ICUs in Spain. Patients: A total of 1,441 patients aged >18 yrs who were consecutively admitted from April 1, 1995, through July 31, 1995. Interventions: Prospective data collection during the stay of the patient in the ICU. Data collected included vital status at hospital discharge as well as all variables necessary for computing the Mortality Probability Models II system at admission and during the first 7 days of stay in the ICU. Measurements and Main Results: Four logistic regression models were obtained. These models contained survival status at hospital discharge as a dependent variable and the following explanatory variables: (model 1) only the probability of dying at admission; (model 2) only the probability of dying during the current day; (model 3) the probability of dying at admission and during the current day; and (model 4) the probabilities of dying at admission and during the previous and current days. Models were evaluated using the Hosmer-Lemeshow statistic and the area under the receiver operating characteristic curve. For survivor and nonsurvivor patients, mortality probabilities obtained using the aforementioned models were compared using linear regression and the paired Student’s t-test. Although severity at admission was a statistically significant variable, models 2 and 3 produced almost the same probabilities of hospital mortality, as shown with the linear regression and paired Student’s t-test results. Conclusions: To have an accurate measurement of the prognosis, it is necessary to update the severity measure. The best estimate of hospital mortality was the probability of death on the current day. Severity at admission and at previous days did not improve the assessment of prognosis. (Crit Care Med 2001; 29:45–50) KEY WORDS: probability models; logistic regression models; hospital mortality; mortality prediction; prognosis; daily assessment; severity of illness index; Mortality Probability Models II system; intensive care; critical illness