OBJECTIVES:

To examine the association between cough status and bacterial infections (BIs) to more accurately stratify risk and predict BIs in febrile infants.

METHODS:

A retrospective cohort study was performed by identifying all infants ≤60 days old with temperature ≥38°C at an urban pediatric emergency department from 2014 to 2016. The Rochester Risk model was used to stratify risk. Cough status (with or without) was the main covariate of interest. The primary outcome was a BI, including urinary tract infection, bacteremia, or meningitis. Analyses consisted of descriptive statistics, simple and multiple regression to compare the odds of BI on the basis of cough status, as well as χ2 statistics to compare the BI rates among high-risk infants with and without cough.

RESULTS:

Of 508 febrile infants ≤60 days old, 46 (9.1%) had a BI, 13 of which were either bacteremia or meningitis. There were no BIs among low-risk infants with a cough. The odds of BI increased progressively, peaking at 14.6 (95% confidence interval: 4.3–49.7) for high-risk infants without a cough. The adjusted odds of BI among infants with cough was 0.47 (95% confidence interval: 0.22–0.99).

CONCLUSIONS:

In our findings, an inverse relationship is demonstrated between presence of cough and odds of BI, suggesting that cough status may be a useful marker of viral infections in febrile infants. Considering that detecting cough status is noninvasive, inexpensive, and immediately available, it represents an attractive value-based risk factor to enhance current BI prediction models.

Management of the febrile infant is a cornerstone of pediatrics because ∼250 000 infants <60 days of age are brought to medical attention for fever each year.13  Because of immature immune systems,4  febrile infants are at risk for a bacterial infection (BI), including urinary tract infection, bacteremia, and meningitis. Diagnosis of a BI is challenging because fever is often the single, sentinel sign of infection, and infected infants can often appear well. However, if left untreated, BIs can have devastating consequences.5 

In previous studies, researchers have consistently demonstrated a lower risk of BIs among febrile infants with a viral respiratory tract infection.616  This finding has proved valuable in refining the risk of BIs8,17 ; however, viral detection tests are expensive and are not consistently performed on-site in rural and community settings, meaning that results may not be available for several days. In addition, positive viral test results from the nasopharynx do not always correlate with clinical disease.18,19  In findings from earlier studies in the 1980s, it is suggested that febrile infants with signs and symptoms of respiratory infection may have fewer BIs.20,21  These studies, however, were limited by small sample sizes, the use of varying definitions of respiratory symptoms, and few infants with bronchiolitis. More recently, this topic has been more systematically examined, as several studies have demonstrated that febrile infants with clinically diagnosed respiratory viral infections, such as bronchiolitis, have lower rates of BIs.2224  Clinical history markers, including cough, wheezing, and rhinorrhea, are noninvasive, inexpensive, instantly available, and may improve the accuracy in detecting infants at high risk of BI. The objective of this study was to evaluate whether cough status may be associated with BIs. We hypothesized that the presence of cough would indicate presence of a viral infection and lower odds of BI.

This pilot study was part of a larger retrospective cohort study from a single, urban, inner-city pediatric emergency department (ED) with ∼70 000 annual encounters. The aim of the larger study was to develop a novel BI predictive model using clinical history risk factors. Our institutional review board approved this study. First, we identified all infants 0 to 90 days old who were brought to the ED between January 1, 2014, and December 31, 2016. Next, using a standard abstraction tool, 3 members of the research team manually reviewed the electronic medical record (EMR) of each infant for presence of fever. Inclusion criteria for this study were as follows: (1) age 0 to 60 days old; (2) fever, defined as a temperature ≥38°C, within 6 hours of arriving to the ED or reported by the caregiver from a measurement before arrival; and (3) documentation of cough status (with or without). We used the Rochester Risk (RR) model to differentiate high-risk from low-risk infants (Supplemental Table 4).25  Documentation and laboratory studies were performed at the discretion of the treating clinician. One author (J.P.Y.) reviewed the abstracted EMR data to assign risk status. On the basis of our clinical experience, it is common practice for low-risk clinical history markers to be omitted from the clinician’s documentation. Therefore, we considered missing data as negative for purposes of risk stratification.

We collected demographic characteristics, including age, sex, race and ethnicity, and insurance type for each infant. Gestational age was recorded as term (≥37 weeks) or preterm (<37 weeks). To identify chronic medical conditions, 1 author (J.P.Y.) manually reviewed all records and created a dichotomous variable (yes or no) on the basis of the likelihood that the condition may contribute to the development of a BI.26  Examples of chronic medical conditions included vesicoureteral reflux, panhypopituitarism, and idiopathic neutropenia.26  We considered a social concern as being present if it was documented as a factor in the clinical decision-making. Per standard practice, ED clinicians consistently used a standard template that included a “Review of Systems” noting the presence or absence of cough. Cough status was recorded if explicitly documented as present or absent in the EMR.

The main outcome was BI, defined as either urinary tract infection, bacteremia, or meningitis. Urine cultures were considered positive results if there were signs of inflammation in the urinary tract, including ≥5 white blood cells per high-power field (unspun) or the presence of leukocyte esterase (at least trace), and growth of at least 10 000 colony-forming units per milliliter of a single organism from a catheterized specimen that was treated clinically as a pathogen.27  Bacteremia and meningitis were defined as growth of a single organism from blood or cerebrospinal fluid cultures that was treated clinically as a pathogen.

We performed descriptive statistics on patient characteristics, including means, medians, and interquartile ranges for continuous variables and frequencies and percentages for categorical variables. Using simple logistic regression, we estimated the odds of BI among groups with different RR (low or high) and cough status (with or without). We also performed multiple logistic regressions to further evaluate the association of cough with BIs after adjusting for potential confounders. A priori variables consisted of age, sex, race and ethnicity, insurance type, RR status, cough status, duration of symptoms, and respiratory season. We used stepwise model selection to identify variables for inclusion. We also used χ2 statistics for proportions to compare rates of urinary tract infections, bacteremia, and meningitis between high-risk infants with and without cough. Statistical analyses were performed with SAS (version 9.4; SAS Institute, Inc, Cary, NC).

A total of 508 infants were ≤60 days old with fever and cough data available. Four infants were excluded because of absence of documented cough status, none of whom had a BI. The median age was 40 days, a majority were male, half were Hispanic, and >90% had public insurance (Table 1). Almost half of infants were high risk (47%), and 223 (44%) had a cough. Forty-six infants (9%) had a BI, 13 of which were due to bacteremia or meningitis (Table 2). For risk stratification purposes, 8% of data were missing, over half of which was due to unclear documentation of the infant’s course in the newborn nursery.

TABLE 1

Subject Characteristics (N = 508)

Characteristicn (%)
Median age in d (IQR) 40 (26) 
Male 291 (57) 
White or other 52 (10) 
African American 195 (38) 
Hispanic 256 (50) 
Unknown race or ethnicity 5 (1) 
Term 474 (93) 
Chronic medical condition 14 (3) 
Public insurance 465 (92) 
Rochester high risk 237 (47) 
Cough present 223 (44) 
BI 46 (9) 
Characteristicn (%)
Median age in d (IQR) 40 (26) 
Male 291 (57) 
White or other 52 (10) 
African American 195 (38) 
Hispanic 256 (50) 
Unknown race or ethnicity 5 (1) 
Term 474 (93) 
Chronic medical condition 14 (3) 
Public insurance 465 (92) 
Rochester high risk 237 (47) 
Cough present 223 (44) 
BI 46 (9) 

IQR, interquartile range.

TABLE 2

Etiology and Type of BI, Categorized by Age

0–30 d31–60 dTotal
No. febrile infants 148 360 508 
No. infants with any BI, n (%) 21 (14.2) 25 (6.9) 46 (9.1) 
No. infants with bacteremia or meningitis, n (%) 4 (2.7) 9 (2.5) 13 (2.6) 
No. infants with UTI, n (%) 17 (11.5) 16 (4.4) 33 (6.5) 
Organism recovered from urine (nEscherichia coli (12)Klebsiella pneumoniae (3)Enterococcus faecalis (1)Group B Streptococcus (1) E coli (13)E faecalis (1)Citrobacter koseri (2) — 
No. infants with UTI and bacteremia, n (%) 2 (0.6) 2 (0.4) 
Organism recovered from urine and blood (n— E coli (2) — 
No. infants with bacteremia only, n (%) 3 (2) 5 (1.4) 8 (1.6) 
Organism recovered from blood (nGroup B Streptococcus (3) Group B Streptococcus (2)Salmonella species (1)K pneumoniae (1)Staphylococcus aureus (1) — 
No. infants with meningitis only, n (%) 1 (0.7) 1 (0.2) 
Organism recovered from cerebrospinal fluid (nE coli (1) — — 
No. infants with bacteremia and meningitis, n (%) 2 (0.6) 2 (0.4) 
Organism recovered from blood and cerebrospinal fluid (n— Group B Streptococcus (2) — 
0–30 d31–60 dTotal
No. febrile infants 148 360 508 
No. infants with any BI, n (%) 21 (14.2) 25 (6.9) 46 (9.1) 
No. infants with bacteremia or meningitis, n (%) 4 (2.7) 9 (2.5) 13 (2.6) 
No. infants with UTI, n (%) 17 (11.5) 16 (4.4) 33 (6.5) 
Organism recovered from urine (nEscherichia coli (12)Klebsiella pneumoniae (3)Enterococcus faecalis (1)Group B Streptococcus (1) E coli (13)E faecalis (1)Citrobacter koseri (2) — 
No. infants with UTI and bacteremia, n (%) 2 (0.6) 2 (0.4) 
Organism recovered from urine and blood (n— E coli (2) — 
No. infants with bacteremia only, n (%) 3 (2) 5 (1.4) 8 (1.6) 
Organism recovered from blood (nGroup B Streptococcus (3) Group B Streptococcus (2)Salmonella species (1)K pneumoniae (1)Staphylococcus aureus (1) — 
No. infants with meningitis only, n (%) 1 (0.7) 1 (0.2) 
Organism recovered from cerebrospinal fluid (nE coli (1) — — 
No. infants with bacteremia and meningitis, n (%) 2 (0.6) 2 (0.4) 
Organism recovered from blood and cerebrospinal fluid (n— Group B Streptococcus (2) — 

UTI, urinary tract infection; —, not applicable.

Cough was absent in 34 infants with a BI (74%). The frequency of BIs in high-risk infants was 18% compared with 1% in low-risk infants. For the simple regression analysis, because no low-risk infants with cough had a BI, we used the low-risk without cough group as the reference. The odds of BI increased in a stepwise fashion, peaking with high-risk infants without cough, whose odds of BI was 14.6 (95% confidence interval [CI]: 4.3–49.7) (Table 3). For the multiple regression analysis, presence of cough was associated with lower odds of BI (adjusted odds ratio: 0.47; 95% CI: 0.22–0.99), and high-risk RR status was associated with higher odds of BI (adjusted odds ratio: 25.5; 95% CI: 7.4–88.3). High-risk infants without cough were significantly more likely to be diagnosed with a urinary tract infection compared with high-risk infants with cough (22.9% vs 9.3%; P = .007). Infants with missing observations for purposes of RR stratification were older than infants with complete data (median = 44 days versus 37 days; P < .00001) and had lower rates of BI (7.7% vs 10.7%), although this difference was not statistically significant. We were unable to detect a significant difference in rates of bacteremia or meningitis for high-risk infants on the basis of cough status. Of 3 low-risk infants with a BI, none had a cough.

TABLE 3

Odds Ratio Estimates by Risk and Cough Status

RRCough StatusBI Rate; % (95% CI)OR (95% CI)
Low With 0 out of 126 (0–2.4) — 
Low Without 3 out of 145; 2.1% (0.4–6.2) 1 (reference) 
High With 12 out of 97; 12.5% (7.1–20.5) 8.7 (2.3–32.5) 
High Without 31 out of 140; 22.1% (16–29.8) 14.6 (4.3–49.7) 
RRCough StatusBI Rate; % (95% CI)OR (95% CI)
Low With 0 out of 126 (0–2.4) — 
Low Without 3 out of 145; 2.1% (0.4–6.2) 1 (reference) 
High With 12 out of 97; 12.5% (7.1–20.5) 8.7 (2.3–32.5) 
High Without 31 out of 140; 22.1% (16–29.8) 14.6 (4.3–49.7) 

OR, odds ratio; —, not applicable.

To our knowledge, this is the first study to examine the relationship between cough status and BIs in febrile infants. Our results indicate that absence of cough is associated with BIs in febrile infants ≤60 days old. This is the same inverse relationship observed in other studies that identified a higher frequency of BIs in febrile infants without a respiratory viral infection.810,1215,22,24  Cough status may be a useful marker of viral infections in febrile infants as suggested by our results. This is compelling because cough status is inexpensive, noninvasive, and often already collected during the clinical encounter, making it an attractive tool to improve value. Specialized equipment or staff are not needed, and the result is immediately available. Furthermore, the use of cough as a risk factor for BIs is attractive because it often indicates the presence of a current respiratory viral infection. In this study, it is indicated that an additional, easily obtained element of the clinical history may provide further guidance in the calculation of BI risk for febrile infants. As with any clinical decision, the application of these results would be contingent on the balance of risk tolerance and aversion in a particular clinician-caregiver dyad. Previously, Byington et al8,28  identified a group of moderate-risk infants, categorized as high risk with virus, who could be safely discharged earlier and with fewer doses of empirical antibiotics than high-risk infants without virus. In our study, there was no such moderate-risk group because the frequency of BI was >10% for high-risk infants regardless of cough status. However, no low-risk infants with cough had a BI, suggesting that cough status may be used to identify febrile infants at extremely low risk of BI. Future researchers should investigate this finding to determine if low-risk febrile infants with cough may represent an additional BI risk subgroup and may benefit from a different management strategy than low-risk infants without cough or high-risk infants.

This study has several limitations. First, as a retrospective study, the treating clinician used their discretion in obtaining a medical history and performing laboratory tests, resulting in missing data, which we considered as negative for purposes of risk stratification. In a sensitivity analysis, it was demonstrated that infants with missing data were older and had a lower rate of BI, suggesting this was a reasonable approach. Second, this study represents the experience at 1 institution. Further investigation is needed at other sites to attempt to replicate these findings. Third, although there is no reason to suspect that cough status was recorded erroneously in the EMR, it is possible. We think that high rates of completion of a standard template to collect the medical history and cough status mitigates this concern, but further investigation of the reliability of reported cough status may be warranted. Fourth, because of a limited sample size, this study was not powered to detect small but potentially clinically meaningful differences in BIs across the 4 risk groups.

In findings from this pilot study, it is demonstrated that cough status is inversely associated with BIs and may represent an inexpensive, noninvasive, value-based clinical history marker of viral infections for febrile infants. Future researchers should use larger data sets to detect the infrequent BI rate in low-risk infants with cough. In addition, researchers should examine how cough status compares with viral testing results as well as how this information, and other clinical history markers, may be integrated with current low-risk models to improve the prediction of BIs in febrile infants.

We thank Lauren G. Solan, MD, MEd (Assistant Professor of Pediatrics, University of Rochester Medical Center), and Constance D. Baldwin, PhD (Professor of Pediatrics, University of Rochester Medical Center), for their thoughtful suggestions and revisions for the final draft of the article. Thanks also to Carlyn Todorow, MD, for her assistance in data abstraction.

Dr Klouda conceptualized and designed the study, collected data, drafted the initial manuscript, and revised the manuscript; Dr Wang conducted the initial and final analyses and reviewed and revised the manuscript; Dr Yaeger conceptualized and designed the study, designed the data collection instruments, collected data, and reviewed and revised the manuscript; and all authors approve the final manuscript as submitted and agree to be accountable for all aspects of the work.

FUNDING: No external funding.

1
Greenhow
TL
,
Hung
YY
,
Pantell
RH
.
Management and outcomes of previously healthy, full-term, febrile infants ages 7 to 90 days
.
Pediatrics
.
2016
;
138
(
6
):
e20160270
2
Baskin
MN
.
The prevalence of serious bacterial infections by age in febrile infants during the first 3 months of life
.
Pediatr Ann
.
1993
;
22
(
8
):
462
466
3
Hamilton
B
,
Martin
J
,
Osterman
M
,
Curtin
S
,
Mathews
T
.
Births: final data for 2014
.
Natl Vital Stat Rep
.
2015
;
64
(
12
):
1
64
4
Melville
JM
,
Moss
TJ
.
The immune consequences of preterm birth
.
Front Neurosci
.
2013
;
7
:
79
5
Bonadio
WA
,
Hegenbarth
M
,
Zachariason
M
.
Correlating reported fever in young infants with subsequent temperature patterns and rate of serious bacterial infections
.
Pediatr Infect Dis J
.
1990
;
9
(
3
):
158
160
6
Kuppermann
N
,
Bank
DE
,
Walton
EA
,
Senac
MO
 Jr
,
McCaslin
I
.
Risks for bacteremia and urinary tract infections in young febrile children with bronchiolitis
.
Arch Pediatr Adolesc Med
.
1997
;
151
(
12
):
1207
1214
7
Antonow
JA
,
Hansen
K
,
McKinstry
CA
,
Byington
CL
.
Sepsis evaluations in hospitalized infants with bronchiolitis
.
Pediatr Infect Dis J
.
1998
;
17
(
3
):
231
236
8
Byington
CL
,
Enriquez
FR
,
Hoff
C
, et al
.
Serious bacterial infections in febrile infants 1 to 90 days old with and without viral infections
.
Pediatrics
.
2004
;
113
(
6
):
1662
1666
9
Levine
DA
,
Platt
SL
,
Dayan
PS
, et al
;
Multicenter RSV-SBI Study Group of the Pediatric Emergency Medicine Collaborative Research Committee of the American Academy of Pediatrics
.
Risk of serious bacterial infection in young febrile infants with respiratory syncytial virus infections
.
Pediatrics
.
2004
;
113
(
6
):
1728
1734
10
Blaschke
AJ
,
Korgenski
EK
,
Wilkes
J
, et al
.
Rhinovirus in febrile infants and risk of bacterial infection
.
Pediatrics
.
2018
;
141
(
2
):
e20172384
11
Oray-Schrom
P
,
Phoenix
C
,
St Martin
D
,
Amoateng-Adjepong
Y
.
Sepsis workup in febrile infants 0-90 days of age with respiratory syncytial virus infection
.
Pediatr Emerg Care
.
2003
;
19
(
5
):
314
319
12
Krief
WI
,
Levine
DA
,
Platt
SL
, et al
;
Multicenter RSV-SBI Study Group of the Pediatric Emergency Medicine Collaborative Research Committee of the American Academy of Pediatrics
.
Influenza virus infection and the risk of serious bacterial infections in young febrile infants
.
Pediatrics
.
2009
;
124
(
1
):
30
39
13
Titus
MO
,
Wright
SW
.
Prevalence of serious bacterial infections in febrile infants with respiratory syncytial virus infection
.
Pediatrics
.
2003
;
112
(
2
):
282
284
14
Hall
CB
,
Powell
KR
,
Schnabel
KC
,
Gala
CL
,
Pincus
PH
.
Risk of secondary bacterial infection in infants hospitalized with respiratory syncytial viral infection
.
J Pediatr
.
1988
;
113
(
2
):
266
271
15
Mahajan
P
,
Browne
LR
,
Levine
DA
, et al
;
Febrile Infant Working Group of the Pediatric Emergency Care Applied Research Network (PECARN)
.
Risk of bacterial coinfections in febrile infants 60 days old and younger with documented viral infections
.
J Pediatr
.
2018
;
203
:
86
91.e2
16
McDaniel
CE
,
Ralston
S
,
Lucas
B
,
Schroeder
AR
.
Association of diagnostic criteria with urinary tract infection prevalence in bronchiolitis: a systematic review and meta-analysis
.
JAMA Pediatr
.
2019
;
173
(
3
):
269
277
17
Byington
CL
,
Reynolds
CC
,
Korgenski
K
, et al
.
Costs and infant outcomes after implementation of a care process model for febrile infants
.
Pediatrics
.
2012
;
130
(
1
).
18
Kalu
SU
,
Loeffelholz
M
,
Beck
E
, et al
.
Persistence of adenovirus nucleic acids in nasopharyngeal secretions: a diagnostic conundrum
.
Pediatr Infect Dis J
.
2010
;
29
(
8
):
746
750
19
Loeffelholz
MJ
,
Trujillo
R
,
Pyles
RB
, et al
.
Duration of rhinovirus shedding in the upper respiratory tract in the first year of life
.
Pediatrics
.
2014
;
134
(
6
):
1144
1150
20
Greene
JW
,
Hara
C
,
O’Connor
S
,
Altemeier
WA
 III
.
Management of febrile outpatient neonates
.
Clin Pediatr (Phila)
.
1981
;
20
(
6
):
375
380
21
King
JC
 Jr
,
Berman
ED
,
Wright
PF
.
Evaluation of fever in infants less than 8 weeks old
.
South Med J
.
1987
;
80
(
8
):
948
952
22
Luginbuhl
LM
,
Newman
TB
,
Pantell
RH
,
Finch
SA
,
Wasserman
RC
.
Office-based treatment and outcomes for febrile infants with clinically diagnosed bronchiolitis
.
Pediatrics
.
2008
;
122
(
5
):
947
954
23
Ralston
S
,
Hill
V
,
Waters
A
.
Occult serious bacterial infection in infants younger than 60 to 90 days with bronchiolitis: a systematic review
.
Arch Pediatr Adolesc Med
.
2011
;
165
(
10
):
951
956
24
Melendez
E
,
Harper
MB
.
Utility of sepsis evaluation in infants 90 days of age or younger with fever and clinical bronchiolitis
.
Pediatr Infect Dis J
.
2003
;
22
(
12
):
1053
1056
25
Dagan
R
,
Powell
KR
,
Hall
CB
,
Menegus
MA
.
Identification of infants unlikely to have serious bacterial infection although hospitalized for suspected sepsis
.
J Pediatr
.
1985
;
107
(
6
):
855
860
26
Yaeger
JP
,
Moore
KA
,
Melly
SJ
,
Lovasi
GS
.
Associations of neighborhood-level social determinants of health with bacterial infections in young, febrile infants
.
J Pediatr
.
2018
;
203
:
336
344.e1
27
Greenhow
TL
,
Hung
YY
,
Herz
AM
.
Changing epidemiology of bacteremia in infants aged 1 week to 3 months
.
Pediatrics
.
2012
;
129
(
3
).
28
Byington
CL
,
Castillo
H
,
Gerber
K
, et al
.
The effect of rapid respiratory viral diagnostic testing on antibiotic use in a children’s hospital
.
Arch Pediatr Adolesc Med
.
2002
;
156
(
12
):
1230
1234

Competing Interests

POTENTIAL CONFLICT OF INTEREST: Dr Yaeger previously published an article that included a subset of patients from this study (Yaeger JP, Moore KA, Melly SJ, Lovasi GS. Associations of Neighborhood-Level Social Determinants of Health with Bacterial Infections in Young, Febrile Infants. J Pediatr. 2018;203:336–344). However, the work for this current study uses a larger sample population, involves a different research question, evaluates a different outcome, and uses different methods and statistical analyses. The data from this research were also presented as an abstract at the Pediatric Hospital Medicine Conference in July 2018. This article is being submitted only to Hospital Pediatrics and will not be submitted elsewhere while under consideration.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.

Supplementary data