Prospective Validation of a Clinical Prediction Model for Lyme Meningitis in Children

Abstract

OBJECTIVE. Lyme meningitis is difficult to differentiate from other causes of aseptic meningitis in Lyme disease–endemic regions. Parenteral antibiotics are indicated for Lyme meningitis but not viral causes of aseptic meningitis. A clinical prediction model was developed to distinguish Lyme meningitis from other causes of aseptic meningitis. Our objective was to prospectively validate this model.

METHODS. Children between 2 and 18 years of age presenting to Hasbro Children's Hospital from April through October of 2006 and 2007 were enrolled if a lumbar puncture for meningitis showed a cerebrospinal fluid white blood cell count of >8 cells per μL. Cerebrospinal fluid was sent for Lyme antibody testing. The probability of Lyme meningitis was calculated by using the percentage of cerebrospinal fluid mononuclear cells, duration of headache, and presence of cranial neuropathy by using the prediction model. Definite Lyme meningitis cases were defined as cerebrospinal fluid pleocytosis with (1) positive Lyme serology confirmed by immunoblot or (2) erythema migrans rash. Possible Lyme meningitis cases were defined as cerebrospinal fluid pleocytosis with positive cerebrospinal fluid Lyme antibody. Sensitivity, specificity, and likelihood ratios for definite and possible Lyme meningitis were determined by using 10% increments of calculated probability of Lyme meningitis.

RESULTS. Fifty children were enrolled, including 14 children with definite Lyme meningitis, 6 with possible Lyme meningitis, and 30 with aseptic meningitis. A calculated probability of <10% for Lyme meningitis had a negative likelihood ratio of 0.006 for definite and possible Lyme meningitis cases. A calculated probability of >50% for Lyme meningitis had a positive likelihood ratio of 100 using these definitions.

CONCLUSIONS. A clinical prediction model using the percentage of cerebrospinal fluid mononuclear cells, headache duration, and presence of cranial neuropathy can differentiate children with Lyme meningitis from children with aseptic meningitis. Our findings suggest categories of low (<10%), indeterminate (10%–50%), and high (>50%) probability of Lyme meningitis.