Eur J Pediatr

DOI 10.1007/s00431-017-2887-y

ORIGINAL ARTICLE

Clinical features and inflammatory markers in pediatric

pneumonia: a prospective study
Are Stuwitz Berg 1,2 & Christopher Stephen Inchley 1 & Hans Olav Fjaerli 1 &
Truls Michael Leegaard 3,2 & Morten Lindbaek 4 & Britt Nakstad 1,2

Received: 22 December 2016 / Revised: 19 February 2017 / Accepted: 23 February 2017

# Springer-Verlag Berlin Heidelberg 2017

Abstract In this prospective, observational study on previ- C-reactive protein (CRP) values, higher age, and SpO2 ≤92%
ously healthy children <18 years, we aimed to study the diag- in multivariate logistic regression, OR 1.06 (95% CI 1.03 to
nostic ability of clinical features and inflammatory markers to 1.09), OR 1.09 (95% CI 1.00 to1.18), and OR 2.71 (95% CI
(i) predict pathologic chest radiography in suspected pneumo- 1.42 to 5.18), respectively. In proven pneumonia, bacterial
nia and (ii) differentiate etiology in radiological proven pneu- pneumonia was significantly differentiated from viral/
monia. In 394 cases of suspected pneumonia, 265 (67%) had atypical pneumonia by increasing CRP values and SpO2
radiographs consistent with pneumonia; 34/265 had proof of >92% in multivariate logistic regression, OR 1.09 (95% CI
bacterial etiology. Of the cases, 86.5% had received pneumo- 1.05 to 1.14) and OR 0.23 (95% CI 0.06 to 0.82), respectively.
coccal conjugate vaccine. In suspected pneumonia, positive Combining high CRP values (>80 mg/L) and elevated white
chest radiography was significantly associated with increasing blood cell (WBC) count provided specificity >85%, positive
likelihood ratios >3, but sensitivity <46% for both radiograph-
Communicated by David Nadal ic proven and bacterial pneumonia.
Conclusion: With relatively high specificity and likelihood
* Are Stuwitz Berg ratio CRP, WBC count and hypoxemia may be beneficial in
[email protected] ruling in a positive chest radiograph in suspected pneumonia
and bacterial etiology in proven pneumonia, but with low
Christopher Stephen Inchley sensitivity, the clinical utility is limited.
[email protected]

Hans Olav Fjaerli

What is Known:
[email protected]
• Pneumonia is recommended to be a clinical diagnosis, and neither
Truls Michael Leegaard clinical features nor inflammatory markers can reliably distinguish
[email protected] etiology.
• The etiology of pneumonia has changed after routine pneumococcal
Morten Lindbaek
conjugate vaccine.
[email protected]
What is New:
Britt Nakstad
• High CRP and WBC counts were associated with infiltrates in children
[email protected]
with suspected pneumonia and with bacterial infection in proven
pneumonia.
1
Department of Pediatric and Adolescent Medicine, Akershus • In the post-pneumococcal vaccination era, viral etiology is expected,
University Hospital, Postal Box 1000, 1478 Lørenskog, Norway and in cases of pneumonia with low CRP and WBC counts, a
2
Institute of Clinical Medicine, Faculty of Medicine, University of watch-and-wait strategy for antibiotic treatment may be applied.
Oslo, Oslo, Norway
3
Department of Microbiology and Infection Control, Akershus
University Hospital, Lørenskog, Norway
4
Institute of Health and Society, Faculty of Medicine, University of Keywords Pneumonia . Clinical features . Inflammatory
Oslo, Oslo, Norway markers . Pneumococcal vaccination
 Eur J Pediatr

Abbreviations Materials and methods

ALRI Acute lower respiratory tract infection
AUC Area under the curve Study design and population
CAP Community-acquired pneumonia
CRP C-reactive protein This prospective, observational diagnostic study was conduct-
IQR Interquartile range ed at the Department of Pediatric and Adolescent Medicine,
PCR Polymerase chain reaction Akershus University Hospital, Norway, from 1 January 2012
ROCs Receiver operating characteristics to 1 January 2014. Children and adolescents 0–18 years of age
WBC White blood cells were considered for inclusion and recruited by either (i) the
attending physician in the pediatric emergency room (ambu-
latory and hospitalized patients) or (ii) the primary care phy-
sician where referral to hospital was not seen as necessary.
Eligible patients were included in two steps. First, the inclu-
Introduction sion criteria into the cohort of clinically suspected CAP cases
were (1) measured body temperature >37.5 °C at inclusion or
WHO and major International Management Guidelines rec- a history of fever to assure the acuteness of the current infec-
ommend that the diagnosis of community-acquired pneumo- tion, (2) clinical sign(s) of lower respiratory tract infection
nia (CAP) in children should be based on clinical criteria [4, (tachypnea, chest retractions, cough), and (3) either a chest
13, 15] and that routine measurements of inflammatory radiograph ordered by the attending clinician due to suspected
markers or routine chest radiography are not warranted, as pneumonia in patients enrolled in hospital (admitted or ambu-
these investigations cannot reliably distinguish between viral latory treated) or intention to treat with antibiotics due to clin-
and bacterial etiologies [4, 15]. Based on clinical signs and ical suspicion of pneumonia in patients enrolled in primary
symptoms, it is difficult to distinguish pneumonia from other care. Patients enrolled from primary care were sent to the
pediatric acute lower respiratory tract infections (ALRIs) [27], hospital for chest radiography and diagnostic tests the follow-
and a recent systematic review and meta-analysis highlights ing day. Secondly, the inclusion criteria into the cohort of
these difficulties [33]. The etiology of pediatric CAP has proven pneumonia consisted of all in the clinically suspected
changed in areas where the pneumococcal conjugate vaccine cohort who were found to have a chest radiograph consistent
has been introduced in routine immunization programs. In a with pneumonia as described below. Exclusion criteria were
previous publication of this cohort, we observed an increase in severe motor impairment, innate or iatrogenic immunodefi-
the proportion of viral etiology and a consequent decrease in ciency, cystic fibrosis or other chronic disease that predisposes
bacterial etiology in 265 cases of radiologically proven CAP for pneumonia, or becoming sick while in hospital or abroad.
cases [2], a finding consistent with other recent CAP etiology Patients over 16 years of age or parents/guardians of youn-
studies [1, 19]. This change in etiology may have an impact on ger patients signed a written informed consent. The Regional
various clinical signs and symptoms of CAP and may alter the Ethics Committee and the local Data Protection Officer ap-
diagnostic ability of clinical features and inflammatory proved the study.
markers. The diagnostic ability of inflammatory markers in
pediatric CAP in the post-pneumococcal vaccination era has Definition of outcomes
recently been reported [9], but as pointed out, clinical features
are viewed as crucial in diagnosing pediatric CAP. We there- A chest radiograph was taken at inclusion, blinded for clinical
fore wanted to examine the diagnostic ability of clinical fea- data, and examined independently by two study radiologists
tures combined with inflammatory markers in a population experienced in pediatric radiology [2]. Localized or interstitial
with a high pneumococcal vaccination rate. infiltrates were regarded as findings consistent with pneumo-
The present study was performed in a routine clinical set- nia except for radiographs with only perihilar changes [5]. To
ting, examining previously healthy children and adolescents. increase specificity, only findings identified by both radiolo-
Childhood immunization rates are >90% in the Norwegian gists were labeled positive, as interrater variability can be sub-
pediatric population [38]. The 7-valent pneumococcal conju- stantial in pediatric chest radiography, especially in non-
gate vaccine was introduced in the Norwegian Childhood alveolar findings [5].
Immunization program in 2006 and replaced with the 13- To identify etiology, a number of microbiological diagnos-
valent vaccine in 2011. We aimed to find if clinical features, tic tests were performed and these tests and their results have
C-reactive protein (CRP), or white blood cell (WBC) count previously been published [2]. In brief, the microbiological
can (1) predict a chest radiography consistent with pneumonia workup consisted of (1) bacterial culture from blood (obtained
in clinically suspected CAP or (2) differentiate bacterial from in 83% of suspected CAP cases) and from pleural fluid (ob-
viral and/or atypical pneumonia in radiologically proven CAP. tained in seven patients where pleural tapping was clinically
Eur J Pediatr

indicated); (2) paired sera (obtained in 77% of suspected CAP Laboratory variables
cases) examined for serological evidence of recent infection
with respiratory syncytial virus (RSV) A/B, influenza virus CRP (mg/L) and WBC count (109/L), including differential
A/B, parainfluenza virus 1–3, adenovirus (all complement fix- count, were routinely analyzed on enrolment. WBC count is
ation tests), Mycoplasma pneumoniae, Chlamydophila presented as an absolute count and age-adjusted ratio (abso-
pneumoniae, and Streptococcus pneumoniae (ELISA for IgG lute WBC count divided by the patient’s age-specific upper
against pneumolysin and the novel flow cytometric analysis of reference range level <1 month 21.0, 2–5 months 19.5,
binding of serum antibodies to live pneumococci which was 6 months to 1 year 17.5, 2–5 years 17.0, 6–12 years 14.0,
strongly correlated to the more widely used pneumolysin and >13 years 13.0 × 109/L) [18], and neutrophils are present-
ELISA test) [2]; and (3) molecular diagnostic tests (PCR) of ed as the percentage of the total WBC count.
nasopharyngeal specimens (obtained in 97% of suspected CAP
cases) tested for RSV A/B, parainfluenza virus 1–4, influenza Analyses
virus A/B, human metapneumovirus, rhinovirus/enterovirus,
human bocavirus, adenovirus, M. pneumoniae, and All statistical analyses were done using IBM SPSS Statistics
C. pneumoniae (interpretation of viral PCR findings were done version 22. Significance levels were two sided and set at
with a strict cycle threshold cutoff of 35 to diminish false pos- p < 0.05. Categorical data was analyzed with chi-squared test.
itives). Based on a positive bacterial culture, a positive serolog- Continuous data were not normally distributed, hence present-
ical test and/or positive PCR according to previously described ed as median with interquartile range (IQR) as a measure of
diagnostic criteria [2], all CAP cases were categorized as (1) variation and analyzed with Mann-Whitney U test or Kruskal-
viral pneumonia without evidence of bacterial co-infection; (2) Wallis test. Cases with missing variables were not included in
atypical pneumonia, infections with M. pneumoniae and/or the statistical analyses.
C. pneumoniae, alone or co-infected with virus; or (3) bacterial Receiver operating characteristic (ROC) curves were used
pneumonia, infections with all other bacteria, predominantly to examine the discriminatory performance of a continuous
S. pneumoniae, alone or co-infected with virus. variable (CRP) in dichotomous outcomes (Bradiography pos-
itive versus radiography negative cases in suspected CAP^
and Bbacterial versus viral/atypical CAP^ with those missing
Clinical variables etiology excluded).
Logistic regression was performed to assess the ability
The attending physician completed a questionnaire on the pa- of several factors to predict the outcome Bpositive com-
tient’s medical history and clinical findings at recruitment. pared to negative chest radiography in suspected
Missing information was extracted from the hospital’s elec- pneumonia^ and to predict the outcome Bbacterial com-
tronic patient record. A certain and substantial fever was de- pared to viral and atypical CAP cases^ (performed sepa-
fined as body temperature ≥38.5 °C as measured at the hospi- rately and together, those with missing etiology excluded).
tal (Bosotherm Basic rectal thermometer, Bosch and Son, A simultaneous entry approach was used with a selection
Germany or Genius2 tympanic thermometer, Covidien, MA, of predictor variables on clinical grounds and in line with
USA) or as reported by patients (to reduce the impact of pre- previous literature [22, 25, 30]: presence of tachypnea,
vious antipyretic treatment). Respiratory rate was counted for localized fine crackles, localized reduced breath sounds,
one full minute and defined as tachypnea if <1 month >70 fever ≥38.5 °C, presence of hypoxemia (all categorical),
breaths/min, <1 year >50 breaths/min, <3 years >40 breaths/ CRP values, and age. CRP was chosen above WBC count
min, and >3 years >30 breaths/min [36]. Hypoxemia was de- in line with previous literature [39] and to avoid co-linear-
fined as peripheral oxygen saturation (SpO2) ≤92% measured ity. Interaction was checked between age and all other var-
by pulse oximetry (Dash 5000 patient monitor, GE iables, and significant interactions are reported.
Healthcare). Pediatric nurses in the pediatric emergency room Sensitivity, specificity, and likelihood ratios were calculat-
measured body temperature, respiratory rate, and SpO2. Chest ed with an online statistical calculator (www.medcalc.org).
retraction score from the Respiratory Distress Assessment
Instrument with assessment of supraclavicular, intercostal,
and subcostal retractions (one point for mild, two for moder- Results
ate, and three for marked retractions; maximum nine points)
was used as a measure of labored breathing and assessed by The patient cohort
the attending physician [24]. In addition, he/she recorded
cough and auscultatory findings, which are expiratory Of the patients with suspected CAP, 394 were enrolled, of
wheeze, localized fine crackles, and reduced breath sounds which 265 cases had radiological findings consistent with
either side. pneumonia (Fig. 1). Two hundred sixty one of the 265 had
 Eur J Pediatr

Fig. 1 Patient inclusion

focal radiological findings, and 4/265 had interstitial radio- Etiological distribution
logical changes. Fifteen of the 265 had parapneumonic ef-
fusions, of which 7 were regarded as complicated Of the 265 radiological proven cases of CAP, 63.4% were
parapneumonic effusion/empyema and tapped [3]. One- viral, 7.9% atypical bacterial, 12.8% bacterial causes (11.3%
hundred-twenty-nine patients did not have radiological pneumococcus, 9.4% with viral co-infection), and 15.8%
findings consistent with pneumonia (82 with normal chest without any proven microbiological agent [2]. The distribu-
radiograph and 42 with only perihilar involvement, 5 with tion of viral, atypical, and bacterial CAP according to place of
other findings, and all 129 cases are denoted as radiography treatment was significantly different (p = 0.03, chi-squared
negative in the following). Details on demographic, clini- test), with a lower proportion of bacterial cases in primary care
cal, and laboratory characteristics by age in all 265 CAP and a lower proportion of atypical cases in the hospitalized
cases compared to the 129 radiography negative cases are group (viral, atypical, and bacterial CAP in primary care pa-
presented in Table 1. Univariate analyses show that chil- tients 63.3, 16.7, and 3.3%, respectively; in ambulatory treat-
dren with radiological proven CAP were older; had higher ed patients 57.6, 12.1, and 12.1%; and hospitalized patients
rates of hospitalization, hypoxemia, and localized reduced 65.7, 4.7, and 14.8%). Clinical and laboratory characteristics
breath sounds; and had higher CRP values and percentage by etiology are presented in Table 2, and univariate analyses
of neutrophils compared to the radiography negative cases. show that (1) atypical pneumonia was associated with older
Seven of the 265 cases were <3 months old. The 84.9% of patients, fewer hospitalizations, and longer duration of disease
the CAP cases and 89.8% of the radiography negative cases and with a tendency of less often being tacyhpnoeic; (2) viral
had received one or more doses of pneumococcal conjugate pneumonia was associated with being hypoxemic and a great-
vaccine. Of the 394 suspected CAP cases, 220 were hospi- er degree of dyspnea; and (3) bacterial pneumonia was asso-
talized, 104 treated ambulatory at hospital level, and 68 ciated with higher CRP values and WBC count. CAP cases
treated in primary care (Fig. 1). Clinical data were missing without wheeze (217/265) showed a similar distribution of
in less than 5/394 cases for most variables, for cough in 14/ etiological agents: viral pneumonia 60.9%, atypical pneumo-
394, and for measured temperature in 12/394 cases. nia 8.8%, and bacterial pneumonia 14%, as did those with
Eur J Pediatr

Table 1 Demographic, clinical, and laboratory characteristics at inclusion of 265 cases of radiologically confirmed CAP by age, compared to 129
radiograph negative cases

<2 years 2–5 years >5 years Significancea All CAP Radiograph Significancec
(n = 127) (n = 98) (n = 40) cases negative
(n = 265) cases (n = 129)

Age in years, median (IQR) 1.2 (0.6–1.6) 2.8 (2.3–3.7) 9 (6.4–14.9) 2 (1.3–3.6) 1.7 (1.2–3) p = 0.044d, *
Male gender 73 (57.5%) 46 (46.9%) 25 (62.5%) p = 0.15 144 (54.3%) 76 (60%) p = 0.39
Hospitalized 93 (73.2%) 58 (59.2%) 18 (45%) p = 0.003* 169 (63.8%) 51 (39.5%) p < 0.001*
Days sick on inclusion, 4 (2–5) 4 (2–6) 3 (2–6) p = 0.55b 4 (2–6) 4 (2–5.75) p = 0.57d
median (IQR)
Fever ≥38.5 °C 104 (81.9%) 79 (80.6%) 31 (77.5%) p = 0.91 214 (80.8%) 96 (74.4%) p = 0.44
Cough 111 (87.4%) 90 (91.8%) 36 (90%) p = 0.72 237 (89.4%) 116 (89.9%) p = 0.96
Tachypnea, age specific 102 (80.3%) 75 (76.5%) 24 (60%) p = 0.014* 201 (75.8%) 89 (70%) p = 0.11
Hypoxemia (SpO2 ≤92%) 34 (26.8%) 31 (31.6%) 4 (10%) p = 0.026* 69 (26%) 15 (11.6%) p = 0.001*
Chest retraction score, 2 (0–4) 2 (0–4) 0 (0–0) p < 0.001b, * 1 (0–3) 1 (0–2) p = 0.33d
median (IQR)
Auscultatory findings
Wheeze 25 (19.7%) 21 (21.4%) 2 (5%) p = 0.059 48 (18.1%) 28 (21.7%) p = 0.23
Fine crackles 31 (24.4%) 20 (20.4%) 10 (25%) p = 0.75 61 (23%) 27 (20.9%) p = 0.67
Reduced breath sounds 11 (8.7%) 19 (19.4%) 15 (37.5%) p < 0.001* 45 (17%) 11 (8.5%) p = 0.025*
Laboratory findings
CRP (mg/L), median (IQR) 75 (30–160) 90 (26–190) 90 (45–240) p = 0.86 b 80(32–190) 42 (13–100) p < 0.001d, *
WBC count (×109/L), 11.6 11.4 9.2 p < 0.001b, * 12.2 11.3 (8.6–15.8) p = 0.17d
median (IQR) (9.4–17.6) (7.8–17.6) (6.2–15.3) (8.9–18.1)
Age-adjusted WBC count 29 (22.8%) 30 (32.7%) 16 (40%) p = 0.11b 75 (28.7%) 22 (17.1%) p = 0.14d
elevatione
Percentage neutrophils, 49 (38–66) 67(52–77) 71 (65–87) p < 0.001b, * 64 (46–76) 52 (37–67) p < 0.001d,*
median (IQR)

All numbers given as n (percent of total in each column), except where otherwise indicated
*Significant differences
a
Comparison between the three age groups. All significance levels are two sided and all analyses are chi-squared test except where otherwise indicated
b
Kruskal-Wallis test
c
Comparison between all CAP cases and other ALRI cases. All significance levels are two sided and all analyses are chi-squared test except where
otherwise indicated
d
Mann-Whitney U test
e
Proportion of cases with age-adjusted WBC count ratio >1

tachypnea (201/265): viral pneumonia 65.4%, atypical pneu- associated with positive chest radiography (Table 1), while
monia 6%, and bacterial pneumonia 13.9%. all three WBC count variables were significantly associated
with etiology (Table 2).

Inflammatory markers
Logistic regression analyses and test characteristics
As stated above, higher median CRP values were significantly
associated with both a bacterial cause and a chest radiograph The full logistic regression models for predicting chest radi-
consistent with pneumonia (Tables 1 and 2). A ROC curve ography consistent with pneumonia and in predicting bacterial
provided an area under the curve (AUC) of 0.72 for CRP in CAP compared to viral, atypical, and to viral/atypical CAP
separating bacterial from viral or atypical CAP, while a ROC were all significant (p < 0.001). Three variables provided
curve for CRP’s discriminatory ability for positive versus neg- unique statistically significant contributions to the models,
ative chest radiograph gave an AUC of 0.64 (Fig. 2). A CRP which are increasing CRP (in four/four models), presence/
cutoff value of 80 mg/L corresponded well with the best sen- absence of hypoxemia (in three/four models), and older age
sitivity and specificity trade-off in both curves. In univariate (in two/four models). Table 3 shows the predictive abilities of
analysis, only percentage of neutrophils was significantly the seven independent variables in the four logistic regression
 Eur J Pediatr

Table 2 Clinical features and inflammatory markers at inclusion by etiology of 265 cases of radiological confirmed CAP

Viral pneumonia Atypical pneumonia Bacterial pneumonia Significancea No cause found

(n = 168) (n = 21) (n = 34) (n = 42)

Age in years, median (IQR) 1.8 (1.1–2.7) 9.6 (5–14.3) 1.7 (1.1–2.9) p < 0.001b, * 3.3 (1.8–7.3)
Male gender 92 (54.8%) 11 (52.4%) 19 (55.9%) p = 0.97 22 (53.4%)
Hospitalized 111 (66.1%) 8 (38.1%) 25 (73.5%) p = 0.020 24 (57.1%)
Days sick on inclusion, median (IQR) 4 (2–6) 5 (3–9.5) 3 (2–4) p = 0.014b, * 2 (2–4)
Fever ≥38.5 °C 135 (80.4%) 16 (76.2%) 31 (91.2%) p = 0.39 32 (76.2%)
Cough 154 (92.3%) 19 (90.5%) 29 (85.3%) p = 0.45 35 (83.3%)
Tachypnea, age specific 132 (78.6%) 12 (57.1%) 28 (82.4%) p = 0.053 29 (69.1%)
Hypoxemia (SpO2 ≤92%) 54 (32.1%) 4 (19.1%) 3 (8.8%) p = 0.013* 8 (19.1%)
Chest retraction score, median (IQR) 2 (0–4) 0 (0–1) 0.5 (0–3) p = 0.002b, * 0 (0–2)
Auscultatory findings
Wheeze 36 (21.4%) 2 (9.5%) 4 (11.8%) p = 0.21 6 (14.3%)
Fine crackles, localized 41 (24.4%) 6 (28.6%) 6 (17.6%) p = 0.60 8 (19.0%)
Reduced breath sounds, localized 22 (13.1%) 6 (28.6%) 8 (23.5%) p = 0.093 9 (21.4%)
Laboratory findings
CRP (mg/L), median (IQR) 70 (27–150) 48 (29–105) 260 (54–320) p < 0.001b, * 160 (75–265)
WBC count (×109/L), median (IQR) 11.1 (8.5–16.1) 7.3 (6.2–12.1) 17.1 (10.5–23.2) p < 0.001b, * 16.6 (12.6–23.4)
Age-adjusted WBC count elevation 34 (20.2%) 3 (14.3%) 15 (47.1%) p = 0.005b, * 24 (57.1%)
Percentage neutrophils, median (IQR) 56 (43–72) 69 (59–71) 70 (50–81) p = 0.004b, * 75 (58–84)

All numbers given as n (percent of total in each column), except where otherwise indicated
*Significant differences
a
All significance levels are two sided and all statistical analyses are chi-squared test except where otherwise indicated
b
Kruskal-Wallis test

models. To mimic the diagnostic process, logistic regression while sensitivity ranged from 4 to 46% when combining two
analyses for all four models were also performed in a two-step or all variables. Low peripheral oxygen concentration is in-
manner by first introducing the clinical variables and then cluded as it is the one clinical variable with unique significant
CRP. This procedure gave similar results as when entering predictive ability in three of the four multivariate models.
clinical variables and CRP simultaneously.
Table 4 provides sensitivity, specificity, and likelihood ra-
tios for CRP, age-adjusted WBC count, and the presence of Discussion
SpO2 ≤92%, alone and in combination. Specificity ranged
from 85 to 98% and positive likelihood ratio from 2.4 to 3.3 In this study, we sought to find clinical features and inflam-
when combining two or all of the variables in both outcomes, matory markers that could aid the clinician in predicting the

Fig. 2 Receiver operating

characteristics curves for CRP’s
diagnostic ability in the two main
outcomes. AUC area under the
curve
Eur J Pediatr

Table 3 Logistic regression predicting likelihood of radiography consistent with pneumonia in suspected CAP and bacterial infection in radiological
proven CAP

Positive chest radiography Bacterial (n = 34) Bacterial (n = 34) Bacterial (n = 34)

(n = 265) versus negative versus viral versus atypical versus viral/atypical
(n = 129) (n = 168) CAP (n = 21) CAP CAP (n = 189)
OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)

Fever ≥38.5 °C 1.15 (0.64–2.08) 0.81 (0.21–3.20) 0.91 (0.06–14.57) 0.84 (0.22–3.29)
Reduced breath sounds 1.47 (0.68–3.21) 1.41 (0.42–4.70) 5.10 (0.25–104.33) 1.53 (0.48–4.93)
Fine crackles 0.98 (0.56–1.71) 0.73 (0.25–2.16) 0.22 (0.02–2.33) 0.69 (0.24–2.00)
Tachypnea 1.61 (0.93–2.80) 1.64 (0.57–4.73) 2.51 (0.24–26.30) 1.62 (0.56–4.65)
Hypoxemia 2.71 (1.42–5.18)* 0.22 (0.06–0.80)* 0.13 (0.01–2.67) 0.23 (0.06–0.82)*
CRP/10a 1.06 (1.03–1.09)* 1.09 (1.05–1.13)* 1.10 (1.01–1.20)* 1.09 (1.05–1.14)*
Age in years 1.09 (1.00–1.18)* 0.96 (0.78–1.18) 0.62 (0.44–0.88)* 0.91 (0.76–1.08)

*Variables providing unique statistically significant contributions to the models, CRP (in four/four models), presence/absence of hypoxemia (in three/
four models), and age (in two/four models)
a
OR for the continuous variable CRP/10 denotes the increase in odds ratio for every increase of 10 in CRP

presence of infiltrates in suspected CAP and in predicting difference in the likelihood of any of the two outcomes is
bacterial etiology in radiological proven CAP. The non- achieved. Hypoxemia at inclusion (SpO2 ≤92%) was the only
specific inflammatory markers CRP and WBC count were in clinical feature significantly predicting the two main outcomes
a combination of univariate and multivariate analyses found to in both univariate and multivariate analyses, but hypoxemia
consistently and significantly predict a positive chest radio- was only found in less than one third in all subgroups.
graph in clinically suspected pneumonia and bacterial etiology Our findings indicate that bacterial pneumonia is associated
in radiologically proven pneumonia (Tables 1, 2, and 3). The with a greater degree of inflammation. This is in line with
low sensitivity for both CRP and WBC count hampers the studies performed both before and after implementation of
clinical utility of these two laboratory tests often used in clin- routine pneumococcal immunization [9, 10, 12, 21, 29, 32,
ical practice. Furthermore, for CRP, a relatively low but yet 37, 40]. In the ROC curve for CRP’s ability to diagnose bac-
significant OR for every increment of 10 mg/L indicates that a terial pneumonia, a CRP value of 80 mg/L gave the best sen-
substantial increase is needed before a clinical relevant sitivity and specificity trade-off. This corresponds to a

Table 4 Test characteristics in the main outcomes

Sensitivity (95% CI) Specificity (95% CI) LR+a (95% CI) LR−b (95% CI)

Chest radiography consistent with pneumonia in suspected CAP

CRP >80c 55 (49–61) 66 (57–74) 1.6 (1.3–2.1) 0.7 (0.6–0.8)
Elevated WBC countd 29 (23–35) 83 (75–89) 1.7 (1.1–2.6) 0.9 (0.8–1)
CRP >80 and elevated WBC count 25 (20–30) 92 (86–96) 3.1 (1.7–5.9) 0.8 (0.7–0.9)
Hypoxemia presente 27 (22–33) 88 (81–93) 2.2 (1.3–3.8) 0.8 (0.7–0.9)
Hypoxemia present, CRP >80, and elevated WBC count 4 (2–7) 98 (94–100) 2.4 (0.5–11) 1 (0.9–1)
Bacterial versus viral/atypical in proven CAP
CRP >80c 71 (53–85) 52 (44–59) 1.5 (1.1–1.9) 0.6 (0.3–1)
Elevated WBC countd 46 (28–64) 80 (74–87) 2.3 (1.4–3.7) 0.7 (0.5–1)
CRP >80 and elevated WBC count 46 (28–64) 85 (79–90) 3 (1.8–5) 0.6 (0.5–0.9)
Hypoxemia absente 91 (76–98) 31 (25–39) 1.3 (1.2–1.5) 0.3 (0.1–0.9)
Hypoxemia absent, CRP >80, and elevated WBC count 39 (23–58) 88 (82–92) 3.3 (1.8–5.8) 0.7 (0.5–0.9)
a
LR+ likelihood ratio of positive test; sensitivity/(1 − specificity), i.e., denotes ratio between true positive and false positive test results
b
LR− likelihood ratio of negative test; 1 − sensitivity/specificity, i.e., denotes ratio between false negative and true negative
c
Cutoff CRP value of 80 selected from the ROC curves considering best sensitivity/specificity trade-off
d
Age-adjusted WBC counts; elevated denotes above upper age-specific reference level
e
The presence of hypoxemia is a positive predictor for the outcome positive chest radiography in suspected CAP but a negative predictor in bacterial
proven CAP versus viral/atypical CAP as seen in Table 3
 Eur J Pediatr

systematic review on the diagnostic value of laboratory tests etiology or radiographic findings in either univariate or mul-
in febrile children [39] but is higher than the cutoff used in a tivariate analyses. Previous studies have found older age to be
meta-analysis of CRP in pediatric CAP [11]. In one systematic associated with bacterial CAP [9, 22], but in our study only
review, WBC count added little extra diagnostic value [39], associated with atypical pneumonia.
but by combining CRP and age-related elevated WBC count, A major strength of this study is that both clinical features
we improved the test characteristics. High values of CRP and laboratory findings are first analyzed in suspected and
(>80 mg/L) and/or elevated WBC count can significantly rule then in proven CAP cases. By this two-step approach, we
in bacterial pneumonia (specificity >85% and positive likeli- simulate the diagnostic approach in the pediatric emergency
hood ratio >3; Table 4), as also found by others [9, 12, 37]. room. The findings reported here should reflect the epidemi-
Poor sensitivity reduces their clinical value, as also concluded ology of pediatric CAP in our and socio-demographically
in a previous systematic review [28]. In predicting positive similar regions. With few bacterial pneumonia cases and few
chest radiography in cases of suspected CAP, a similar statis- school-aged children with pneumonia, conclusions in these
tical association with relatively high specificity and positive groups must be interpreted cautiously. Furthermore, and as
likelihood ratios, but with limited clinical utility due to low discussed in our previous publication [2], there are several
sensitivity, were found for both CRP and WBC count, in line obstacles in the microbiological diagnosis of pneumonia, in-
with a previous study [23]. Although CRP was associated troducing some uncertainty in our etiological classification
with etiology, the cases treated only in primary care were too and hence influencing the ability of clinical and laboratory
few to conclude if our findings collide with a previous primary features in predicting etiology. On the other hand, the validity
care study that found no association between CRP and etiol- of our etiological results is corroborated by similar results in a
ogy [17]. Furthermore, we found fewer bacterial cases in the recent, large US multicenter CAP study [19]. We believe that
primary care group, in line with a study that found predomi- combining a variety of univariate and multivariate statistical
nantly viral cause in ALRI, where the primary care physician analyses makes our findings more robust. The present study
considered antibiotics [14]. includes patients at all treatment levels including primary care,
Previous studies on diagnostic ability of clinical features increasing the usefulness of our study for clinical decisions
have focused on their ability to either predict radiographic concerning admission and antibiotic use. Caution should be
pneumonia [25, 26, 30, 34, 41] or differentiate etiology [10, taken, however, when making direct comparison to CAP stud-
22, 40]. Other studies found reduced peripheral oxygen con- ies of only hospitalized patients. In order to make an easily
centration as one of several clinical features associated with a communicable pneumonia definition, intention to treat with
positive chest radiograph [7, 25, 26, 30, 31], while in our antibiotics for suspected pneumonia in primary care was set
cohort with a high coverage of pneumococcal vaccination as an inclusion criterion in these patients. This may have lead
and with a large impact of viral etiology, this was the only to selection bias compared to patients included by hospital
clinical feature associated with a positive chest radiograph. In pediatricians. Measurements of clinical features were done
differentiating between etiologies in proven CAP, our findings by staff on call, and although all staff there are trained in these
are in line with previous literature; most clinical features did routine measurements, we cannot exclude errors in data col-
not distinguish significantly between viral, atypical, or bacte- lection. Due to the lack of a uniform clinical definition [27],
rial pneumonia [6, 20, 22, 27]. Only the presence of hypox- comparison with other studies may also be hampered as pneu-
emia at inclusion was significantly associated with viral etiol- monia definition varies. Our approach in only assigning radio-
ogy. Age-related tachypnea, a cornerstone in the WHO clini- graphic findings seen by both radiologists as positive to re-
cal case definition, did not predict radiological confirmed duce interrater variability and increase specificity may have
CAP, a finding consistent with other studies [16, 34, 35, 41]. been at the cost of reduced sensitivity. Our low proportion of
Perhaps more surprisingly in our cohort, wheeze was equally interstitial infiltrates may reflect this. Our one-center setting is
prevalent across etiological and radiological categories and a limitation, and representativeness must be kept in mind
hence not associated with either of our outcomes. This is in when applying our results to socio-demographically similar
contrast to previous studies, finding wheeze as a negative populations.
predictor of positive chest radiography in suspected CAP In conclusion, in this study of pediatric pneumonia in a
[30] and associated with viral CAP by others [29]. The new population with a high coverage of pneumococcal vaccina-
epidemiological situation with a large impact of viral infec- tion, clinical features are of little diagnostic value, in accor-
tions may explain wheeze’s lack of predictive ability, and the dance with previous studies. An exception is reduced periph-
relatively high proportion of bacterial-viral co-infection may eral oxygen saturation at admission, which is predictive of
additionally obscure its predictive ability in differentiating eti- viral etiology. Elevated CRP and WBC counts gave relatively
ology. Fever has been reported as the most consistent clinical high specificity and positive likelihood ratios and may thus be
feature in pediatric CAP [8]. In our study, fever ≥38.5 °C was beneficial in (i) ruling in positive chest radiography in
a common finding but not significantly associated with suspected CAP and (ii) ruling in bacterial etiology in
Eur J Pediatr

radiological proven CAP. Although poor sensitivity, and a 2. Berg AS, Inchley CS, Aase A, Fjaerli HO, Bull R, Aaberge I,
Leegaard TM, Nakstad B (2016a) Etiology of pneumonia in a pe-
relatively low but significant OR for CRP, reduces their clin-
diatric population with high pneumococcal vaccine coverage: a
ical usefulness, our findings may contribute to reductions in prospective study. Pediatr Infect Dis J 35(3):e69–e75. doi:10.
antibiotic and chest radiography use in resource-rich settings 1097/INF.0000000000001009
with good routine immunization programs. In cases with low 3. Berg AS, Inchley CS, Fjaerli HO, Leegaard TM, Nakstad B
(2016b) Microbial aetiology of paediatric pneumonia complicated
CRP and low white blood cell counts, a watch-and-wait strat-
with parapneumonic effusion in the era of pneumococcal vaccina-
egy in children with suspected or proven pneumonia may be tion. Infect Dis (Lond) 48(9):712–714. doi:10.1080/23744235.
applied without detrimental effects. 2016.1192721
4. Bradley JS, Byington CL, Shah SS, Alverson B, Carter ER,
Acknowledgements We are grateful to all participating children and Harrison C, Kaplan SL, Mace SE, McCracken GH Jr, Moore
their parents and nurses and doctors on call for recruiting patients and MR, St Peter SD, Stockwell JA, Swanson JT, Pediatric Infectious
collecting data. We thank Drs. R Bull and R Grotli at the Department for Diseases S, the Infectious Diseases Society of A (2011) The man-
Radiology, Akershus University Hospital, for performing the radiological agement of community-acquired pneumonia in infants and children
examinations; Dr. I Aaberge and A Aase for the pneumococcal serolog- older than 3 months of age: clinical practice guidelines by the
ical analyses at the Department of Bacteriology and Immunology, the Pediatric Infectious Diseases Society and the Infectious Diseases
Norwegian Institute for Public Health; and Dr. Line Sletner at our depart- Society of America. Clin Infect Dis 53(7):e25–e76. doi:10.1093/
ment for the advise on statistical analyses and staff at the laboratories at cid/cir531
Akershus University Hospital for performing the laboratory analyses. 5. Cherian T, Mulholland EK, Carlin JB, Ostensen H, Amin R, de
Campo M, Greenberg D, Lagos R, Lucero M, Madhi SA,
Authors’ contributions All authors provided substantial contributions O’Brien KL, Obaro S, Steinhoff MC (2005) Standardized interpre-
to the study’s conception and design and acquisition, analysis, or inter- tation of paediatric chest radiographs for the diagnosis of pneumo-
pretation of the data. nia in epidemiological studies. Bull World Health Organ 83(5):
Drs. Berg, Inchley, Fjaerli, and Nakstad executed the clinical part of 353–359
the work. 6. Chiappini E, Venturini E, Galli L, Novelli V, de Martino M (2013)
Dr. Leegaard was responsible for the microbiological laboratory Diagnostic features of community-acquired pneumonia in children:
analyses. what’s new? Acta Paediatr Suppl 102(465):17–24. doi:10.1111/apa.
Dr. Lindbaek was responsible for the primary care part of the study. 12502
Dr. Berg drafted the initial manuscript, and all other authors revised it 7. Craig JC, Williams GJ, Jones M, Codarini M, Macaskill P, Hayen
critically for important intellectual content. All authors approved the final A, Irwig L, Fitzgerald DA, Isaacs D, McCaskill M (2010) The
manuscript as submitted and agreed to be accountable for all aspects of accuracy of clinical symptoms and signs for the diagnosis of serious
the work. bacterial infection in young febrile children: prospective cohort
study of 15 781 febrile illnesses. BMJ 340:c1594. doi:10.1136/
bmj.c1594
Compliance with ethical standards 8. Don M, Canciani M, Korppi M (2010) Community-acquired pneu-
monia in children: what’s old? What’s new? Acta Paediatr 99(11):
Conflict of interest The authors declare that they have no conflicts of 1602–1608. doi:10.1111/j.1651-2227.2010.01924.x
interest. 9. Elemraid MA, Rushton SP, Thomas MF, Spencer DA, Gennery
AR, Clark JE (2014) Utility of inflammatory markers in predicting
Ethical approval All procedures performed in studies involving hu- the aetiology of pneumonia in children. Diagn Microbiol Infect Dis
man participants were in accordance with the ethical standards of the 79(4):458–462. doi:10.1016/j.diagmicrobio.2014.04.006
institutional and national research committee and approved by both the 10. Esposito S, Bosis S, Cavagna R, Faelli N, Begliatti E, Marchisio P,
Regional Ethics Committee and the local Data Protection Officer. Blasi F, Bianchi C, Principi N (2002) Characteristics of
Furthermore, the study was in accordance with the 1964 Helsinki decla- Streptococcus pneumoniae and atypical bacterial infections in chil-
ration and its later amendments or comparable ethical standards. dren 2–5 years of age with community-acquired pneumonia. Clin
Infect Dis 35(11):1345–1352. doi:10.1086/344191
11. Flood RG, Badik J, Aronoff SC (2008) The utility of serum C-
Informed consent Informed consent was obtained from all individual
reactive protein in differentiating bacterial from nonbacterial pneu-
participants included in the study.
monia in children: a meta-analysis of 1230 children. Pediatr Infect
Dis J 27(2):95–99. doi:10.1097/INF.0b013e318157aced
Source of funding The study was supported by research grants from 12. Galetto-Lacour A, Alcoba G, Posfay-Barbe KM, Cevey-Macherel
Akershus University Hospital and South-Eastern Norway Regional M, Gehri M, Ochs MM, Brookes RH, Siegrist CA, Gervaix A
Health Authority and grants from the Grimsgaard Foundation and the (2013) Elevated inflammatory markers combined with positive
Norwegian Organization for Surveillance of Antimicrobial Resistance. pneumococcal urinary antigen are a good predictor of pneumococ-
cal community-acquired pneumonia in children. Pediatr Infect Dis J
32(11):1175–1179. doi:10.1097/INF.0b013e31829ba62a
13. Gove S (1997) Integrated management of childhood illness by out-
References patient health workers: technical basis and overview. The WHO
Working Group on Guidelines for Integrated Management of the
1. Angoulvant F, Levy C, Grimprel E, Varon E, Lorrot M, Biscardi S, Sick Child. Bull World Health Organ 75(Suppl 1):7–24
Minodier P, Dommergues MA, Hees L, Gillet Y, Craiu I, Zenkhri F, 14. Harnden A, Perera R, Brueggemann AB, Mayon-White R, Crook
Dubos F, Guen CG, Launay E, Martinot A, Cohen R (2014) Early DW, Thomson A, Mant D (2007) Respiratory infections for which
impact of 13-valent pneumococcal conjugate vaccine on general practitioners consider prescribing an antibiotic: a prospec-
community-acquired pneumonia in children. Clin Infect Dis tive study. Arch Dis Child 92(7):594–597. doi:10.1136/adc.2007.
58(7):918–924. doi:10.1093/cid/ciu006 116665
 Eur J Pediatr

15. Harris M, Clark J, Coote N, Fletcher P, Harnden A, McKean M, lower respiratory tract: systematic review. BMJ 331(7507):26. doi:
Thomson A (2011) British Thoracic Society guidelines for the man- 10.1136/bmj.38483.478183.EB
agement of community acquired pneumonia in children: update. 29. Michelow IC, Olsen K, Lozano J, Rollins NK, Duffy LB, Ziegler T,
Thorax 66(Suppl 2):ii1–i23. doi:10.1136/thoraxjnl-2011-200598 Kauppila J, Leinonen M, McCracken GH Jr (2004) Epidemiology
16. Hazir T, Nisar YB, Qazi SA, Khan SF, Raza M, Zameer S, Masood and clinical characteristics of community-acquired pneumonia in
SA (2006) Chest radiography in children aged 2–59 months diag- hospitalized children. Pediatrics 113(4):701–707
nosed with non-severe pneumonia as defined by World Health 30. Neuman MI, Monuteaux MC, Scully KJ, Bachur RG (2011)
Organization: descriptive multicentre study in Pakistan. BMJ Prediction of pneumonia in a pediatric emergency department.
333(7569):629. doi:10.1136/bmj.38915.673322.80 Pediatrics 128(2):246–253. doi:10.1542/peds.2010-3367
17. Heiskanen-Kosma T, Korppi M (2000) Serum C-reactive protein 31. Oostenbrink R, Thompson M, Lakhanpaul M, Steyerberg EW,
cannot differentiate bacterial and viral aetiology of community- Coad N, Moll HA (2013) Children with fever and cough at emer-
acquired pneumonia in children in primary healthcare settings. gency care: diagnostic accuracy of a clinical model to identify chil-
Scand J Infect Dis 32(4):399–402 dren at low risk of pneumonia. Eur J Emerg Med 20(4):273–280.
18. Hinchliffe R (1992) Reference values. In: Lilleyman J, Hann I (eds) doi:10.1097/MEJ.0b013e32835771fd
Pediatric hematology, 2nd edn. Churchill Livingstone, Edinburgh, 32. Prat C, Dominguez J, Rodrigo C, Gimenez M, Azuara M, Jimenez
pp 1–22 O, Gali N, Ausina V (2003) Procalcitonin, C-reactive protein and
19. Jain S, Williams DJ, Arnold SR, Ampofo K, Bramley AM, Reed C, leukocyte count in children with lower respiratory tract infection.
Stockmann C, Anderson EJ, Grijalva CG, Self WH, Zhu Y, Patel A, Pediatr Infect Dis J 22(11):963–968. doi:10.1097/01.inf.
Hymas W, Chappell JD, Kaufman RA, Kan JH, Dansie D, Lenny 0000095197.72976.4f
N, Hillyard DR, Haynes LM, Levine M, Lindstrom S, Winchell JM, 33. Rambaud-Althaus C, Althaus F, Genton B, D’Acremont V (2015)
Katz JM, Erdman D, Schneider E, Hicks LA, Wunderink RG, Clinical features for diagnosis of pneumonia in children younger
Edwards KM, Pavia AT, McCullers JA, Finelli L, Team CES than 5 years: a systematic review and meta-analysis. Lancet Infect
(2015) Community-acquired pneumonia requiring hospitalization Dis 15(4):439–450. doi:10.1016/s1473-3099(15)70017-4
among U.S. children. N Engl J Med 372(9):835–845. doi:10. 34. Rothrock SG, Green SM, Fanelli JM, Cruzen E, Costanzo KA,
1056/NEJMoa1405870 Pagane J (2001) Do published guidelines predict pneumonia in
20. Juven T, Mertsola J, Toikka P, Virkki R, Leinonen M, Ruuskanen O children presenting to an urban ED? Pediatr Emerg Care 17(4):
(2001) Clinical profile of serologically diagnosed pneumococcal 240–243
pneumonia. Pediatr Infect Dis J 20(11):1028–1033 35. Shah S, Bachur R, Kim D, Neuman MI (2010) Lack of predictive
value of tachypnea in the diagnosis of pneumonia in children.
21. Korppi M, Heiskanen-Kosma T, Leinonen M (1997) White blood
Pediatr Infect Dis J 29(5):406–409. doi:10.1097/INF.
cells, C-reactive protein and erythrocyte sedimentation rate in pneu-
0b013e3181cb45a7
mococcal pneumonia in children. Eur Respir J 10(5):1125–1129
36. Solevag AL, Eggen EH, Schroder J, Nakstad B (2013) Use of a
22. Korppi M, Don M, Valent F, Canciani M (2008) The value of
modified pediatric early warning score in a department of pediatric
clinical features in differentiating between viral, pneumococcal
and adolescent medicine. PLoS One 8(8):e72534. doi:10.1371/
and atypical bacterial pneumonia in children. Acta Paediatr 97(7):
journal.pone.0072534
943–947. doi:10.1111/j.1651-2227.2008.00789.x
37. Toikka P, Irjala K, Juven T, Virkki R, Mertsola J, Leinonen M,
23. Koster MJ, Broekhuizen BD, Minnaard MC, Balemans WA, Ruuskanen O (2000) Serum procalcitonin, C-reactive protein and
Hopstaken RM, de Jong PA, Verheij TJ (2013) Diagnostic proper- interleukin-6 for distinguishing bacterial and viral pneumonia in
ties of C-reactive protein for detecting pneumonia in children. children. Pediatr Infect Dis J 19(7):598–602
Respir Med 107(7):1087–1093. doi:10.1016/j.rmed.2013.04.012 38. Trogstad L, Ung G, Hagerup-Jenssen M, Cappelen I, Haugen IL,
24. Lowell DI, Lister G, Von Koss H, McCarthy P (1987) Wheezing in Feiring B (2012) The Norwegian immunisation register—
infants: the response to epinephrine. Pediatrics 79(6):939–945 SYSVAK. Euro Surveill 17 (16)
25. Lynch T, Platt R, Gouin S, Larson C, Patenaude Y (2004) Can we 39. Van den Bruel A, Thompson MJ, Haj-Hassan T, Stevens R, Moll H,
predict which children with clinically suspected pneumonia will Lakhanpaul M, Mant D (2011) Diagnostic value of laboratory tests
have the presence of focal infiltrates on chest radiographs? in identifying serious infections in febrile children: systematic re-
Pediatrics 113(3 Pt 1):e186–e189 view. BMJ 342:d3082. doi:10.1136/bmj.d3082
26. Mahabee-Gittens EM, Grupp-Phelan J, Brody AS, Donnelly LF, 40. Virkki R, Juven T, Rikalainen H, Svedstrom E, Mertsola J,
Bracey SE, Duma EM, Mallory ML, Slap GB (2005) Identifying Ruuskanen O (2002) Differentiation of bacterial and viral pneumo-
children with pneumonia in the emergency department. Clin Pediatr nia in children. Thorax 57(5):438–441
(Phila) 44(5):427–435 41. Wingerter SL, Bachur RG, Monuteaux MC, Neuman MI (2012)
27. McIntosh K (2002) Community-acquired pneumonia in children. N Application of the world health organization criteria to predict ra-
Engl J Med 346(6):429–437. doi:10.1056/NEJMra011994 diographic pneumonia in a US-based pediatric emergency depart-
28. van der Meer V, Neven AK, van den Broek PJ, Assendelft WJ ment. Pediatr Infect Dis J 31(6):561–564. doi:10.1097/INF.
(2005) Diagnostic value of C reactive protein in infections of the 0b013e31824da716