Validation of the bacterial meningitis score in adults presenting to the ED with meningitis

a b s t r a c t

Objectives: The Bacterial meningitis Score classifies children with meningitis and none of the following high-risk predictors at Very low risk for bacterial meningitis: positive cerebrospinal fluid (CSF) Gram stain, CSF protein

>= 80 mg/dL, CSF absolute Neutrophil count (ANC) >= 1000 cells/mm³, peripheral ANC >= 10,000 cells/mm³, and seizure at or prior to presentation. Although extensively validated in children, the Bacterial Meningitis Score has not been rigorously evaluated in adults.

Methods: We performed a single-center cross-sectional retrospective study of adults presenting to the emergency department between 2003 and 2013 with meningitis (defined by CSF white blood cell count >=10 cells/mm³). We defined a case of bacterial meningitis with either a positive CSF or blood culture. We report the performance of the Bacterial Meningitis Score in the study population.

Results: We identified 441 eligible patients of which, 4 (1%) had bacterial meningitis. The Bacterial Meningitis Score had a sensitivity of 100% [95% confidence interval (CI) 40%-100%], specificity 51% (95% CI, 46%-56%) and negative predictive value of 100% (95% CI, 98%-100%). None of the low risk adults had bacterial meningitis. If Bacterial Meningitis Score had been applied prospectively, the Hospital admission rate would have dropped from 84% to 49% without missing any patients with bacterial meningitis.

Conclusions: The Bacterial Meningitis Score accurately identified patients at low risk for bacterial meningitis and could assist clinical decision-making for adults with meningitis.

Background

Although rare in the conjugate vaccine era [1], bacterial meningitis is a medical emergency with high morbidity and mortality [2]. While most adults with meningitis have viral rather than bacterial infections, clini- cians frequently initiate broad-spectrum antibiotics and hospitalize pa- tients while awaiting Bacterial culture results, which take several days to reliably exclude bacterial growth.

Accurate identification of low risk meningitis patients at the time of presentation could allow for selective outpatient management. The Bac- terial Meningitis Score is a validated clinical prediction rule to identify children at low risk for bacterial meningitis [3,4]. Although extensively

* Corresponding author at: St. Luke’s Hospital, Emergency Medicine, 101 Page Street, New Bedford, MA 02740.

E-mail address: [email protected] (R. McArthur).

evaluated in children, the performance of this prediction rule has been examined only in a small number of adults with meningitis [5]. To this end, we assembled a 10-year retrospective cohort of adult emergency department (ED) patients with meningitis and report the performance of the Bacterial Meningitis Score in the study population.

Methods

Study design

We conducted a single-center cross-sectional retrospective study of all patients aged 17 and older presenting to the ED of a single urban ac- ademic institution with meningitis. The study protocol was approved by the institutional review board of the study institution with a waiver of patient informed consent.

In order to detect a sensitivity of 100% (95% CI, 99%-100%) for the clinical prediction rule, a sample of at least 100 patients with bacterial meningitis would be required. Based on our estimate that 5% of adults with CSF pleocytosis have bacterial meningitis, we would need to cap- ture 2000 adults with CSF pleocytosis.

http://dx.doi.org/10.1016/j.ajem.2016.04.003

case definitions”>1266 R. McArthur et al. / American Journal of Emergency Medicine 34 (2016) 1265-1267

Case identification

We identified patients who underwent a lumbar puncture in the ED between November 2003 and October 2013. We included all pa- tients with meningitis defined as a cerebrospinal (CSF) white blood cell count >= 10 cells/mm³ after correction for the presence of CSF red blood cells using a standardized 500:1 ratio [6,7].

Exclusion criteria

We excluded patients with any of the following: recent neurosurgi- cal procedures, indwelling ventriculoperitoneal shunt, immune deficits (congenital or acquired), presence of purpura, coexisting bacterial infec- tion requiring Parenteral antibiotic therapy, critical illness or antibiotic therapy within the last 72 hours [3]. Critical illness was defined as obtundation, need for ventilator support or hemodynamic instability re- quiring repeat fluid boluses or vasoactive agents.

Data collection

We abstracted the following from the medical record: demo- graphics, clinical presentation, laboratory and culture results, neuroim- aging (cranial computed tomography or magnetic resonance imaging) as well as treatment received. When recorded, we noted whether anti- biotics were given prior to the diagnostic LP.

Case definitions

We defined a case of bacterial meningitis as a patient with a positive CSF culture or with a CSF pleocytosis and a positive blood culture for a bacterial pathogen. Cultures positive for the following pathogens were considered to be contaminants: Corneybacterium diptheroids, Staphylo- coccus epidermidis, and Streptococcus viridans [3]. We defined a case of Aseptic meningitis as a patient with CSF pleocytosis and negative CSF and blood cultures.

Statistical analysis

The ED encounter was the unit of analysis. We present proportions as percentages with a 95% confidence interval (CI) and continuous var- iables as a median with an interquartile ranges.

We then applied the Bacterial Meningitis Score to the study popula- tion which includes the following five high-risk predictors: positive CSF Gram stain, CSF absolute neutrophil count (ANC) >= 1000 cells/mm³, CSF protein >= 80 mg/dL, peripheral blood ANC >= 10 000 cells/mm³ or a histo- ry of seizure prior to or at time of presentation [3]. Patients with none of the high-risk predictors were classified as very low risk for bacterial meningitis. Those with one or more high-risk predictors were classified as not low risk for bacterial meningitis. Patients with one or more miss- ing predictors and no high risk predictors could not be classified. We re- port the sensitivity, specificity, negative predictive value, and positive predictive value of the Bacterial Meningitis Score rule for bacterial meningitis.

All data analysis was performed using IBM SPSS Statistics version

23.0 (IBM SPSS Statistics, IBM Corporation).

Results

We identified 711 ED encounters that met study inclusion criteria. Of these, 270 encounters (40% of identified encounters) were excluded for the one or more of the following reasons: 77 due to a recent neurosur- gical procedure, 150 an immune deficiency, 12 co-existing bacterial in- fection, and 31 critical illness at presentation.

Of the 441 remaining ED encounters, the median patient age was 40 years [interquartile range 27-54] and 190 patients (43%) were

male. Neuroimaging was performed in 285 patients (65%) prior to the performance of the diagnostic LP.

Of the 4 patients with bacterial meningitis [0.9%; 95% confidence in- terval (CI) 0.4%-2.4%], 2 had a positive CSF culture alone and 2 had CSF pleocytosis and a positive blood culture. The causative bacterial patho- gens were as follows: Neisseria meningitides (1 patient), Staphylococcus aureus [1], Streptococcus Group G [1] and Streptococcus pneumoniae [1]. The remaining 437 patients had aseptic meningitis (99.1%; 95% CI,

97.6%-99.6%).

We then successfully applied the Bacterial Meningitis Score to 429 ED encounters (97% of encounters). The 4 patients with bacterial men- ingitis were classified as not low risk (Table 1). More than half of the ED patients with meningitis were classified at very low risk for bacterial meningitis score by the Bacterial Meningitis Score.

Of the 441 study encounters, treating clinicians hospitalized 372 pa- tients (84%). If the Bacterial Meningitis Score had been prospectively ap- plied and very low risk patients had been discharged, the admission rate could have dropped to 48%.

Discussion

In the era of widespread conjugate vaccination, bacterial meningitis has become quite rare [8]. We only identified four adults with bacterial meningitis meeting study criteria that presented to the ED of a large ac- ademic referral center over the 10-year study period reflecting wide- spread conjugate vaccine uptake [1]. In our study, the Bacterial Meningitis Score, a clinical prediction rule derived and validated in chil- dren, accurately identified all the adults in our study with bacterial meningitis. Prospective implementation of this clinical prediction rule has the potential to substantially reduce the hospitalization rate for adults with meningitis by identifying low risk patients who could po- tentially be managed as outpatients while awaiting bacterial culture results.

Bacterial meningitis, although rare, remains a disease with high morbidity and mortality. No single laboratory test or readily available biomarker has demonstrated an adequate ability to discriminate be- tween aseptic and bacterial meningitis. A previous multivariate model derived and validated in adults with meningitis presenting to a single academic institution in the 1980s had excellent accuracy [9]. However, the clinical applicability of this model has been challenged due to the dramatic changes in bacterial meningitis epidemiology over the past 3 decades stemming from the widespread uptake of bacterial conjugate vaccines.

The Bacterial Meningitis Score has derived and subsequently validat- ed in 9 studies conducted in 6 countries [4,10]. However, of the more than 5000 patients included in the meta-analysis of the clinical predic- tion rule performance, none were adults. One study of 111 adults was not included in this meta-analysis due to missing data including antibi- otic pretreatment status [5]. While the prior adult study included a smaller number of adult patients they found a larger number of bacterial meningitis cases and also demonstrated a NPV of 100%. Our study is the largest external validation of the Bacterial Meningitis Score in adult ED patients to date and suggests that the Bacterial Meningitis Score

Table 1

Performance of the Bacterial Meningitis Score in adult ED patients with meningitis.

	Bacterial meningitis	Aseptic meningitis
No Bacterial Meningitis Score Predictors	0	218
One or more Bacterial Meningitis Score	4	207

predictors

Sensitivity 100% (95% CI, 40%-100%)

Specificity 51% (95% CI, 46%-56%)

Negative Predictive Value 100% (95% CI, 98%-100%)

Positive Predictive Value 2% (95% CI, 1%-5%)

R. McArthur et al. / American Journal of Emergency Medicine 34 (2016) 1265-1267 1267

performs with a high degree of accuracy in a wide variety of clinical set- tings and across the age spectrum.

Our study has several limitations. First, our study was retrospective. However, we collected objective laboratory results and clinical factors that were likely to be accurately recorded in the medical record. Second, we were unable to accurately abstract the timing of antibiotics adminis- tered either at home or in the ED relative to performance of the LP in many cases. As a result, we may have falsely included patients with pretreated and potentially falsely negative bacterial cultures. Third, bac- terial cultures also may be falsely negative due to delay in laboratory processing or low volume blood samples. However, none of the culture-negative patients later developed symptoms concerning for bacterial meningitis after stopping antibiotic treatment. Fourth, patients with bacteremia and CSF pleocytosis may have had an inflammatory re- action but not true bacterial meningitis. However, we chose a conserva- tive bacterial meningitis case definition for our low risk clinical prediction rule. Fifth, patients with known immunocompromised states at time of presentation were excluded from analysis, but those with un- diagnosed conditions or leukopenia may have been inadvertently in- cluded. Last, bacterial meningitis was very uncommon in our study population reducing the certainty in our estimates of the prediction rule performance. Our results may not be applicable to regions with lower conjugate vaccination and thus higher bacterial meningitis rates. However, this rarity highlights the importance of identifying low risk patients in order to minimize unnecessary interventions.

Bacterial meningitis is rare among adults presenting to the ED, al- though hospitalization rates remain high. The Bacterial Meningitis Score accurately identifies low risk adults with meningitis and could

be used to assist clinical decision-making to avoid unnecessary hospital- izations for low risk patients. Given the low numbers of patients with meningitis included in our study, we recommend testing the accuracy of this prediction model in other populations. Looking ahead, careful im- pact analysis of prospective application of the Bacterial Meningitis Score of adult patients with meningitis will be needed.

References

AJEM