To identify simple, contemporary predictors of both morbidity and mortality in pediatric patients with purpuric sepsis syndrome in order to provide a basis for future study of innovative interventions. Retrospective study. An 18-bed multidisciplinary intensive care unit (ICU) in a large pediatric hospital. A total of 53 patients, ranging in age from 18 days to 17 yrs (mean 4.9 yrs) with either culture-proven meningococcal sepsis or the systemic inflammatory response syndrome with purpura, who were admitted to the ICU during the period from January 1, 1982 through March 15, 1992. A computerized database was constructed containing the characteristics of these patients at presentation, during the first 24 hrs of hospitalization, and on discharge. Single variables were screened for significance between "good" (intact survival) and "poor" (mortality or survival with significant morbidity) outcome groups. Those variables found to be most significant were then tested for sensitivity, specificity, and predictive value. The best predictors identified in this manner were then compared with the two most-cited prognosticating strategies as applied to these patients. Coagulopathy (defined as a partial thromboplastin time > 50 secs or serum fibrinogen concentration < 150 mg/dL [4.4 mumol/L]) at the referral site or on ICU admission was identified as an excellent predictor of poor outcome: sensitivity, specificity, positive and negative predictive values of a low serum fibrinogen value, being 81%, 95%, 93%, and 88%, and of prolonged partial thromboplastin time, being 95%, 90%, 86%, and 97%, respectively. Classical prognosticating strategies were found to be inadequately associated with mortality, yet comparable with coagulopathy in identifying patients destined for clinically important morbidity. We conclude that: a) outcome of pediatric patients with meningococcal sepsis or the systemic inflammatory response syndrome with purpura can be predicted rapidly, more easily, and with overall accuracy superior to classical prognostication strategies by the simple presence or absence of coagulopathy; b) when applied to a contemporary population, classical prognostication strategies lack value for prediction of mortality, yet remain valid for prediction of "poor outcome" (significant morbidity + mortality); c) when evaluating treatment strategies for such patients, the presence of serious coagulopathy may potentially be useful as an index of illness severity.