Abstract & Commentary
Source: Nigrovic LE, et al. Development and validation of a multivariable predictive model to distinguish bacterial from aseptic meningitis in children in the post-Haemophilus influenzae era. Pediatrics 2002;110:712-719.
Immunization against Haemophilus influenzae and, more recently, Streptococcus pneumoniae, has shifted etiologies to viral agents in pediatric meningitis. Clinicians need rapid, dependable means to differentiate viral from bacterial disease. To enhance diagnostic accuracy among ill children, Nigrovic and colleagues retrospectively analyzed 696 pediatric patients admitted with meningitis to Children’s Hospital in Boston between 1992 and 2000. Ages ranged from 1 month to 19 years, with a median age of 5 months. Fever averaged 39.0°C for 2.1 days. Seizures were present in 7%. Final diagnoses included 125 (18%) with bacterial meningitis and 571 (82%) with presumed viral etiologies. All cases with bacterial meningitis had a cerebrospinal fluid (CSF) culture positive for a bacterial pathogen, or CSF pleocytosis of more than seven white blood cells (WBCs)/mm3 together with a positive blood culture or CSF latex agglutination test positive for a bacterial pathogen. Aseptic meningitis was diagnosed by CSF pleocytosis with negative bacterial cultures and latex agglutination tests. Children were excluded if they had clinical sepsis; recent neurosurgical procedures; focal bacterial infections such as urinary tract infection or cellulitis; or were immunosuppressed.
Etiologies of bacterial meningitis in 125 children included the following pathogens: Streptococcus pneumoniae in 79 (63%); Neisseria meningitidis in 21 (17%); Group B Streptococcus in 13 (10%); non-typable H. influenzae in six (5%); enteric gram-negative rods in two (2%); and other streptococci in one (1%). Forty-two percent of patients studied had received antibiotics within 72 hours of lumbar puncture. Four children (3%) died.
The authors randomly selected 456 patients to perform logistic regression analysis of five objective, readily available predictors of bacterial meningitis: positive CSF Gram’s stain, CSF protein greater than 80 mg/dL, peripheral absolute neutrophil count (ANC) greater than 10,000/mm3, CSF ANC greater than 1000/mm3, and occurrence of seizures. A Bacterial Meningitis Score (BMS) with a point value range of 0 to 6 was developed based on these five selected variables. Gram’s stain had a weight of two points on the six-point scoring system, while the other four variables were one point each. Average CSF ANC was 2928 vs. 108/mm3 for bacterial vs. aseptic meningitis. Average CSF protein was 230 vs. 57 mg/dL for bacterial vs. aseptic cases. The BMS then was validated by application to the other 240 randomized cases. Discrimination among variables between bacterial and viral meningitis was determined by receiver operating characteristic (ROC) curves to optimize sensitivity and specificity calculations.
In the validation cohort of 240 patients, a BMS of 0 accurately correlated with aseptic meningitis, with a negative predictive value of 100% for bacterial meningitis. A BMS of greater than 2 predicted bacterial meningitis with a sensitivity of 87% (95% CI: 72-96%) and a positive predictive value of 87% (same confidence intervals). When applied to the entire cohort of 696 patients, the BMS misclassified a mere 3.3% of cases. Of all patients with BMS greater than 2, only 21 were later proven to have aseptic meningitis. Of all children with a BMS of 0, only 2/404 (0.5%) actually had bacterial meningitis. One was a pretreated 7-month-old with S. pneumoniae meningitis but only 11 cells in CSF and peripheral ANC of only 4500. The other was a 6-month-old with N. meningitidis with 10 cells in CSF and peripheral ANC of 8000. The authors conclude that the clinically and statistically derived, five-item BMS accurately distinguishes patients with bacterial and aseptic meningitis.
Commentary by Michael Felz, MD
A BMS score greater than 2 was correlated highly with bacterial etiologies, while scores of 0 identified aseptic meningitis with high specificity. The authors further mention that theirs is the first-ever validated, multivariate study on pediatric meningitis in the post-immunization era. I was impressed that the variables selected for use in the BMS are ones readily obtainable in the ED setting, where ill children with possible meningitis are first evaluated. The cohort of nearly 700 patients lends strength to the data analysis; most were infants, in whom clinical diagnosis is much more problematic. The shifting epidemiology of meningitis to viral etiologies still demands a high index of suspicion for bacterial pathogens, infrequent though they have become. This new tool, the BMS, is a promising weapon in the systematic investigation of the febrile, seriously ill child in the pediatric ED where time, and precision, are life-and-death priorities. I give this five-item BMS a "high-five."
Dr. Felz, Associate Professor, Department of Family Medicine, Medical College of Georgia, Augusta, GA, is on the Editorial Boar d of Emergency Medicine Alert.