The severity of illness scoring systems in use incorporate both a “case-mix description” and an “outcome prediction” capability. The review of the literature of scoring systems and their role in management and utilization programs of intensive care units has shown that the most relevant research addresses the prediction capability of the researched models. Whereas their description capability seems undisputed, the prediction capability of the scoring systems faces yet unsolved serious methodologic problems. In this review it is contended that the use of the prediction capability of the severity of illness scoring systems cannot replace for the approach of the management and utilization of intensive care units from a scientific managerial perspective, in which “improvement” is the leading keyword. “Improvement” can only be operationalized by enlightened changes of the ways of working and processes of care.