282 The Bacterial Meningitis Score

Original Article

External validation of the Bacterial

Meningitis Score in children
hospitalized with meningitis
Tuerlinckx D1, El Hayeck J1, Van der Linden D, Bodart E1, Glupczynski Y
1
Dpartement de Pdiatrie, Universit Catholique de Louvain, Cliniques Universitaires de Mont
Godinne, Yvoir, Belgium, Dpartement de Pdiatrie, Universit Catholique de Louvain,
Cliniques Universitaires Saint Luc, Brussels, Belgium and Laboratoire de Microbiologie,
Universit Catholique de Louvain, Cliniques Universitaires de Mont Godinne, Yvoir, Belgium

Correspondence and offprint requests to: David Tuerlinckx, Email: david.tuerlinckx@uclouvain.be

Abstract
The Bacterial Meningitis Score (BMS) is considered as
the rule with the highest sensitivity to safely distinguish
between aseptic and bacterial meningitis (BM).
Objective: The objective of our study was to evalu-
ate the performance of the score and its usefulness for
the clinician.
Method: Retrospective analysis of two Belgian aca-
demic hospitals-based cohort studies. All consecutive
children aged 29days to 18years admitted for acute
meningitis between January 1996 and December 2008
was eligible. The BMS (risk of bacterial meningitis if sei-
zure, positive cerebrospinal fluid (CSF) Gram staining, CSF
protein level 80mg/dl, CSF neutrophil count 1,000/
mm or blood neutrophil count 10,000/mm) was
applied to all patients with meningitis defined by CSF
pleocytosis >8 WBC/mm.
Results: 174 patients were included in the final anal-
ysis of whom 26 (15%) had BM. Of the 93 patients cate-
gorized as having with no risk for BM (BMS score=0),
2patients had BM, one of which had petechial rash (neg-
ative predictive value 97.8%). BMS had a sensitivity of
92.3%. Risk of BM was significantly related to the BMS
score: 6/147 (4%) patients with BMS 1 had BM com-
pared to 20/27 (74%) patients with BMS >1.
Conclusions: Our study reports a lower sensitivity of
the BMS than observed in previous validation studies. We
suggest to include the BMS in a decision tree aiming to
optimize the ordering of laboratory investigations includ-
ing viral and bacterial PCR testing in any child with CSF
pleocytosis.
Keys words: aseptic meningitis, bacterial meningitis, decision rules
Introduction
Distinguishing between bacterial (BM) and aseptic men-
ingitis (AM) is difficult in children. This situation leads the phy-
sician to start the empiric administration of antibiotics imme-
diately after making the diagnosis of acute meningitis based
on clinical signs and/or cerebrospinal fluid (CSF) pleocytosis
and to continue them until BM is definitively ruled out (i.e.
until negative culture or PCR results as well as good clinical
evolution) (1, 2).
Even with optimal treatment, patients with BM are
exposed to the risk of neurological sequelae especially when
diagnosis and antibiotic administration are delayed (3, 4, 5, 6).
The strategy of treating all acute meningitis ensures the
swift treatment of children with BM. However in an era of rou-
tine immunization with protein conjugate vaccines against
Haemophilus influenzae type b, Streptococcus pneumoniae and
Neisseria meningitidis group C, BM only represents 3.7% to 6%
of all episodes of acute meningitis in western countries beyond
the neonatal period (7, 8). Enabling the differentiation between
BM and AM could help to reduce unnecessary antibiotic use
and hospital admissions for viral meningitis. Because of the risk
of erroneous decisions and the potentially dramatic conse-
quences that a delayed diagnosis of BM may carry, any diag-
nostic tool should achieve a sensitivity close to 100% (9).
Currently, no single clinical nor laboratory criterion can
achieve alone the objective of distinguishing between BM
and AM with 100% sensitivity and with a high specificity.
Therefore, several clinical decision rules based on a combina-
tion of clinical and biological data have been described in the
literature (10). However only one rule, the Bacterial Meningitis
Score (BMS), developed by Nigrovic et al (11), has satisfied the
methodological standards and quality criteria proposed by
the Evidence- Based Medicine Working Group as well as those
proposed by Stiell and Wells for emergency medicine (12, 13).
These rules have to address and fulfill three conditions;

Acta Clinica Belgica, 2012; 67-4 doi: 10.2143/ACB.67.4.2062673


namely to yield a 100% sensitivity for BM, with at the same
time the best specificity and to be easy to apply in clinical
settings. To date the BMS has been subjected to five external
evaluations (7, 8, 14, 15, 16). The objectives of our study were
twofolds: (1) to make an external evaluation of the BMS and
(2) to incorporate this score model into an algorithm aiming
to improve the ordering of diagnostic laboratory tests as well
as the management of the patients.
METHODS
Study design and setting
This retrospective cohort study comprised two cohorts of
hospitalized children aged 29days to 18years. The first
cohort included all children who were admitted for meningi-
tis between January 1996 and December 2006 at the Mont-
Godinne Hospital, Catholic University of Louvain, Belgium,
Yvoir. The second cohort included all children who were
admitted with a diagnosis of meningitis between January
2007 and December 2008 at the Saint-Luc Hospital, Catholic
University of Louvain, Belgium, Brussels.
Patients
All children aged 29days up to 18years hospitalized dur-
ing the study period with a diagnosis of acute meningitis
were retrospectively included. The analysis was based on
computerized medical charts and on CSF data base from the
laboratory. Meningitis was defined by CSF pleocytosis
>8WBC/mm. Patients were excluded if they had any of the
following criteria : purpura, clinical sepsis, predisposing fac-
tors (congenital or acquired immunodeficiency, history of
neurosurgery), Lyme meningitis, antibiotic treatment admin-
istered at least 72 h before lumbar puncture, and traumatic
lumbar puncture with >10,000 RBC/mm.
The diagnosis of BM was retained in patients with positive
blood and/or CSF cultures or in case of a positive CSF latex
antigen or nucleic acid amplification test for bacteria known to
cause meningitis. Aseptic meningitis (AM) was defined as the
occurrence of an episode of acute meningitis with negative
bacterial culture from blood and/or CSF and when available,
the presence of virus detected by CSF culture and/or PCR.
Data analysis
The BMS score was applied to each patient fulfilling the
criteria of acute meningitis with CSF pleocytosis to calculate
its specificity and sensitivity with 95% CI. The BMS proposed
by Nigrovic et al (11) included a list of 5 items (table 1): sei-
zure before or/at presentation, blood neutrophil count
10,000/mm, positive CSF gram-stain, CSF neutrophil count
1,000/mm and CSF protein concentration 80mg/dl. All
Table 1: Bacterial Meningitis Score (ranges from 0 to 6)
Predictor Points
Positive gram stain 2
CSF protein 80mg/dl 1
Peripheral absolute neutrophil count 10,000/mm 1
CSF absolute neutrophil count 1000/mm 1
Seizure at or before admission 1 ANC, absolute neutrophils count; CSF, cerebrospinal fluid.
The Bacterial Meningitis Score 283
items are quoted by 1 excepted CSF gram-stain which is
quoted 2; this makes the possible score from 0 to 6. Children
with missing data were excluded.
RESULTS
The computerized database system identified 242patients
who had a diagnosis of meningitis. Sixty- height patients
were excluded for the following reasons: Lyme meningitis
(n=26), antibiotic before lumbar puncture (n=11), the pres-
ence of an underlying pathology or predisposing factor
(n=5), purpura (n=5), traumatic lumbar puncture (n=10),
missing biological data (n=11).
Hence, 174 patients were eligible for evaluation of the
BMS decision rule; these patients were distributed as 148
(85%) with AM and 26(15%) with BM. The distribution of men-
ingitis according to the location was as follow: Yvoir 130
patients of whom 112 (86%) with AM, 18 (14%) with BM and
Brussels 44 patients of whom 36 (82%) with AM and 8 (19%)
with BM. The bacterial pathogens were identified by positive
blood and/or CSF culture in 23/26 patients with BM of whom
13 positive blood culture and 20 positive CSF culture (Neisseria
meningitidis (n=13), Streptoccocus pneumoniae (n=8), Esche-
richia coli (n=1), Streptococcus agalactiae (group B streptococ-
cus) (n=1)). The diagnosis of BM was further based on the
results of the CSF PCR testing in 2 patients (one N. meningitidis
and one S pneumoniae) and on CSF latex antigen agglutina-
tion test for N. meningitidis in 1 patient. Enterovirus infection
was diagnosed by positive CSF PCR assay in 16patients, posi-
tive CSF viral culture in 2 patients and cytomegalovirus infec-
tion by positive blood and CSF PCR in 1 patient.
The characteristics of the 174 patients are shown in Table2.
The BMS was calculated for each of these patients. Of the
93patients categorized as with very low score (BMS=0), 2 had
BM (negative predictive value 97.8%). Table 3 shows the distri-
bution of acute meningitis in relation to the BMS score: 6/147
(4%) patients with BMS 1 had BM compared to 20/27 (74%)
patients with BMS >1 (p<0.001). The sensitivity of the BMS (for
a score 1) for BM was 92.3% (95% CI: 82.1-100) and the spec-
ificity 61.5% (95%CI 53.6 -69.3.). The two patients with BM not
detected by the BMS were 2.5 and 15years old, and were both
infected with N. meningitidis, one with a petechial rash. The two
Table 2: Characteristics of the 174 patients with acute
meningitis included in the final analysis
Aseptic meningitis Bacterial meningitis
(n=148) (n=26)
Age, median (range) (years) 9.4 (0.15-15.7) 0.8 (0.08-16.2)
Seizure (n(%)) 5 (3.4%) 3 (11.5%)
Blood ANC 104/mm (n(%)) 50 (34%) 13(50%)
CSF ANC (x106/l)* 110 ( 242) 1313 (2037)
CSF ANC 1000/l (n (%)) 4(2.7%) 9 (35%)
CSF protein (mg/dl)* 38.2 ( 26) 213.73 ( 258)
CSF protein 80mg/dl (n(%)) 6 (4%) 18 (70%)
Positive gram-stain (n(%)) 1 18 (70%)
Petechial rash (n(%)) 12 (8%) 9 (35%)
* Means and SD.
Acta Clinica Belgica, 2012; 67-4
284 The Bacterial Meningitis Score
Table 3: Distribution of meningitis in relation to the
Bacterial Meningitis Score (BMS)
BMS score Number (%) patients Number patients with
(patient number) with bacterial meningitis aseptic meningitis
BMS=0 (n=93) 2 (2.1%) 91
BMS=1 (n=54) 4 (7.4%) 50
BMS=2 (n=6) 1 (16%) 5 rules showed 100% sensitivity, with a specificity higher for the
BMS=3 (n=10) 8 (80%) 2 BMS (52%) compared with the Meningitest (36%).
BMS 4 (n=11) 11 (100%) 0 While awaiting results of large prospective multicenter
patients with a diagnosis of N. meningitidis meningitis based on
positive CSF latex agglutination or positive CSF PCR were
respectively 9years old (BMS=5) and 4.5years old (BMS=1 formed in priority in a child with acute meningitis. This rule
plus petechial rash). The patient with S pneumoniae meningitis
based on CSF PCR was 8 months old and had a BMS=4.
DISCUSSION
The present study constitutes the sixth external evaluation
conducted to assess the clinical value of the BMS in children
presenting with a clinical picture of acute meningitis (7, 8, 14,
15, 16). This score was initially designed to help the physician
in the differential diagnosis of the less problematic AM and the
potentially devastating/dreadful BM. The BMS was first devel-
oped in the USA before the pneumococcal conjugate vaccine
era in the setting of a retrospective cohort study which almost
included seven hundred hospitalized children aged 1 month
or more with CSF pleocytosis (11). The score was tested on a
validation set including 234 patients of whom 38 with BM with
a sensitivity and a negative predictive value of 100%. Subse-
quently, the BMS was subjected to external evaluations both in
Europe and in the USA and all these studies confirmed its high
sensitivity (98.3% to 100%) and negative predictive value
(99.9% to 100%) (7, 8, 14, 15, 16). In comparison to these previ-
ous reports, , we found a lower sensitivity (92.3%) of the BMS
and the two patients not correctly identified by the BMS were
both infected with N. meningitidis, including one with a pete-
chial rash. In the external evaluation from the French Bacterial
Meningitis Surveillance Network which included 889 children
with BM, the diagnosis would have been missed in five patients
by assigning the BMS (14). Four of these patients had devel-
oped a meningitis caused by N.meningitidis none had purpura
but two had petechial rash.
If the main aim of a prediction tool such as the BMS is to
assist the clinician in deciding which children with CSF pleo-
cytosis should be hospitalized and treated with antibiotic
while awaiting for the blood and CSF culture results, then a
sensitivity of 100% should be considered in the rule owing to
the potentially devastating consequences that a delayed
diagnosis or a misdiagnosis of BM may have. Several authors
have proposed to refine the BMS by another score called
Meningitest in which only certain items of the BMS were
retained (positive CSF gram-stain, CSF protein concentration
with a lower threshold of 50mg/dl, seizures), while other ele-
ments were excluded (absolute neutrophil count in blood
and CSF) and in which one additional biological parameter
(procalcitonin serum level 0.5ng/ml) and two other clinical
criteria (presence of purpura and toxic appearance) were
Acta Clinica Belgica, 2012; 67-4
included (17). The Meningitest was first evaluated in a single-
center retrospective cohort and it was found to yield a sensi-
tivity of 100% (95% CI: 96-100) and a specificity of 62% and
51% in the derivation and internal validation sets respectively
(17). The Meningitest was further retrospectively compared
with the BMS by a secondary analysis of a retrospective mul-
ticentre cohort study in five European countries (15). Both
studies evaluating the usefulness of these predictive score in
the daily clinical practice, we suggest to use the BMS as a
diagnostic tool assisting both the physician and the labora-
tory to better select which laboratory tests should be per-
only concerns patients who fulfilled the same inclusion and
exclusion criteria as those described in our study and other
derivations sets including no antibiotic prior lumbar puncture
and patient age 29days (11).
In children presenting with acute meningitis and none of
the above mentioned exclusion criteria, we propose a deci-
sion tree based on the BMS score (fig 1). Since over than 96%
of patients with a BMS 1 will not have BM and as prevalence
of enteroviral meningitis in children with CSF pleocytosis
range from 66% to 90%, enteroviral PCR testing would be rec-
ommended in all patients with BMS score 1. Several studies
have demonstrated a significant impact of CSF enteroviral
PCR testing on length of hospital stay and duration of antibi-
otic use (18, 19, 20). In a recent prospective study, a positive
CSF enterovirus PCR testing allowed discharge within 24h in
95% of children evaluated for meningitis and hence before
bacterial culture information was available (18). Some authors
consider that the clinical usefulness of PCR could be enhanced
by performing the test only during the peak enterovirus sea-
son (mid summer to early fall) (19). We propose to perform
enteroviral PCR testing according to the BMS score and to
give intravenous antibiotic until results of PCR are available.
This approach will allow to discharge a high proportion of
well appearing children with meningitis before results of bac-
terial culture are available if a diagnosis of viral meningitis is
confirmed by CSF enterovirus PCR testing. On the other hand,
a diagnosis of BM should be strongly considered in patients
with a BMS 2 as it is demonstrated that 60 to 87% of these
patients will have BM (11, 16). The main dilemma for the clini-
cian will concern a limited number of patients with a BMS 2
but with negative bacterial cultures from blood and CSF. For
these patients, the diagnosis will be essentially relying on the
results of the Gram-stain as well as on PCR testing. It should
also be underlined that in the various studies, including ours
which evaluated the BMS, the sensitivity of the CSF gram-
stain ranged from 61 to 89% and the specificity from 99.4%
to 100% (8, 11, 14, 16). Due to the high specificity of the Gram
stain when associated to pleocytosis, these patients should
received a full antibiotic course (21). Whenever available, bac-
terial PCR testing from CSF and serum samples should be
performed to detect the causative organism.
In conclusion , we believe that the BMS will need further
refinement before it can be considered as a diagnostic tool
enabling the physician to discharge a child with meningitis if
BMS=0 pending for the availability of bacterial culture results
(i.e. duration of 48-72 h). We would rather suggest using the

Figure 1: Decision tree based on Bacterial Meningitis Score (BMS)
BMS as a diagnostic tool for assisting the physician to better
manage a child with acute meningitis.
REFERENCES
1. Kim KS. Acute bacterial meningitis in infants and children. Lancet Infect Dis
2010; 10: 32-42.
2. Tunkel AR, Hartman BJ, Kaplan SL, et al. Practice guidelines for the management
of bacterial meningitis. Clin Infect Dis 2004; 39: 1267-1284.
3. Chang CJ, Chang WN, Huang LT, et al. Bacterial meningitis in infants: the epide-
miology, clinical features, and prognostic factors. Brain Dev 2004; 26: 168-175.
4. de Louvois J, Halket S, Harvey D. Effect of meningitis in infancy on school leav-
ing examination results. Arch Dis Child 2007; 92: 959-962.
5. Mace SE. Acute bacterial meningitis. Emerg Med Clin North Am 2008; 26: 281-
317.
6. Roine I, Peltola H, Fernndez J, et al. Influence of admission findings on death
and neurological outcome from childhood bacterial meningitis. Clin Infect Dis
2008; 46: 1248-1252.
7. Dubos F, Lamotte B, Bibi-Triki F, et al. Clinical decision rules to distinguish
between bacterial and aseptic meningitis. Arch Dis Child 2006; 91: 647-650.
8. Nigrovic LE, Kuppermann N, Macias CG, et al. Clinical prediction rule for identi-
fying children with cerebrospinal fluid pleocytosis at very low risk of bacterial
meningitis. JAMA 2007; 297: 52-60.
9. Haruda FD. Meningitis viral versus bacterial. Pediatrics 2003; 112: 447-448.
10. Dubos F, Martinot A, Gendrel D, Brart G, Chalumeau M. Clinical decision rules
for evaluating meningitis in children. Curr Opin Neurol 2009; 22: 288-293.
The Bacterial Meningitis Score 285
11. Nigrovic LE, Kuppermann N, Malley R. Development and validation of a multi-
variable predictive model to distinguish bacterial from aseptic meningitis in
children in the post-Haemophilus influenzae era. Pediatrics 2002; 110: 712-719.
12. McGinn TG, Guyatt GH, Wyer PC, Naylor CD, Stiell IG, Richardson WS. Users
guides to the medical literature: XXII: How to use articles about clinical decision
rules. Evidence- Based Medicine Working Group. JAMA 2000; 284: 79-84.
13. Stiell IG, Wells GA. Methodologic standards for the development of clinical deci-
sion rules in emergency medicine. Ann Emerg Med 1999; 33: 437-447.
14. Dubos F, De la Rocque F, Levy C, et al. Sensitivity of the bacterial meningitis
score in 889 children with bacterial meningitis. J Pediatr 2008; 152: 378-382.
15. Dubos F, Korczowski B, Aygun DA, et al. Distinguishing between bacterial and
aseptic meningitis in children: European comparison of two clinical decision
rules. Arch Dis Child 2010; 95: 963-967.
16. Pierart J, Lepage P. Value of the Bacterial Meningitis Score (BMS) for the dif-
ferential diagnosis of bacterial versus viral meningitis. Rev Med Liege 2006; 61:
581-585.
17. Dubos F, Moulin F, Raymond J, Gendrel D, Brart G, Chalumeau M. Distinction
between bacterial and aseptic meningitis in children: refinement of a clinical
decision rule. Arch Pediatr 2007; 14: 434-438.
18. Archimbaud C, Chambon M, Bailly JL et al. Impact of rapid enterovirus molecu-
lar diagnosis on the management of infants, children, and adults with aseptic
meningitis. J Med Virol 2009; 81: 42-48.
19. Hamilton MS, Jackson MA, Abel D. Clinical utility of polymerase chain reaction
testing for enteroviral meningitis. Pediatr Infect Dis J 1999; 18: 533-537.
20. Nigrovic LE, Chiang VW. Cost analysis of enteroviral polymerase chain reaction
in infants with fever and CSF pleocytosis. Arch pediatr Adolesc Med 2000; 154:
817-821.
21. Neuman MI, Tolford S, Harper MB. Test characteristics and interpretation of cere-
brospinal fluid gram stain in children. Pediatr Infect Dis J 2008; 27: 309-313.
Acta Clinica Belgica, 2012; 67-4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.