Use of a stochastic simulation model to identify an efficient protocol for ovarian cancer screening