MULTIVARIABLE PREDICTIVE MODELS FOR ADVERSE OUTCOME OF INVASIVE MENINGOCOCCAL DISEASE IN CHILDREN

For prediction of adverse outcome (AO, defined as death or limb amputa tion) of invasive meningococcal disease (IMD) in children, two multiva riable models were derived and validated by reviewing the data in the medical records of patients with IMD, who ranged from birth to 19 year s of age, at three pediatric referral hospitals between 1985 and 1990 (derivation set, n = 153, 19 AO) and between 1991 and 1994 (validation set, n = 92, 11 AO). Variables in the derivation set significantly as sociated with AO (p <0.05) were entered into a logistic regression ana lysis. Because coagulation studies (prothrombin time, partial thrombop lastin time, and serum fibrinogen concentration) were available for on ly 50% of patients, two analyses were performed, either excluding (mod el 1) or including (model 2) coagulation studies. These analyses ident ified an absolute neutrophil count less than 3000/mm(3), poor perfusio n, and ct platelet count less than 150,000/mm(3) (model 1), and a seru m fibrinogen concentration less than 2.5 gm/L (250 mg/dl) and an absol ute neutrophil count less than 3000/mm(3) (model 2), as independent pr edictors of AO (p <0.05). When the models were tested on the validatio n set, the presence of at least two of the three predictors in model 1 had a sensitivity of 82% and a specificity of 97% in predicting AO; t he presence of both predictors in model 2 had a sensitivity of 89% and a specificity of 97%. These models can reliably identify patients wit h IMD at high risk of AO for whom consideration of novel therapies is justified.