Multivariate prognostication in bacterial meningitis of childhood.

A multivariate method for prognosticating the outcome of bacterial meningitis was computerized on the basis of initial clinical and laboratory data in 123 patients. The model, based on likelihood ratios, is simple and rapid. At the time of diagnosis, it predicted the outcome correctly in 104/123 cases (85%) in the basic group and in 67/98 independent controls (68%). The sensitivity of the method in predicting death or mild to severe neurological sequelae was 83% in both groups. The predictive value of a calculated good prognosis was 94% in the basic group and 97% in the control group, and that of a calculated poor prognosis was 63% and 26%, respectively. The predictive system can be used for rapid assessment of prognosis in individual patients and for comparisons between groups.