Large-scale adoption of transgenic crops that produce Bacillus thuringiensis (Bt) toxins could greatly reduce populations of target pests and their specialist parasitoids. We used a spatially explicit model of host-parasitoid population dynamics to examine effects of Bt crops on parasitoid persistence. We modeled a 9,000-ha region with 900 fields of 10 ha, each planted with either a Bt crop or a non-Bt crop refuge. In the 13,524 simulations we conducted, the pest (i.e., host) was never lost from the region before the parasitoid. The parasitoid was lost from the region in 36% of the simulations. The host was lost after the parasitoid in 12% of all simulations. The probability of regional parasitoid loss increased as the percentage of Bt fields and rotation of field types (Bt and non-Bt) increased and as host reproductive rate and parasitoid attack rate decreased. Regional parasitoid loss also increased as the insecticide spray threshold for target pests in non-Bt fields decreased and as parasitoid susceptibility to insecticide increased. With field types fixed, regional parasitoid loss increased as parasitoid movement increased. However, with field types rotated, parasitoid movement did not influence regional parasitoid loss. Host movement did not influence regional parasitoid loss. Factors not modeled that would reduce risk of regional parasitoid loss include use of alternate hosts by the parasitoid, use of alternate host plants by the target pest, and evolution of resistance to the Bt crop by the pest. The results suggest that risk of regional parasitoid loss can be assessed from the parasitoid life history traits and reduced by increasing the percentage of refuge fields, fixing refuge locations, and increasing the insecticide spray threshold for target pests in refuges.