Existing scoring systems have failed to reflect the pathophysiological changes during ICU therapy, and do not provide reliable criteria for the prediction of outcome in surgical patients. The aim of the present project was to establish a comprehensive scoring system for daily evaluation of physiological parameters and therapeutic interventions in a surgical intensive care unit, and to identify score patterns in the course of ICU treatment to be used for prospective clinical decisions. In a prospective study of 123 consecutive patients who required intensive care for more than two consecutive days we documented 10 physiological parameters and a set of 14 therapeutic interventions on a daily basis over a total of 1274 days. Evaluation of the new scoring system, called the Heidelberg Intensive Ward Score (HDWS), included comparison with APACHE II and a set of unfavorable HDWS-patterns at different time points during ICU treatment. All variables were implemented into a notebook computer to be used at bedside. Neither HDWS nor APACHE II obtained on admission differentiated between survivors and non-survivors. In contrast, scores of non survivors were significantly higher than scores of survivors at day 7. At that point, HDWS was superior to APACHE II with respect to the predictive power as assessed by receiver operator characteristic curves. No patient who fulfilled all four unfavorable HDWS-patterns during the first week of ICU treatment survived (but these were only two patients). We conclude that the limited power of scores obtained on admission to predict outcome in surgical patients may be improved by trend analysis of scores over time which also take into account the patients' response to therapy.(ABSTRACT TRUNCATED AT 250 WORDS)