We studied the hospital course of 1148 consecutive intensive care unit (ICU) admissions to test the feasibility of identifying patients suitable for early transfer. Based on the type of treatment each admission received during the initial 16 hours in ICU, we divided the patients into active treatment or monitored categories. Which of the 513 monitored admissions received active treatment before discharge was analyzed with a multivariate logistic regression analysis, using variables such as age, sex, indication for admission, and a new severity-of-illness scale. The most important variable in identifying low-risk monitored patients was the severity of illness measure, which performed well in both estimation and validation data sets. Within the 513 monitored admissions, 154 had predicted risks of requiring active intensive therapy of less than 5 per cent. Only five persons actually received such treatment. This approach might assist in reducing the ever-increasing demand for intensive care.