010;17:163-7.
6. Barg AA, Kozer E, Mordish Y, Lazarovitch T, Kventsel I, Goldman M. The risk of seri- ous bacterial infection in Neutropenic Immunocompetent febrile children. J Pediatr Hematol Oncol. 2015;37:e347-51.
7. Pascual C, Trenchs V, Hernandez-Bou S, Catala A, Valls AF, Luaces C. Outcomes and infectious etiologies of febrile neutropenia in non-immunocompromised children who present in an emergency department. Eur J Clin Microbiol Infect Dis. 2016; 35:1667-72.
8. DeAngelis C, Joffe A, Wilson M, Willis E. Iatrogenic risks and financial costs of hospi- talizing febrile infants. Am J Dis Child. 1983;137:1146-9.
9. Shrestha P, Cooper BS, Coast J, et al. Enumerating the economic cost of antimicrobial resistance per antibiotic consumed to inform the evaluation of interventions affect- ing their use. Antimicrob Resist Infect Control. 2018;7:98.
10. Shamseer L, Moher D, Clarke M, et al. Preferred reporting items for systematic re- view and meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation. BMJ. 2015;350:g7647.
11. https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=186862.
12. Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al. Meta- analysis of observational studies in epidemiology: a proposal for reporting. Meta- analysis Of Observational Studies in Epidemiology (MOOSE) group. JAMA. 2000; 283:2008-12.
13. Wells GA, Shea B, O’Connell D, et al. The Newcastle- Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses Available from: http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp. (Accessed Oct 1, 2017).
14. Gilbert EH, Lowenstein SR, Koziol-McLain J, et al. Chart reviews in emergency med- icine research: where are the methods? Ann Emerg Med. 1996;27(3):305-8.
15. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta- analyses. BMJ. 2003;327:557-60.
16. Barendregt JJ, Doi SA, Lee YY, Norman RE, Vos T. Meta-analysis of prevalence. J Epidemiol Community Health. 2013;67:974-8.
17. Husain EH, Mullah-Ali A, Al-Sharidah S, Azab AF, Adekile A. Infectious etiologies of transient neutropenia in previously healthy children. Pediatr Infect Dis J. 2012;3: 575-7.
18. Alexandropoulou O, Kossiva L, Haliotis F, et al. Transient neutropenia in children with febrile illness and associated infectious agents: 2 years’ follow-up. Eur J Pediatr. 2013;172:811-9.
19. Karavanaki K, Polychronopoulou S, Giannaki M, et al. Transient and chronic neutropenias detected in children with different viral and bacterial infections. Acta Paediatr. 2006;95:565-72.
20. Serwint JR, Dias MM, Chang H, Sharkey M, Walker AR. Outcomes of febrile children presumed to be immunocompetent who present with leukopenia or neutropenia to an ambulatory setting. Clin Pediatr (Phila). 2005;44:593-600.
21. Hall MK, Kea B, Wang R. Recognising Bias in studies of diagnostic tests part 1: patient selection. Emerg Med J. 2019;36:431-4.
22. Kohn MA, Carpenter CR, Newman TB. Understanding the direction of bias in studies of Diagnostic test accuracy. Acad Emerg Med. 2013;20:1194-206.
23. Alpern ER, Alessandrini EA, Bell LM, Shaw KN, McGowan KL. Occult bacteremia from a pediatric emergency department: current prevalence, time to detection, and out- come. Pediatrics. 2000;106:505-11.
24. Shaikh N, Morone NE, Bost JE, Farrell MH. Prevalence of urinary tract infection in childhood: a meta-analysis. Pediatr Infect Dis J. 2008;27:302-8.
25. Harris M, Clark J, Coote N, et al. British Thoracic Society guidelines for the manage- ment of community acquired pneumonia in children: update 2011. Thorax. 2011; 66(Suppl. 2) ii1-23.
26. Tamma PD, Avdic E, Li DX, et al. Association of Adverse Events with Antibiotic use in hospitalized patients. JAMA Intern Med. 2017;177:1308.
27. Thigpen MC, Whitney CG, Messonnier NE, Zell ER, Lynfield R, Hadler JL, et al. Emerg- ing infections programs network. bacterial meningitis in the United States, 1998- 2007. N Engl J Med. 2011 May 26;364(21):2016-25. https://doi.org/10.1056/ NEJMoa100538421612470.