This cross-sectional study uses single health system data on acute hospital visits over 6 years to determine the yield of head computed tomography examinations for actionable pathology in common psychiatric presentations and to characterize low-risk scenarios in which imaging may be avoidable.