04 October 2009

Differential, schmifferential

Nice dodge there, Dr. Shadowfax. I'd like to review a couple of journal articles for all the ER folks out there who may not be familiar with the pediatric literature, to aid you in the management of patients such as the school-age kid with pleocytosis. I'm not sure about the literature in the adult world, since I stopped caring following it closely a dozen years ago or more. In the world of kids though, it can be devilishly tricky to figure out which kids have aseptic (viral) and which have bacterial meningitis, particularly since the incredibly effective vaccines for Haemophilus influenza B and Streptococcus pneumonia have drastically reduced the number of kids who get bacterial meningitis in the first place.

Enter Dr. Lise Nigrovic from haavad (say hi while you're in Boston, Dr. S!), and her group's development of a multivariable predictive model to distinguish the two entities. Using a set of nearly 700 patients with meningitis (1:4 ratio of bacterial to viral meningitis), they first randomly-selected 1/3 of them as the "derivation set." They then analyzed this subgroup to find the variables that differ in patients with bacterial vs. viral sources of infection. They found a model that used a Bacterial Meningitis Score (BMS) with one point for any of the following:

  • CSF Protein >= 80
  • CSF ANC >= 1,000
  • Peripheral ANC >= 10,000
  • Seizure before or at presentation
Plus 2 points for:
  • CSF Gram Stain + for bacteria
Interestingly, the types of white cells found in the CSF did not enter this model. Using these variables, they used the other 2/3 of patients as the "validation set." In these patients a BMS of 0 was found to have a negative predictive value for bacterial meningitis of 100%!! OK, OK, nothing is 100% in medicine, but this is pretty close with confidence intervals 97-100%. The sensitivity of a BMS >=2 for predicting bacterial meningitis was also impressively high, 87%.

This study has since been validated in a multicenter study done in the era of widespread Pneumococcal-vaccine use.

Certainly there are many reasons for admitting a child at low risk (BMS = 0) of bacterial meningitis: vomiting, intractable pain, lack of appropriate follow-up, or lack of an appropriate (or sufficiently reassured) caregiver. However in many instances the BMS can be a useful tool in changing your disposition from "slam-dunk admit" to "home to rest and recover."

  1. Development and Validation of a Multivariable Predictive Model to Distinguish Bacterial From Aseptic Meningitis in Children in the Post-Haemophilus influenzae Era. PEDIATRICS Vol. 110 No. 4 October 2002, pp. 712-719
  2. Clinical Prediction Rule for Identifying Children With Cerebrospinal Fluid Pleocytosis at Very Low Risk of Bacterial Meningitis. JAMA Vol. 297 No. 1, January 3, 2007


  1. Like so many clinical scenarios (chest pain, syncope, PE) trying to define a good decision rule is usually a bunch of mental masturbation that does not help all that much with the individual patient sitting before you.

    I don't sweat CSF pleocytosis. If they look well they get Rocephin and go home with next day follow up and culture results. We probably send home a lot of viral syndromes that would have some pleocytosis if we tapped everyone. If they look sick they get Vanco, Rocephin and get admitted.

    BTW, excluding newborns, I can't remember the last time I saw a pediatric bacterial meningitis.

    Pediatricians are about 1-2 vaccinations away from becoming irrelevent.

  2. Anon: Thanks goodness we have the Jenny McCarthys of the world to keep us in business then! /sarcasm