This paper addresses the problem of emergency surgery insertion into a given elective surgery schedule of an operating theater (OT) composed of multiple operating rooms (ORs). Emergency surgeries with different emergency levels characterized by waiting time targets (WTTs) arrive according to a nonhomogeneous Poisson process and can be inserted into any OR. An event-based stochastic programming model is proposed to minimize the total cost incurred by exceeding WTTs of emergency surgeries, elective surgery delay, and surgery team overtime. A perfect information-based lower bound is proposed and the properties of the optimal policies are proved. Simple heuristic policies and a stochastic optimization (SO) approach derived from the simple policies by policy improvement are proposed. Numerical experiments show that the SO significantly outperforms the others and efficient emergency insertion significantly improves the system performance. A principal component analysis is performed to show how near-optimal policies differ from simple heuristic policies.