Reliability of observation variables in distinguishing infectious outcome of febrile young infants.