Neurocrit Care (2020) 33:776–784

https://doi.org/10.1007/s12028-020-00964-w

ORIGINAL WORK

Utility of the Cell Index in Predicting External

Ventricular Drain‑Related Ventriculo‑Meningitis
Sarah Liew1*, Stephen Richards1, Kwok Ming Ho2,3,4 and Ronan Murray5,6,7

© 2020 Springer Science+Business Media, LLC, part of Springer Nature and Neurocritical Care Society

Abstract
Background/Objective: Ventriculo-meningitis (VM) is an important complication of external ventricular drains
(EVDs) in neurosurgical patients. Consequences include increased morbidity, mortality, and duration of hospital stay.
Early diagnosis of EVD-associated VM allows earlier treatment intervention. The cell index (CI) may provide a simple
measure that overcomes the limitations of isolated cerebrospinal fluid (CSF) parameters and other diagnostic tests,
allowing earlier prediction of VM.
Methods: All patients admitted to a tertiary hospital and requiring EVD insertion during 2015 and 2016 were
assessed for inclusion in this retrospective case–control study. Patients with a known or suspected intracranial infec-
tion were excluded. Of the 186 patients who underwent EVD insertion, 95 patients were included in the final cohort.
Data pertaining to patient characteristics and laboratory indices were extracted from health records and the micro-
biology laboratory database. The CI was calculated as the ratio of temporally related CSF leukocytes/erythrocytes to
peripheral blood leukocytes/erythrocytes. Data from patients with microbiologically confirmed VM were analyzed in
comparison with those not developing VM during the course of their stay. Categorical and continuous variables with
skewed distributions were analyzed by Chi square and Mann–Whitney tests, respectively.
Results: EVD-associated VM developed in 7.4% of patients. The highest CSF CI (within 3 days prior to diagnosis of VM
or at any time for those not developing VM) differed significantly between the two groups (16; IQR 10.8–48.5 vs. 3.3;
IQR 1.0–12.8, respectively; p = .046). The area under the receiver operating characteristic curve (AUROC) for the high-
est CI was 0.727 (95% confidence interval [CI] 0.526–0.929; p = .027). A CI of 10.4 provided a sensitivity and specificity
of 80.5% and 70.5%, respectively, for the early diagnosis of VM.
Conclusions: In neurosurgical patients with an EVD, the CSF CI significantly predicted the development of VM.
Keywords: Cell index, Cerebral ventriculitis, Ventriculostomy, Neurosurgery, Cerebrospinal fluid

Introduction have both diagnostic and therapeutic utility, allowing

Ventriculo-meningitis (VM) is a serious complication of
external ventricular drains (EVDs), with consequences
including both increased mortality and morbidity [1–5]
EVDs, otherwise known as ventriculostomy catheters,
are commonly used in neurocritical care [4, 6, 7]. They
*Correspondence: Sarah.liew@health.wa.gov.au
1
Department of Intensive Care Medicine, Sir Charles Gairdner Hospital,
Perth, WA, Australia
Full list of author information is available at the end of the article
Research conducted at Sir Charles Gairdner Hospital, Nedlands, WA,
Australia.
measurement and modulation of intracranial pressure
(ICP) [7, 8]. The incidence of EVD-associated VM is
reported to range between 2 and 27%, with some of this
wide variation attributable to differing diagnostic criteria
[6–10]. Risk factors and preventative strategies have been
clearly documented [1, 3, 10, 11].
Early diagnosis of EVD-associated VM may facili-
tate timely institution of appropriate therapy, thereby
mitigating the adverse consequences of infection [3,
12]. Diagnosis depends largely upon laboratory meas-
urements in a setting where clinical assessment may

be confounded by the primary neurocritical condition
and consequences thereof [4, 8, 11, 13–16]. Positive
microbiological culture of cerebrospinal fluid (CSF)
together with CSF pleocytosis, appropriate clinical con-
text and exclusion of contamination has been widely
used to define EVD-associated VM and initiate appro-
priate treatment [10, 11, 17, 18]. This dependency upon
microbiological culture result has the disadvantage of
delayed diagnosis and hence treatment and may con-
tribute to worse outcomes [4, 12, 14].
Various surrogate parameters have been examined
and used in order to expedite the diagnosis of EVD-
associated VM [7, 14]. Examination of traditional CSF
components has been studied, with varying results [7,
14–16]. Importantly, interpretation of CSF pleocytosis
in this population may not be used in a manner similar
to traditional interpretation of CSF results in patients
suspected of meningitis without an indwelling device
[7, 11]. The absolute number of CSF erythrocytes and
leukocytes may be influenced by a number of processes
other than infection [4, 8, 11, 16]. These include pri-
mary or device-related intraventricular hemorrhage,
intraventricular leukocyte infiltration as a consequence
of foreign device presence, and physiological clearance
of intraventricular blood [6, 11, 19]. A variety of new
direct and indirect laboratory tests have been proposed
for the purpose of early diagnosis; however, these may
have limited utility due to varying discriminatory abil-
ity, availability, and expense [7, 15, 20–22].
Cognizant of variations in CSF cell counts in the con-
text of EVDs, the cell index (CI) has been proposed as a
simple means to predict early development of VM [19].
The CI is the ratio of CSF leukocyte/erythrocyte count
to the leukocyte/erythrocyte ratio in peripheral blood
[8, 19]. The formula theoretically corrects for aber-
rations in cell counts, thereby retaining the ability of
these simple parameters to determine the early devel-
opment of infection.
A significant rise in CI was shown to predict develop-
ment of VM up to 3 days prior to positive CSF culture
[19]. A fivefold increase in the CI has been suggested
to be highly indicative of EVD-related infection, even
being incorporated into various local guidelines [6, 8,
9].
The advantages of the CI in predicting the early diag-
nosis of ventriculitis, namely its availability, speed of
processing and cost, are attractive. The existing data
describing the application of the CI are limited [8, 11,
14]. We hypothesized that changes in the white cell to
erythrocyte count ratio in the CSF relative to the same
in the blood could precede the presence of positive CSF
bacterial culture. This study was undertaken in order to
777
further examine the utility of the CI in predicting EVD-
related VM in a larger neurocritical care population.
Methods
Study Design
A retrospective case–control study was conducted
of patients admitted to Sir Charles Gairdner Hospi-
tal (SCGH) between 2016 and 2017. All neurosurgical
patients over the age of 18 who underwent insertion of an
EVD for any indication during their admission, and had
two or more samples of CSF sent for laboratory analysis
were included in the study. Patients were excluded if they
had an admission diagnosis of known or suspected pri-
mary intracranial infection. Only the first presentation
(index case) was analyzed if the patient had two or more
hospital admissions with EVD insertion.
CSF Sampling
SCGH is a 600-bed tertiary hospital located in Western
Australia. It is the state neurosurgical center, providing
emergency neurovascular intervention services predomi-
nantly for the treatment of primary and secondary neuro-
logical malignancies and infections, stroke, and cerebral
aneurysm rupture. Antibiotic-impregnated EVDs are
inserted in theatre via a sterile, standardized technique,
by an experienced member of the neurosurgical team.
In our facility, patients may be admitted to the intensive
care unit (ICU) or the neurosurgical high dependency
unit (HDU) following neurosurgery and EVD insertion,
depending on stability and requirements for additional
organ support. In our neurosurgical HDU, CSF sampling
is not routinely performed in all patients. In patients at
increased risk of intracranial infection or where the EVD
remains in situ for a prolonged period, CSF sampling is
generally analyzed every 3 to 4 days. In our ICU popu-
lation, neurosurgical patients are often under chemical
sedation or have a reduced conscious state as a result of
their underlying neurological condition. CSF sampling is
necessary to delineate intracranial infection from other
common causes for fever and reduced conscious state.
In our ICU, CSF sampling is conducted every 3 days or
more frequently, depending on the index of suspicion for
VM.
Data Collection
The following demographic and clinical parameters
were obtained from medical records: primary diagnosis,
age, gender, World Federation of Neurosurgical Soci-
eties (WFNS) grading of subarachnoid hemorrhage,
immunosuppression, length of hospital and ICU stay,
duration of mechanical ventilation, other neurosurgical
intervention(s) performed within 24 h of EVD insertion,
date and total number of EVD insertions, and disposition
778
following discharge from hospital. Laboratory data were
obtained from the electronic hospital database com-
prised of: CSF gram stain, CSF culture result, CSF eryth-
rocyte and leukocyte count, and positive culture results
from other sites. Only CSF samples obtained from an
EVD were included in analysis; peripheral blood leuko-
cyte and erythrocyte count obtained within 24 h of the
corresponding CSF sample were analyzed concordantly.
The presence of intraventricular blood was examined
according to computed tomography (CT).
Definitions
VM was defined as having an organism(s) isolated
from CSF by culture. This is consistent with the Cent-
ers for Disease Control and Prevention (CDC) criteria
for VM [23], within the limitations of the clinical con-
dition imposed by neurocritical care patients, of which
our cohort had a majority (78.9%). Where CSF culture
was positive for a coagulase-negative staphylococcus or
another typical skin commensal, two positive CSF cul-
tures were required. Contamination was defined as hav-
ing one isolated positive CSF culture for a typical skin
commensal or a positive CSF culture within 24 h of EVD
insertion.
Cell index was calculated using the following formula:
Leukocytes (CSF)/Erythrocytes (CSF)
Cell index (CI) =
Leukocytes (blood)/Erythrocytes (blood)
In the VM group, the CI was calculated for each CSF
sample until the time of positive CSF culture (defined as
day 0). The CI of five preceding CSF samples was com-
pared to the baseline CI (derived from the first CSF sam-
ple following EVD insertion) to obtain the CI escalation
ratio. In the non-ventriculitis group, the cell index was
calculated for each CSF sample obtained during admis-
sion. The CSF sample with the highest CI during admis-
sion was defined as day 0, and the cell index escalation
ratios for the five preceding CSF samples were calculated.
If more than one CSF sample or peripheral blood count
was sent in a 24-h period, only the first CSF sample or
blood count was used in the analysis.
Statistical Analyses
Categorical and continuous variables with skewed distri-
butions were analyzed by Chi square and Mann–Whit-
ney tests, respectively. The ability of the highest CSF CI
and CSF CI escalation ratio (= highest/baseline CSF cell
index) to discriminate between patients with and with-
out culturally proven VM was reported by area under
the receiver operating characteristic curve (AUROC).
The association between the primary diagnosis leading
to the requirement of an EVD and development of VM
was assessed by Cox proportional hazards regression
analysis. All analyses were conducted by the MedCalc
Statistical Software (version 19.0.4, MedCalc Software
bvba, Ostend, Belgium; https​://www.medca​lc.org; 2019).
All statistical tests were two-tailed without adjusting
for multiplicity, and an alpha error < 5% was taken as
significant.
Results
General Characteristics
One hundred and eighty-six patients underwent EVD
insertion over a 2-year period and were assessed for
inclusion into this study. A final study cohort of 95
patients was included after application of exclusion cri-
teria (Fig. 1). The median age of the study cohort was
58 (IQR 47–67) years, and 58% were female (Table 1).
Seventy-nine (83%) of patients receiving an EVD were
admitted to the intensive care unit (ICU). VM developed
in 7 (7.4%) of patients, with a gram-negative bacillus iso-
lated in 4 (57.1%) patients and coagulase-negative staphy-
lococci (CoNS) in the remainder (42.9%).
Subarachnoid hemorrhage constituted the primary
neurosurgical condition (59 [62.1%]), with intraparen-
chymal hemorrhage (18 [18.9%]) and central nervous
system neoplasm (5 [5.3%]) the second and third most
frequent diagnoses. Intraventricular blood was present
in 66 (69.4%) of patients on CT scan at the time of EVD
insertion. An open neurosurgical procedure, including
decompressive craniectomy, or craniotomy (for debulk-
ing/resection/biopsy of lesion, evacuation of hematoma,
or clipping of aneurysm) was undertaken in 16 (16.8%)
of patients. The mean duration of EVD insertion was 12
(IQR 8–15) days. The median length of hospital stay was
26 (IQR 18–41) days. Seventy-eight (82.1%) of patients
survived to hospital discharge, with the majority of these
patients (52[66.6%]) discharged to a rehabilitation facility,
while 18 (23%) returned home.
Cell Index Parameters
The baseline CI was similar between those patients
developing VM (0.23; IQR 0.07–3.2) and those not devel-
oping VM (0.37 IQR 0.11–2.1), p = .598. The highest
CSF CI (within 3 days prior to diagnosis of VM or at any
time for those not developing VM) differed significantly
between the two groups (16; IQR 10.8–48.5 vs. 3.3; IQR
1.0–12.8, respectively; p = .046). The CSF CI escalation
ratio did not differ significantly between those developing
and those not developing VM (39.2; IQR 1–467 vs. 4.7;
IQR 1.3–23.1, respectively; p = .305). The results pertain-
ing to the highest CSF CI and CI escalation ratio for both
groups are depicted in Fig. 2.

Fig. 1 Flow diagram depicting study cohort derivation
The AUROC for the highest CI parameter as previ-
ously described was 0.727 (95% confidence interval [CI]
0.526–0.929; p = 0.027). The best cut point for the highest
CI was 10.4 with a corresponding sensitivity and specific-
ity of 80% and 70.5%, respectively (Fig. 3).
Discussion
This study demonstrates that the CI was able to success-
fully predict the subsequent development of VM in a
population of neurosurgical patients with an EVD. A CI
above 10.4 provided predictive ability to determine VM
development with a good sensitivity and moderate speci-
ficity. These findings are clinically significant and relevant
for a diverse population of neurosurgical patients with an
EVD and thus require further discussion.
First, this study is the largest reported observational
study to our knowledge specifically comparing the CI in
a diverse population of neurosurgical patients with an
EVD. Pfausler et al. [19] initially proposed the CI and
investigated its discriminatory ability in a cohort of 13
patients with intraventricular hemorrhage. The CI calcu-
lated in CSF samples up to 3 days prior to definitive diag-
nosis correlated with the development of ventriculitis.
Wiegand et al. [13] recently investigated the diagnostic
utility of the CI among other parameters, within a cohort
of 39 patients with confirmed EVD-associated infections.
779
A comparison of CI from the time of EVD insertion to
time of EVD microbiological diagnosed infection was
undertaken with no significant correlation demonstrated
[13]. Schade et al. [7] mentioned a lack of diagnostic or
predictive value of the CI in a prospective study, although
do not provide related data. Our study investigated the
utility of the CI across a large group of patients with an
EVD in situ, both with and without radiologic evidence
of intraventricular hemorrhage, thereby increasing the
applicability of our results.
Recently, Montes et al. evaluated the CI in a study
attempting to develop a diagnostic model for the develop-
ment of healthcare-associated ventriculitis and meningi-
tis [14]. Of the 111 patients with intracranial hemorrhage
studied, they found that the CI was the best individual
variable predicting development of VM [14]. A signifi-
cant difference in CI was found between those patients
developing VM (4.3), compared with those patients who
did not (0.007) [14]. Of the variables assessed by Montes
et al., the CI was found to be the best predictor of subse-
quent VM development [14].
Our study provides data on the utility of the CI across
a diverse range of neurocritical care patients with an
EVD. This recognizes the corrective nature of the CI with
respect to CSF cellular variations due to macroscopic or
microscopic erythrocyte contamination, and leukocyte
780

Table 1 Differences in characteristics between patients with and without ventriculo-meningitis (N = 95)
Variable All patients Ventriculo-meningitis No ventriculo-men- p ­valuea
N = 95 n = 7 (7.4%) ingitis
n = 88 (92.6%)

Age, yrs (IQR) 58 (46–66) 58 (53–59) 57 (46–66) 0.938

Female, no. (%) 55 (58) 4 (57) 51 (58) 0.999
Immunosuppressedb, no. (%) 4 (4) 0 (0) 4 (5) 0.999
Primary diagnosis, no. (%) 0.111
Subarachnoid hemorrhage 59 (62) 4 (57) 55 (63)
Intraparenchymal hemorrhage 18 (19) 1 (14) 17 (19)
Brain tumor with obstructive hydrocephalus 5 (5) 2 (29) 3 (3)
Cerebral infarction 4 (4) 0 (0) 4 (5)
Obstructive hydrocephalus 6 (6) 0 (0) 6 (7)
Others 3 (3) 0 (0) 3 (3)
WFNS grading no. (%) 0.616
Not applicable 36 (38) 3 (43) 33 (38)
I (GCS 15, no motor deficit) 13 (14) 0 (0) 13 (15)
II (GCS 13–14, no motor deficit) 18 (19) 2 (29) 16 (18)
III (GCS 13–14, with motor deficit) 1 (1) 0 (0) 1 (1)
IV (GCS 7–12) 13 (14) 0 (0) 13 (15)
V (GCS 3–6) 14 (15) 2 (29) 12 (14)
Intraventricular blood, no. (%) 66 (70) 2 (29) 64 (73) 0.026
APACHE II score (IQR)c 18 (14–21) 21 (14–22) 18 (14–21) 0.762
Open neurosurgical ­proceduresd, no. (%) 16 (17) 2 (29) 14 (16) 0.335
CSF cell index at baseline on admission (IQR) 0.37 (0.11–2.13) 0.23 (0.07–3.2) 0.37 (0.11–2.1) 0.598
Highest CSF cell ­indexe (IQR) 3.7 (1.0–15.4) 16.0 (10.8–48.5) 3.3 (1.0–12.8) 0.046
CSF cell index escalation r­ atiof (IQR) 5.2 (1.3–23.7) 39.2 (1.0–467) 4.7 (1.3–23.1) 0.305
Time to ventriculitis or EVD removal, days (IQR) 11 (8–15) 9 (3–17) 11 (8–15) 0.445
Total duration of EVD in situ, days (IQR) 12 (8–15) 16 (12–20) 11 (8–15) 0.115
No. of subsequent EVDs needed (IQR) 0 (0–0) 1 (0–1) 0 (0–0) 0.010
Total no. of CSF samples sent for culture during the 4 (3–6) 7 (6–8) 4 (2–5) 0.004
entire hospital stay (IQR)
Ventilated days (IQR) 6 (0–11) 3 (0–12) 6 (0.3–11) 0.575
Length of ICU stay, days (IQR) 11 (4–14) 11 (0–20) 11 (4–14) 0.683
Length of hospital stay, days (IQR) 26 (18–42) 34 (23–48) 26 (17–40) 0.131
Discharge destination, no. (%) 0.238
Home 18 (19) 0 (0) 18 (21)
Other hospital 3 (3) 0 (0) 3 (3)
Rehabilitation facility 52 (55) 6 (86) 46 (52)
Deceased 17 (18) 0 (0) 17 (19)
Residential facility 5 (5) 1 (14) 4 (5)
APACHE, Acute Physiology and Chronic Health Evaluation; CSF, cerebrospinal fluid; EVD, external ventricular drain; GCS, Glasgow Coma Score; ICU, intensive care unit;
IQR, interquartile range; WFNS, World Federation of Neurosurgical Society
a
Chi square or Mann–Whitney test
b
Received chemotherapeutic agents within 30 days; currently taking systemic corticosteroids or antirejection drugs; neutropenia (< 1 × 109/l) or leukopenia
(< 4 × 109/l)
c
Including only 79 patients admitted to the intensive care unit (5 developed VM and 74 did not)
d
Including evacuation of intracranial hematoma, decompressive craniectomy, biopsy/resection/debulking of brain tumors, resection of arteriovenous malformation,
or clipping of cerebral aneurysm in addition to an EVD insertion. All continuous data are median values
e
Highest CSF cell index = within 3 days prior to the diagnosis of ventriculitis or during the entire admission for those without ventriculitis
f
CSF cell index escalation ratio = highest cell index within 3 days prior to the diagnosis of ventriculitis or during the entire admission for those without ventriculitis
and CSF cell index at baseline
 781

Fig. 2 a Difference in highest Cell Index within 3 days prior to the diagnosis of ventriculo-meningitis (VM) and those without VM during the entire
admission, and b difference in cerebrospinal fluid (CSF) Cell Index Escalation Ratio (= highest/baseline CSF Cell Index)

pathological and physiological infiltration. Our results Second, investigation of the CI in relation to VM
therefore confirm and extend the applicability and utility outcome allowed development of a discriminatory
of the CI with respect to the limited available data previ- test with which to predict EVD-associated infection.
ously described. Although the AUC was moderate (0.727), a CI of 10.4
782
Fig. 3 Area under the receiver-operating-characteristic curve
(AUROC) for the highest cerebrospinal fluid (CSF) Cell Index within
3 days prior to the diagnosis of ventriculo-meningitis (VM) was
0.727 (95% confidence interval [CI] 0.526–0.929; p = 0.027)—the best
CSF Cell Index cut point for this analysis was 10.4 and the correspond-
ing sensitivity and specificity to diagnose ventriculitis were 80% and
70.5%, respectively. The AUROC for the CSF Cell Index Escalation Ratio
(= highest/baseline CSF Cell Index) was 0.617 (95% CI 0.335–0.899;
p = 0.417). Subgroup analyses: (1) Including only patients with
intraventricular blood improved the predictive ability of the CSF
highest Cell Index appeared to increase its predictive ability but this
was not statistically significant (AUROC = 0.891, 95% CI 0.814–0.967;
p = 0.061). Conversely, excluding patients with intraventricular
blood made the predictive ability of the highest Cell Index worse
(AUROC = 0.500, 95% CI 0.173–0.827; p = 0.999). (2) Area under the
AUROC for the CSF Cell Index Escalation Ratio (i.e. highest/baseline
CSF Cell index) = 0.617 (95% CI 0.335–0.899; p = 0.417). Including only
patients with intraventricular blood improved the predictive ability
of the CSF Cell Index Escalation (AUROC = 0.922, 95% CI 0.801–0.999;
p = 0.043), but excluding patients with intraventricular blood did not
improve the predictive ability of the CSF Cell Index Escalation Ratio
(AUROC = 0.533, 95% CI 0.260–0.807; p = 0.817)
suggests development of microbiologically confirmed
VM within the subsequent 3 days, with a sensitivity
of 80% and specificity of 70.5%. Montes et al. recently
evaluated the diagnostic accuracy of the CI and found
the CI to have good diagnostic accuracy, with an
AUROC of 0.825 [14]. Although both results were rea-
sonably comparable with regard to diagnostic utility,
the difference may reflect a more diverse neurosurgical
population (patients with an EVD both with and with-
out intraventricular hemorrhage were included in our
study) investigated and differences in CSF sampling
frequency. Both studies highlight the significant signal
that the CI may contribute. Furthermore, our results
emphasize the potential importance and utility of the
CI evaluation and calculation when analyzing EVD-
obtained CSF samples. By calculating and analyzing the
CI, EVD-associated VM may be anticipated, resulting
in further rigorous investigation and possible changes
in management.
Third, our study justifies the continued regular inves-
tigation and reporting of the constituent simple param-
eters used in the CI calculation. A number of factors
with potential to diagnose EVD-associated VM have
been investigated and advocated [7]. These include CSF
Interleukin-6 and 8, interleukin 1 beta, CD 62-L on neu-
trophils, soluble triggering factor on myeloid cells-1,
CSF procalcitonin, polymerase chain reactions (PCR)
for infectious agents, and CSF lactate. These have signifi-
cant disadvantages including varying predictive ability,
limited availability, and high cost [7, 8, 11, 12, 15]. Our
finding of a significant predictive ability of the CI to iden-
tify those patients likely to develop EVD-associated VM
underscores the continued utility of the measurement
and reporting of the CI parameters. In isolation, numer-
ous studies have shown the lack of correlation between
individual parameters (CSF cell count, glucose, and pro-
tein) and prediction of VM and are therefore not recom-
mended in this context [7, 8, 11, 15]. The CI overcomes
the shortcoming of these individual parameters and so
accounts for ventricular blood contamination and the
presence of leukocytes due to blood, foreign body inflam-
mation, and phagocytic migration [13, 19]. The finding
of significant utility of the CI justifies the financial and
logistic implications of continued testing and reporting
of these simple CSF parameters in neurosurgical patients
managed with an EVD. Our study results support the
value of the CI and the ongoing utility, evaluation, and
reporting of CSF parameters from samples in this regard.
Fourth, our study confirmed the continued incidence
of EVD-associated VM, despite implementation of risk-
reduction factors (such as antibiotic coated catheters).
The incidence of EVD-associated VM in our study (7.4%)
is similar to that found in other recent studies [7, 12, 15].
Although the incidence of VM in our study was relatively
low (historically reported incidence up to 27%) [13], the
continued significant presence of VM in this popula-
tion underscores the importance of continued attention
to preventative measures and early identification. The
importance of and potential aid of the CI in this context
is further suggested.
Finally, we would like to acknowledge the limitations
of our study. The retrospective study design may have
resulted in the presence of unrecognized confound-
ers, although care was taken to describe likely relevant
baseline characteristics. The limited number of patients
developing VM may have contributed to inequivalent

baseline characteristics in some areas; number of subse-
quent EVDs and incidence of immunosuppression rep-
resent two possible instances. Variation in CSF sampling
intervals—indicative of a break from standardized CSF
collection procedure—did not allow for more detailed
day-to-day CSF analysis. Variation in EVD-associated
VM diagnosis among studies may limit comparison.
This underscores the difficulty of diagnosis, particularly
within a critical care setting. Consistent with other stud-
ies and reviews, the definition used within this study
emphasizes the objectivity of the microbiological culture
and the limitations of clinical and hyperthermia related
features [10, 17, 18]. The Infectious Disease Society of
America (IDSA) recent guidelines acknowledge the rela-
tive lack of discriminatory ability available from clinical
signs and individual CSF cell parameters and emphasize
the importance of CSF culture in the diagnosis of VM
[11]. Differentiating cultures representing contamination
may be a further limitation. Care was, however, taken to
identify contaminant samples by applying organism and
EVD-sample collection criteria. Our retrospective study
was also limited in its ability to establish the effect of sys-
temic antibiotics on the cell index and cell counts. The
effect and use of systemic antimicrobial for other indi-
cations (e.g., pneumonia) were not recorded. Given the
specific route and dose of antimicrobials required to treat
VM, it is questionable whether antimicrobials for other
indications would influence the CI. Notably, no patients
in our study received intrathecal antimicrobials, which,
we suspect, would have a greater influence on CSF cell
counts than systemic antibiotics for extracranial infec-
tion. This is certainly an area that could benefit from fur-
ther study, but likely requires a prospective study design
to investigate this thoroughly.
Conclusions
In neurosurgical patients with an EVD, the CSF CI sig-
nificantly predicts development of VM. An absolute CI
value of 10.4 demonstrated utility as a simple predictive
tool to aid in the early diagnosis of VM in this popula-
tion. Further prospective studies are required to confirm
and refine these findings.
Author details
1
Department of Intensive Care Medicine, Sir Charles Gairdner Hospital, Perth,
WA, Australia. 2 Department of Intensive Care Medicine, Royal Perth Hospital,
Perth, WA, Australia. 3 School of Veterinary and Life Sciences, Murdoch Univer-
sity, Perth, WA, Australia. 4 Faculty of Medicine and Health Science, University
of Western Australia, Perth, WA, Australia. 5 Department of Infectious Diseases,
Sir Charles Gairdner Hospital, Perth, WA, Australia. 6 Department of Microbiol-
ogy, PathWest Laboratory Medicine, Queen Elizabeth II Medical Centre, Perth,
WA, Australia. 7 School of Pathology and Laboratory Medicine, University
of Western Australia, Perth, WA, Australia.
783
Author Contributions
All authors contributed to drafting and critical revision of the manuscript. SL
contributed to the study design and data collection. SR contributed to the
study conception and design, data interpretation, and analysis. KMH con-
ducted statistical analysis and interpretation of the data. RM provided clinical
insights and contributed to data interpretation.
Source of Support
Intensive Care Unit, Sir Charles Gairdner Hospital.
Conflict of interest
The authors have no conflicts of interests to declare.
Ethical Approval
No ethics approval was deemed necessary for this study as assessed by the
National Statement on Ethical Conduct in Human Research (5.1.22). This study
has been exempt from review by the Sir Charles Gairdner Hospital Human
Research and Ethics Committee, as assessed by the Sir Charles Gairdner Hospi-
tal Manager for Quality Improvement.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in pub-
lished maps and institutional affiliations.
Published online: 21 April 2020
References
1. Williamson RA, Phillips-Bute BG, McDonagh DL, et al. Predictors of
extraventricular drain-associated bacterial ventriculitis. J Crit Care.
2014;29:77–82.
2. Rahman M, Whiting JH, Fauerbach LL, et al. Reducing ventriculostomy-
related infections to near zero: the eliminating ventriculostomy infection
study. Jt Comm J Qual Patient Saf. 2012;38(10):459–64.
3. Humphreys H, Jenks PJ. Surveillance and management of ventriculitis
following neurosurgery. J Hosp Infect. 2015;89:281–6.
4. Pfisterer W, Muhlbauer M, Czech T, et al. Early diagnosis of external
ventricular drainage infection: results of a prospective study. J Neurol
Neurosurg Psychiatry. 2003;74:929–32.
5. Murthy SB, Moradiya Y, Shah J, et al. Incidence, predictors and outcomes
of ventriculostomy associated infections in spontaneous intracerebral
hemorrhage. Neurocrit Care. 2016;24:389–96.
6. Beer R, Lackner P, Pfausler B, et al. Nosocomial ventriculitis and meningitis
in neurocritical care patients. J Neurol. 2008;255:1617–24.
7. Schade RP, Schinkel J, Roelandse FWC, et al. Lack of value of routine
analysis of cerebrospinal fluid for prediction and diagnosis of external
drainage-related bacterial meningitis. J Neurosurg. 2006;104:101–8.
8. Martin RM, Zimmermann LL, Huynh M, Polage CR. Diagnostic approach
to health care and device-associated central nervous system infections. J
Clin Microbiol. 2018;56:e00861-18.
9. Beer R, Pfausler B, Schmutzhard E. Management of nosocomial
external ventricular drain-related ventriculomeningitis. Neurocrit Care.
2009;10:363–7.
10. Lozier AP, Sciacca RR, Romagnoli MF, et al. Ventriculostomy-related infec-
tions: a critical review of the literature. Neurosurgery. 2002;51:170–82.
11. Tunkell AR, Hasbun R, Bhimraj A, et al. Infectious diseases society of Amer-
ica’s Clinical practice guidelines for healthcare-associated ventriculitis and
meningitis. Clin Infect Dis. 2017;64(6):e34–365.
12. Gordon M, Ramirez P, Soriano A, et al. Diagnosing external ventricular
drain-related ventriculitis by means of local inflammatory response:
soluble triggering receptor expressed on myeloid cells-1. Crit Care.
2014;18:567.
13. Wiegand J, Hickson L, Merz TM. Indicators of external drainage-related
infections—a retrospective observational study. Acta Neurochir.
2016;158:595–601.
14. Montes K, Jenkinson H, Habib OB, et al. Corrected white blood cell count,
cell index, and validation of a clinical model for the diagnosis of health
784
care-associated ventriculitis and meningitis in adults with intracranial
hemorrhage. Clin Neurol Neurosurg. 2019;178:36–41.
15. Hoogmoed J, van de Beek D, Coert BA, et al. Clinical and laboratory
characteristics for the diagnosis of bacterial ventriculitis after aneurysmal
subarachnoid hemorrhage. Neurocrit Care. 2017;26:362–70.
16. Lenski M, Huge V, Schmutzer M, et al. Inflammatory markers in serum and
cerebrospinal fluid for early detection of external ventricular drain-associ-
ated ventriculitis in patients with subarachnoid hemorrhage. J Neurosurg
Anesthesiol. 2019;31:227–33.
17. Brown EM, de Louvois J, Bayston R, et al. The management of neurosurgi-
cal patients with postoperative bacterial or aseptic meningitis or external
ventricular drain-associated ventriculitis. Br J Neurosurg. 2000;14(1):7–12.
18. Mayhall CG, Archer NH, Lamb VA, et al. Ventriculostomy-related infections:
a prospective epidemiologic study. N Eng J Med. 1984;310:553–9.
19. Pfausler B, Beer R, Engelhardt K, et al. Cell Index- a new parameter for the
early diagnosis of ventriculostomy (external ventricular drainage)-related
ventriculitis is patients with intraventricular hemorrhage? Acta Neurochir.
2004;146:477–81.
20. Rath PM, Schoch B, Adamzik M, et al. Value of multiplex PCR using cer-
ebrospinal fluid for the diagnosis of Ventriculostomy-related meningitis in
neurosurgery patients. Infection. 2014;42:621–7.
21. Schoch B, Regel JP, Nierhaus A, et al. Predictive Value of Intrathecal
Interleukin-6 for ventriculostomy-related infection. Zentralbl Neurochir.
2008;69:80–6.
22. Boeer K, Vogelsang H, Deufel T, et al. Immediate diagnosis of ventriculitis:
evaluation of parameters independent of microbiological culture. Acta
Neurochir. 2011;153:1797–805.
23. CDC/NHSN Surveillance Definitions for Specific Types of Infections
January 2019. https​://www.cdc.gov/nhsn/PDFs/pscMa​nual/17psc​NosIn​
fDef_curre​nt.pdf. Accessed 10 July 2019.