A total of 5,790 intensive care unit (ICU) admissions from 13 tertiary care institutions were studied to identify patients who were at such low risk of receiving unique ICU therapies that admission might have been avoided or the length of ICU stay reduced. We used acute severity of disease on admission to the ICU along with the type of disease or surgery to risk stratify individual ICU patients. Among 1,941 patients who only received monitoring services on admission to the ICU, 1,358 (70 percent) were predicted to have less than a 10 percent risk of requiring subsequent active ICU treatment. Only 58 (4.3 percent) of these low-risk patients actually received active treatment. The identification of low-risk patients was equally accurate in estimation and validation data sets. Our methods should allow physicians and hospitals to assess their current ICU utilization and, if appropriate, guide reductions in use.