Continuous observation is a procedure used in most psychiatric inpatient units to manage acute and escalating risk in patients. Yet, it is virtually unstudied and unreported. The present study, conducted in a psychiatric teaching hospital, compared 102 inpatients who required continuous observation with 102 control subjects. Continuous observation was provided by nursing staff, and was used for 13% of inpatients. The most common reasons for continuous observation were risk of self-harm, overstimulation by the environment, and violence. Five clinical factors that predicted the use of continuous observation as well as another set of factors that correlated with a history of violence were found. Almost two thirds of patients required brief episodes lasting less than 72 hours (mean, 28 hours). These appeared to be effective and practical. However, when continuous observation exceeded 72 hours, it often became problematic and ineffective. We recommend that after 72 hours of continuous observation, patients' clinical management should be reviewed by the treatment team and changed.