Clinical Predictors of Occult Pneumonia in the

Febrile Child
Charles G. Murphy, MD, Alma C. van de Pol, MD, Marvin B. Harper, MD, Richard G. Bachur, MD

Abstract
Background: The utility of chest radiographs (CXRs) for detecting occult pneumonia (OP) among pediatric
patients without lower respiratory tract signs has been previously studied, but no predictors other than
white blood cell count (WBC) and height of fever have been investigated.
Objectives: To identify predictors of OP in pediatric patients in the postconjugate pneumococcal vaccination era.
Methods: This was a retrospective cross sectional study that was conducted in a large urban pediatric hospital. Physician records of emergency department (ED) patients of age 10 years or less who presented with
fever (R38 C) and had a CXR obtained for suspected pneumonia were reviewed. Patients were classified
into two groups: signs of pneumonia and no signs of pneumonia on the basis of the presence or absence of respiratory distress, tachypnea, or lower respiratory tract findings. Occult pneumonia was defined
as radiographic pneumonia in a patient without signs of pneumonia.
Results: Two thousand one hundred twenty-eight patients were studied. Among patients categorized as
having no signs of pneumonia (n = 1,084), 5.3% (95% CI = 4.0% to 6.8%) had OP. Presence of cough and
longer duration of cough (greater than 10 days) had positive likelihood ratios (LR+) of 1.24 (95% CI =
1.15 to 1.33) and 2.25 (95% CI = 1.21 to 4.20), respectively. Absence of cough had a negative likelihood ratio
(LR) of 0.19 (95% CI = 0.05 to 0.75). The likelihood of OP increased with increasing duration of fever
(LR+ for more than three days and more than five days of fever, respectively: 1.62; 95% CI = 1.13 to 2.31
and 2.24; 95% CI = 1.35 to 3.71). When obtained (56% of patients), WBC was a predictor of OP, with a
LR+ of 1.76 (95% CI = 1.40 to 2.22) and 2.17 (95% CI = 1.58 to 2.96) for WBC of >15,000/mm3 and
>20,000/mm3, respectively.
Conclusions: Occult pneumonia was found in 5.3% of patients with fever and no lower respiratory tract
findings, tachypnea, or respiratory distress. There is limited utility in obtaining a CXR in febrile children
without cough. The likelihood of pneumonia increased with longer duration of cough or fever or in the
presence of leukocytosis.
ACADEMIC EMERGENCY MEDICINE 2007; 14:243249 2007 by the Society for Academic Emergency
Medicine
Keywords: occult pneumonia, febrile child

ever is one of the most common chief complaints

of children presenting to the pediatric emergency
department (ED). The majority of febrile children
who present to the ED are presumed to have a viral
illness, with the clinician relying on the history, clinical
From the Division of Emergency Medicine (CGM, MBH, RGB)
and Division of Infectious Disease (MBH), Childrens Hospital
Boston, Harvard Medical School, Boston, MA; and Division of
Infectious Disease, Wilhelmina Childrens Hospital University
Medical Center Utrecht (ACvdP), Utrecht, The Netherlands.
Received June 29, 2006; revision received August 15, 2006;
accepted August 21, 2006.
Presented in part at the annual meeting of Pediatric Academic
Societies, Washington, DC, May 2005.
Contact for correspondence and reprints: Charles G. Murphy,
MD; e-mail: charles.murphy@mountsinai.org.

2007 by the Society for Academic Emergency Medicine

doi: 10.1197/j.aem.2006.08.022

presentation, and physical examination to determine the

degree of concern for serious bacterial infection. Some
children who present with fever and no apparent source
on physical examination (fever without a source) may
have serious bacterial infections. The inability to rely on
the clinical examination has led to the adoption of certain
empiric testing practices of young febrile children in an
attempt to maximize detection of occult infections.
The diagnosis of bacterial pneumonia in children remains challenging, with numerous previous studies evaluating laboratory data and physical examination findings
as potential clinical predictors of this disease process.
These studies have understandably focused on patients
with respiratory distress or lower respiratory tract findings on physical examination, whereas there has been
little study of patients without respiratory findings or
clinically occult presentations.15 Several studies have

ISSN 1069-6563
PII ISSN 1069-6563583

243

244

Murphy et al.

CLINICAL PREDICTORS OF OCCULT PNEUMONIA IN THE FEBRILE CHILD

evaluated the use of empiric chest radiographs (CXRs)

in the evaluation of febrile infants (age, %3 months) without lower respiratory symptoms and found them not to
be indicated.6 There are limited data regarding the utility
of a CXR in the evaluation of older febrile children without an apparent source of fever. This raises some questions:
is there value in obtaining a CXR in a child with high
fever despite clear lung findings and no tachypnea?
What about the same child with a history of prolonged
cough or fever? What is the value of a radiograph in a
febrile child without auscultatory findings but with
leukocytosis?
It has been previously demonstrated that there is a
high rate of occult pneumonia among febrile children
with leukocytosis.7 However, the introduction and widespread use of the conjugate pneumococcal vaccine has
altered the approach to young febrile children without
an apparent source for the fever on physical exam; notably, physicians are obtaining fewer white blood cell
(WBC) counts.8 In addition, although there is a modest
reduction in cases of pneumonia among vaccinated children, the impact of vaccine use on the prevalence of
occult pneumonia remains unknown.9
The objectives of this study were to 1) investigate the
prevalence of occult pneumonia in the postpneumococcal vaccine era and 2) identify variables in the patients
history and physical examination that are predictive of
patients with occult pneumonia.
METHODS
Study Design
This was a retrospective cross sectional study conducted
in a large pediatric medical center with approximately
50,000 patient visits to the ED annually. The study was
approved by the institutional review board, and data
collection was compliant with the Health Insurance Portability and Accountability Act of 1996.
Study Setting and Population
To avoid seasonal variations, we examined data for one
entire calendar year, 2002. Records of ED patients aged
%10 years who presented to the ED in 2002 with fever
and who had a CXR obtained for suspected pneumonia
were reviewed. Fever was defined as a temperature of
R38 C at triage or as measured at home within the prior
24 hours. Patients were excluded from study if they met
one of the following exclusion criteria: CXR within the
previous 24 hours, radiograph with evidence of pneumonia within the past month, suspicion of foreign-body
aspiration, chronic lung disease (excluding asthma),
sickle cell disease, immunodeficiency, or severe cardiac
or neurological disorders.
Study Protocol
The physicians medical records were reviewed for history and physical examination findings. If a trainee was
involved in the patients care, both the trainee and
attending physicians notes were reviewed. Abnormal
physical findings were considered present if they were
documented by either the resident or attending physician. In the case of any discrepancies, the attending physicians note was recorded as the final determination. In

cases in which the presence or absence of a particular

finding was not specifically documented, it was presumed to be absent. The history was reviewed and the
following variables were recorded: the presence, height,
and duration of fever before presentation; the presence
and duration of cough; history of increased work of
breathing noted by caregiver at home (audible wheezing,
rapid breathing, or retractions); the presence of chest
pain; and the presence of nasal congestion. The recorded
symptoms of congestion, rhinorrhea, or URI symptoms were collectively coded as the presence of nasal
congestion. Physical examination findings that were
reviewed included the following variables: general
appearance of the patient, body temperature at triage,
respiratory rate, oxygen saturation, the presence of visible signs of respiratory distress, and lung auscultation
findings (lower respiratory tract findings). When multiple
temperatures or respiratory rates were noted in the ED
chart, only the highest value was recorded. The lowest
oxygen saturation measured by pulse oximetry at the
ED visit was recorded. A child was defined to be wellappearing if documentation of well-appearing, happy,
playful, active and playing, or non-toxic was in the
record. A child was classified as ill-appearing if described
as ill-appearing, toxic, lethargic, poorly perfused, apneic, or cyanotic. Signs of respiratory distress were defined as any description in the medical
record of respiratory distress, retractions, nasal flaring,
or grunting. Physical examination findings of wheeze,
rales or crackles, rhonchi, decreased breath sounds, or
an asymmetric lung exam were considered to represent
lower respiratory tract findings. The description of
coarse breath sounds was described as normal. Respiratory rates were classified as either normal or tachypneic,
based on age-based normal values. Tachypnea was defined as a respiratory rate of R60 breaths/min for age
%2 years, R50 breaths/min for age 35 years, or R30
breaths/min for age 6 years to 10 years.
Chest radiograph results were based on the attending
radiologists reading of the radiograph. CXRs were classified into one of three groups: positive, negative, or equivocal. A CXR was classified as positive if it was read
as having the presence of an infiltrate, consolidation,
multiple infiltrates, pneumonia, or pneumonia with
effusion. A CXR was classified as negative if it was read
as no pneumonia, normal radiograph, clear lungs,
peri-bronchial cuffing, or bronchial wall thickening.
A CXR was classified as equivocal if it was read as focal
atelectasis, atelectasis vs. infiltrate, or cannot exclude
pneumonia. To achieve the most conservative estimate of
the prevalence of occult pneumonia, equivocal CXRs were
excluded from analysis even if the clinician made the
final diagnosis of pneumonia.
Patients were classified into two groups: signs of
pneumonia and no signs of pneumonia. Patients were
classified as having no signs of pneumonia if they met
the following criteria: no visual signs of respiratory distress (retractions, nasal flaring, or grunting), no tachypnea (based on described age-based normal values), no
hypoxia (room air O2 saturation (SaO2) > 95%) when
measured, and no lower respiratory tract findings on
physical examination (wheeze, rales or crackles, rhonchi,
decreased breath sounds, or an asymmetric lung exam).

ACAD EMERG MED

March 2007, Vol. 14, No. 3

www.aemj.org

Any patients who failed to meet all of these criteria were

classified as having signs of pneumonia (i.e., signs of
respiratory distress, tachypnea by age, lower respiratory
tract findings, or SaO2 % 95% in room air). Occult pneumonia was defined as radiographic pneumonia in a patient classified in the no signs of pneumonia group
(i.e., absence of respiratory distress, tachypnea, hypoxia,
or lower respiratory tract signs).
Data Analysis
Within the no signs of pneumonia subgroup, we compared patients with positive CXRs (occult pneumonia)
with those with negative CXRs. Univariate analysis was
performed for elements of the history, general appearance of the patient, and for WBC count. For patients
with signs of pneumonia, we in addition analyzed specific
physical examination findings (distress, tachypnea, each
auscultatory sign) as a predictor of pneumonia. To add
clinical relevance, we chose to calculate likelihood ratios
(LRs) as opposed to simple sensitivity and specificity.
We performed statistical analysis by using SPSS (version 12.0; SPSS Inc., Chicago, IL). Chi-square analysis
was performed for categorical variables and for threshold values of continuous variables. T-test was used to
compare mean values of continuous variables. In general, statistical significance was considered p % 0.05.
Logistic regression was used to develop a multivariate
prediction model. Variables with p % 0.2 were considered candidate predictors for entry; automated backward stepwise regression analysis was performed by
using SPSS. Confidence intervals (CI) and LRs were

245

also calculated by using Stata (version 8; Stata Corp.,

College Station, TX).
Duplicate abstraction and data entry were performed
for 4% of eligible study patients. Interrater agreement
classifying patient subgroups (signs vs. no signs of pneumonia) and for coding of radiographic readings was
96.1% and 98.7%, respectively.
RESULTS
During the one-year study period, there were 2,128 eligible patients. Of these, 1,084 were classified as having no
signs of pneumonia. One hundred eighty-three (16.9%)
of these patients were excluded from analysis for having
equivocal CXR readings. Fifty-seven or 5.3% (95% CI =
4.0% to 6.8%) had positive CXR readings and were classified as having occult pneumonia. This compares with
a prevalence rate of pneumonia in patients with lower
respiratory tract findings or signs of pneumonia of
12.6% (95% CI = 10.7% to 14.8%; Figure 1).
The median age of the patients in the signs of pneumonia group was 1.7 years (interquartile range [IQR], 0.7 to
3.3 yr). The median age of the patients in the no signs of
pneumonia group was 1.7 years (IQR, 0.8 to 3.5 yr).
Analysis of Patients without Physical-examination
Findings Suggestive of Pneumonia
Among patients without signs of pneumonia (by physical
examination), we compared clinical characteristics among
those with positive CXRs (occult pneumonia) with
those with negative CXRs (Table 1). Patients with occult

Figure 1. Distribution of eligible patients on the basis of the presence or absence of clinical signs of pneumonia and chest
radiograph (CXR) results. Grayed boxes are the study patients who are compared for predictors of occult pneumonia.

246

Murphy et al.

CLINICAL PREDICTORS OF OCCULT PNEUMONIA IN THE FEBRILE CHILD

Table 1
Comparison of Clinical and Laboratory Characteristics of
Patients without Signs of Pneumonia by Physical Examination
Positive CXR
Negative
(Occult Pneumonia)
CXR

pvalue

Parameter

Age (yr)
Temperature
at home ( C)
Triage temp ( C)
WBC count
( 103 per mm3)
Respiratory rate
(respirations/
min)

901
658

3.14 (2.3)
39.5 (0.7)

2.50 (2.3)
39.4 (0.8)

893
508

38.5 (1.0)
21.5 (8.8)

38.6 (2.3) 0.74

14.6 (7.8) <0.01*

828

31.6 (9.2)

30.5 (9.4)

0.05*
0.40

0.44

Data are mean (SD).

CXR = chest radiograph; WBC = white blood cell.
* Statistical significance of p % 0.05.

pneumonia were statistically older in age. Fifty-six percent of the patients had a WBC count obtained; patients with occult pneumonia had higher mean WBC
count as compared with those without pneumonia (21.5 vs.
14.6  103/mm3, respectively, p < 0.01).
In Table 2, we further examined specific historical variables in patients with no signs of pneumonia and
either positive (occult pneumonia) or negative CXRs. The
positive and negative likelihood ratios for each clinical
finding are shown in Table 3. The duration of cough was
found to be statistically significant in patients with a prolonged history of cough. The LR+ increased with increasing duration of fever. Although the duration of fever

Table 2
Frequencies of Symptoms or Signs by Chest Radiograph (CXR)
Results among Children without Signs of Pneumonia

Symptom

Positive CXR
(Occult Pneumonia;
n = 57)

Presence of cough
Duration of cough (d)
>1
>3
>5
>10
Presence of nasal
congestion
Ill appearance
Fevery
>39 C
>40 C
Duration of fever (d)
>1
>3
>5
WBC count (per mm3)
>15,000
>20,000

Negative
CXR
(n = 844)

pvalue

95.7 (44/46)z

77.3 (515/666)z <0.01*

78.8
48.5
42.4
27.3
64.5

78.3
40.0
26.2
12.1
69.9

(26/33)
(16/33)
(14/33)
(9/33)
(20/31)

(278/355)
(142/355)
(93/355)
(43/355)
(411/588)

1.0
0.36
0.07
0.03*
0.04*

1.8 (1/57)

3.0 (26/844)

1.00

71.4 (40/56)
32.1 (18/56)

63.8 (534/837)
29.3 (245/837)

0.31
0.65

68.8 (33/48)
41.7 (20/48)
27.1 (13/48)

55.1 (387/702) 0.07

25.8 (181/702) 0.03*
12.1 (85/702) <0.01*

74.2 (23/31)
61.3 (19/31)

42.1 (201/477) <0.01*

28.3 (135/477) <0.01*

Data are % (n).

WBC = white blood cell.
* Statistical significance at p % 0.05.
y Fever as determined in emergency department.
z Denominators refer to the number of patients for whom the variable is
known.

was significant, the magnitude of the fever at cutoffs of

39.0 C and 40.0 C was not significantly different between
the two groups.
Other variables studied (the presence of chest pain, the
presence of nasal congestion, a history of increased work
of breathing noted at home, and ill appearance) did not
appear to occur at a significantly different rate between
patients with and without pneumonia in the no signs
of pneumonia group.
Candidate predictors (fever for more than 1 day, presence of cough, nasal congestion, and WBC) were analyzed with stepwise logistic regression. Other cutoffs for
the duration of cough and fever were not entered separately because of collinearity. The model retained cough
(OR = 14.4), nasal congestion (OR = 0.083), and WBC
(OR = 1.07) with overall significance of p = 0.001.
Analysis of Patients with Physical-examination
Findings Suggestive of Pneumonia
In Table 4, there is a comparison of the respiratory tract
findings in patients with signs of pneumonia on physical
examination who had either positive or negative CXR
readings. The presence of either rales or decreased
breath sounds was a significant predictor of patients
with pneumonia on CXR; wheeze was a statistically significant negative predictor. Tachypnea, visible signs of
respiratory distress, and the presence of rhonchi were
not significant predictors of pneumonia. Ill appearance
was uncommon among study patients (2.5%) but was a
significant predictor of nonoccult pneumonia (LR+,
5.57; 95% CI = 2.46 to 13.0).

DISCUSSION
Fever remains one of the most common chief complaints
of children presenting to the pediatric ED. The approach
to children with fever and no apparent source on physical examination continues to evolve, as rates of occult
infections change with new vaccine development and their
more widespread distribution. The CXR is an important
tool that is used in the ED for the diagnosis of pneumonia, whereas the diagnosis of pneumonia in the ambulatory setting has typically relied on suggestive physical
examination findings to help determine appropriate empiric initiation of therapy. Although the finding of an infiltrate or radiographic pneumonia cannot be directly
equated with bacterial pneumonia, the majority of clinicians will prescribe antibiotics for presumed bacterial
pneumonia in the setting of these radiographic findings.
This study investigated patients with fever and no physical examination findings suggestive of pneumonia
(auscultatory signs, tachypnea, hypoxia, or respiratory
distress) to determine the prevalence rate of occult pneumonia in the post pneumococcal vaccine era. We also examined whether clinical variables in the patients history
and physical examination might be predictive of pneumonia. Is there value in obtaining a CXR in a child with
high fever despite clear lung findings and no tachypnea?
In addition, does the duration of cough, duration of fever,
or WBC count influence this decision? This study was undertaken to examine whether there exist specific components of the patients acute illness that can help guide the

ACAD EMERG MED

March 2007, Vol. 14, No. 3

www.aemj.org

247

Table 3
Percentage of Patients with Occult Pneumonia Who Have Specific Historical and Laboratory Features, with Positive (LR+) and Negative
(LR) Likelihood Ratios

Feature

% (n) of Patients
with Specific Finding
Who Have Occult Pneumonia

LR+ (95% CI)

LR (95% CI)

7.9 (44/559)z

1.24* (1.15, 1.33)

0.19* (0.05, 0.75)

Presence of cough
Duration of cough (d)
>1
>3
>5
>10
Presence of nasal congestion
Ill appearance
Fevery
>39 C
>40 C
Duration of fever (d)
>1
>3
>5
WBC count (per mm3)
>15,000
>20,000

8.6
10.1
13.1
17.3
4.6
3.7

(26/304)
(16/158)
(14/107)
(9/52)
(20/431)
(1/27)

1.01
1.21
1.62*
2.25*
0.92
0.56

(0.81,
(0.83,
(1.05,
(1.21,
(0.71,
(0.08,

1.21)
1.76)
2.50)
4.20)
1.20)
4.09)

0.98
0.86
0.78
0.83
1.18
1.01

(0.49,
(0.61,
(0.58,
(0.67,
(0.72,
(0.98,

1.49)
1.21)
1.05)
1.02)
1.92)
1.05)

7.0 (40/574)
6.8 (18/263)

1.12 (0.94, 1.33)

1.10 (0.74, 1.63)

0.79 (0.52, 1.21)

0.96 (0.80, 1.15)

7.9 (33/420)
10.0 (20/201)
13.3 (13/98)

1.62* (1.05, 2.50)

1.62* (1.13, 2.31)
2.24* (1.35, 3.71)

0.78 (0.58, 1.05)

0.79 (0.62, 1.00)
0.83* (0.7, 0.99)

10.3 (23/224)
12.3 (19/154)

1.76* (1.40, 2.22)

2.17* (1.58, 2.96)

0.45* (0.24, 0.81)

0.54* (0.35, 0.84)

WBC = white blood cell.

* Statistical significance of likelihood ratio.
y Fever as determined in the emergency department.
z Denominators are the total number of a clinical feature from Table 2 (i.e., 44 + 515 = 559).

clinician in determining which patients are most at risk

for having occult pneumonia.
Numerous studies have focused on clinical predictors
for a positive radiograph in patients with clinically suspected pneumonia. These studies, however, focused on
physical examination findings as predictors for this
disease process. Leventhal1 set out in 1982 to establish a
set of clinical signs and symptoms that could serve as a
guide to when a CXR should be obtained in children suspected of having pneumonia. In his prospective study,
physicians were asked to record the presence or absence
of certain symptoms and physical findings in patients for
whom a radiograph was to be obtained to evaluate for
suspected pneumonia. In that study, tachypnea had the
highest predictive accuracy. The presence of any pulmonary findings, defined as tachypnea, rales, or decreased
breath sounds in his study, was also a useful predictor
of a positive CXR.1 This topic was examined again in a
more recent article by Lynch et al.3 In that prospective
study, patients with positive CXRs were found more

likely to have a history of fever, crackles, decreased

breath sounds, grunting, or retractions than patients
with normal CXRs. These variables, although statistically
significant, had varying positive predictive values of 39%
to 80%. These variables were then combined by using a
multivariate model to attempt to develop a prediction
rule for obtaining CXRs in children with a clinical suspicion of pneumonia. The highest sensitivity was obtained
when fever, decreased breath sounds, crackles, and tachypnea were all found to be present; however, each individual variable, when combined with fever, was highly
sensitive (sensitivities of 93% to 98%) but had relatively
low specificity (6% to 20%).3 We found a prevalence
rate of pneumonia of 12.6% among patients with signs
of pneumonia (signs of respiratory distress, lower respiratory tract auscultatory findings, or tachypnea). This
rate is lower than that seen in some other studies, possibly as a result of the exclusion of patients with wheeze
or recent asthma exacerbation from other studies. In
our study, the presence of rales (crackles) or decreased

Table 4
Comparison of Respiratory Findings in Patients with Signs of Pneumonia by Physical Examination
Sign or Symptom
Wheeze
Rales
Rhonchi
Decreased breath sounds
Signs of respiratory distress
Tachypnea (age defined)

Positive CXR,
% (n; N = 132)
31.8
40.1
18.9
31.8
38.6
71.9

CXR = chest radiograph.

* Statistical significance of likelihood ratio (LR).

(42)
(53)
(25)
(42)
(51)
(95)

Negative CXR,
% (n; N = 613)
48.3
23.3
13.1
10.9
40.1
70.6

(296)
(143)
(80)
(67)
(246)
(433)

p-value
<0.01
<0.01
0.1
<0.01
0.77
1.0

LR+ (95% CI)

0.66*
1.72*
1.45
2.91*
0.96
1.00

(0.51,
(1.34,
(0.57,
(2.08,
(0.76,
(0.89,

0.86)
2.22)
2.18)
4.08)
1.22)
1.12)

LR (95% CI)
1.32*
0.78*
0.93
0.77*
1.02
1.01

(1.15,
(0.67,
(0.85,
(0.68,
(0.88,
(0.75,

1.52)
0.90)
1.02)
0.86)
1.19)
1.37)

248

Murphy et al.

CLINICAL PREDICTORS OF OCCULT PNEUMONIA IN THE FEBRILE CHILD

breath sounds was also found to be a significant predictor for pneumonia among patients with lower respiratory tract findings (LR+ of 1.7 and 2.9, respectively) but
with relatively low positive predictive values of 27.0%
(95% CI = 21.1% to 24.4%) and 39.1% (95% CI =
30.0% to 48.9%), respectively. Tachypnea, although significant in some past studies, was not seen at a statistically higher rate in our patients with pneumonia, when
compared with patients with negative radiographs. Our
study population included a large number of younger
patients (median age for both groups = 1.7 yr), many
of whom were tachypneic, regardless of a positive or
negative CXR. This large number of younger patients
may have made the value of tachypnea as a predictor
of pneumonia more difficult to appreciate.
These previous studies, although helpful guides for the
patient who presents with respiratory signs that are suggestive for pneumonia, leave the clinician without clear
recommendations on the indication for a CXR in the absence of overt signs of pneumonia. The presumption for
previous studies would be that a CXR would not be indicated on the basis of an absence of these clinical signs.
However, these studies did not specifically evaluate the
prevalence rate of occult pneumonia in febrile children.
We find the rate of pneumonia to be 5.3% (95% CI =
4.0% to 6.8%), or 1 in 20 among febrile children without
any lower respiratory tract findings, signs of respiratory
distress, tachypnea, or hypoxia. A study in 1999, before
the use of the conjugate pneumococcal vaccine, reported
a high rate of occult pneumonia (no clinical evidence of
pneumonia) among febrile children with leukocytosis.
In this study, positive radiographs were found in 26%
of febrile children aged %5 years of age with no clinical
signs of pneumonia by examination, if they had fever of
R39 C and a WBC count of R20,000/mm3.7 These results
suggested a much higher incidence of occult pneumonia
in this patient population than previously realized.
With the advent of a conjugate pneumococcal vaccine,
the incidence of invasive Streptococcus pneumoniae disease has decreased significantly.8,10 Historically, the ED
evaluation of febrile children aged 336 months with no
apparent source of infection frequently included evaluation for possible occult bacteremia, with S. pneumoniae
being the most common pathogen isolated in these cases.
As the rates of invasive pneumococcal disease declined
with widespread vaccine use, the practice of evaluating
febrile young children for occult bacteremia has declined, and with that has come a reduction in the frequency in which a complete blood count is obtained in
this population. As a result, the frequency with which
the clinician will have an opportunity to identify leukocytosis to guide evaluation of young febrile children has appropriately decreased. In the absence of available WBC
counts, it is even more important to examine other variables that may guide the clinician in the evaluation of
these patients for occult pneumonia.
In our study, the prevalence of occult pneumonia was
5.3% (95% CI = 4.0% to 6.8%). Among the historical variables studied, the presence of cough was found to be a
statistically significant predictor of patients with occult
pneumonia. Conversely, in febrile children with no signs
of pneumonia on physical examination and no history of
cough, the prevalence of occult pneumonia was only

0.28%. This finding is in agreement with a previous study

among young febrile infants without respiratory symptoms.6 Interestingly, among patients with temperature
of R39 C and a WBC count of R20,000/mm3, the rate
of occult pneumonia was 14.2%, as compared with previous estimates of 26% in the pre-pneumococcal vaccine
era.7 We also found that a longer duration of a patients
cough or fever increased the likelihood of pneumonia,
whereas height of fever was not predictive.

LIMITATIONS
There are several limitations of our study. First, we were
only able to review patients who had a CXR obtained;
therefore, the exact prevalence of occult pneumonia
can only be estimated. The need to have a CXR also introduces potential bias into our study, as well as limits its
possible applicability, as we are studying a population
of patients in whom the clinician has already made the
decision to obtain a CXR. Furthermore, documentation
in medical records can be incomplete or inaccurate. To
help minimize the inherent limitations of a retrospective
review, we developed strict data abstraction rules. In
the case of physical examination, findings were presumed to be absent unless specifically recorded in the
medical record. This introduces a potential source of error into our study, as the possibility exists that specific
physical examination findings were in fact present but
were not noted in the medical record. The presence
and height of fever in several cases was based on information provided from the guardian, because the patient
was afebrile on arrival in the ED. This could raise the potential concern of the accuracy of this data element. In
addition, we were unable to assess whether the radiograph result was known before documentation in the
record; this could lead to recall bias in the physicians
record. Similarly, CXR readings were based on the reading from a single attending radiologist who may not have
been blinded to the clinical presentation of the patient.
Chest radiographs were ordered on patients at the physicians discretion, rather than on the basis of prespecified criteria, raising the potential concern for possible
selection bias. We also adopted a strict definition of
what we term occult pneumonia, which may limit the applicability of our findings. Furthermore, the prevalence
of pneumonia may have been underestimated by our
exclusion of all patients with apparently equivocal CXR
readings; although many of the patients with equivocal
radiographs were treated as having pneumonia, we felt
that this would also overestimate the incidence of pneumonia. Finally, although the presence of leukocytosis
continues to be a predictor of pneumonia, it is important
to note that a WBC count was only obtained in approximately half the patients in the study.

CONCLUSIONS
Radiographic pneumonia can be identified in a subset of
febrile children, even in the absence of lower respiratory
signs by physical examination. These occult pneumonias
are more common in children with prolonged fever and
cough. Conversely, empiric CXRs have limited diagnostic

ACAD EMERG MED

March 2007, Vol. 14, No. 3

www.aemj.org

value in children with the absence of cough, even with

high or prolonged fever.

References
1. Leventhal JM. Clinical predictors of pneumonia as a
guide to ordering chest roentgenograms. Clin Pediatr
(Phila). 1982; 21:7304.
2. Losek JD, Kishaba RG, Berens RJ, Bonadio WA,
Wells RG. Indications for chest roentgenogram in
the febrile young infant. Pediatr Emerg Care. 1989;
5:14952.
3. Lynch T, Platt R, Gouin S, Larson C, Patenaude Y.
Can we predict which children with clinically suspected pneumonia will have the presence of focal infiltrates on chest radiographs? Pediatrics. 2004; 113:
e1869.
4. Grossman LK, Caplan SE. Clinical, laboratory, and
radiological information in the diagnosis of pneumonia in children. Ann Emerg Med. 1988; 17:436.
5. Zukin DD, Hoffman JR, Cleveland RH, Kushner DC,
Herman TE. Correlation of pulmonary signs and

249

symptoms with chest radiographs in the pediatric

age group. Ann Emerg Med. 1986; 15:7926.
6. Bramson RT, Meyer TL, Silbiger ML, Blickman JG,
Halpern E. The futility of the chest radiograph in the
febrile infant without respiratory symptoms. Pediatrics. 1993; 92:5246.
7. Bachur R, Perry H, Harper MB. Occult pneumonias:
empiric chest radiographs in febrile children with
leukocytosis. Ann Emerg Med. 1999; 33:16673.
8. Herz AM, Greenhow TL, Alcantara J, et al. Changing
epidemiology of outpatient bacteremia in 3- to 36month-old children after the introduction of the heptavalent-conjugated pneumococcal vaccine. Pediatr
Infect Dis J. 2006; 25:293300.
9. Black SB, Shinefield HR, Ling S, et al. Effectiveness of
heptavalent pneumococcal conjugate vaccine in children younger than five years of age for prevention
of pneumonia. Pediatr Infect Dis J. 2002; 21:8105.
10. Haddy RI, Perry K, Chacko CE, et al. Comparison
of incidence of invasive Streptococcus pneumoniae
disease among children before and after introduction
of conjugated pneumococcal vaccine. Pediatr Infect
Dis J. 2005; 24:3203.