University of Michigan
Department of Pediatrics
Evidence-Based Pediatrics Web Site

Question

• In a healthy, non-toxic appearing pediatric patient with suspected meningitis, what are the best predictors for bacterial meningitis and how good are they in ruling out this disease?

Clinical Bottom Lines

1. Some patients with low suspicion for bacterial meningitis with supporting clinical presentation, blood, and CSF studies may be candidates for outpatient management.¹
2. Because the ramifications of delayed diagnosis of bacterial meningitis are great, the ultimate decision to hold antibiotics or commit to outpatient management depends on clinician's degree of risk preference rather than any strict models or guidelines to predict bacterial meningitis.²

Summary of Key Evidence

1. Case control study of 696 patients utilized a multivariable predictive model to identify 5 predictors of bacterial meningitis: CSF Gram stain positive for organism, CSF protein >80, CSF absolute neutrophil count (ANC) >1,000, peripheral ANC >10,000, and presence of seizure.¹
2. Bacterial Meningitis Score (BMS) constructed from analytical models was validated through an independent data set separate from the derivation set.
3. Through the calculation of BMS (+2 for +CSF Gram stain, +1 for +CSF protein, +1 for +CSF ANC, +1 for peripheral ANC, +1 for seizure), a score of 0/6 produces NPV of 100% and -LR ~0. A score of >2/6 produces PPV of 87% and +LR of 33.

Additional Comments

• Although the BMS ruled out bacterial meningitis with remarkable NPV and -LR in the validation sample, there are limitations. Application of BMS opens to possibility of misdiagnosing bacterial meningitis as cited in two cases (0.8%) with BMS of 0 in the derivation set. A clinician should carefully consider the benefit of reduced hospitalization vs. cost of delayed diagnosis of bacterial meningitis in the decision algorithm.
• The study revealed significant referral bias with high prevalence of bacterial meningitis at a major tertiary center. (18%). With lower prevalence of bacterial meningitis in anon-tertiary center, the BMS score may allow lower post-test probability with greater window of opportunity for outpatient management.
• Advent of universal adoption of pneumococcal vaccine may further decrease the prevalence of pneumococcal meningitis and affect post-test probability; this was not fully accounted for in the study.
  

Citation

1. Nigrovic LE, Kuppermann N, Malley R. Development and validation of multivariable predictive model to distinguish bacterial from aseptic meningitis in children in the post-Haemophilus influenzae era. Pediatrics 2002;110:712-9.
2. Frohna J, Park S, Gopal S. Diagnosing bacterial meningitis after the Haemophilus influenzae vaccine. Arch Pediatr Adol Med 2001;155: 1307-10.
   

CAT Author: John Kim, MD

CAT Appraisers: Robert Schumacher, MD

Date appraised: January 19, 2005