A goal of fire management in wilderness is to allow fire to play its natural ecological role without intervention. Unfortunately, most unplanned ignitions in wilderness are suppressed, in part because of the risk they might pose to values outside of the wilderness. Although the fire management community has embraced the concept of risk assessment for all fire and fuels management activities, existing analysis approaches were not developed for the wilderness context. They focus on the management of fuels, or on suppression resource allocation, rather than managing ignitions for resource benefits. The purpose of this research was to develop a risk-based decision support metric to support wilderness fire management. The metric, escape probability, was developed to quantify the likelihood that an unplanned ignition in wilderness would escape the wilderness boundary. Using the Bob Marshall Wilderness Complex, Montana, USA, (BMWC) as the case study landscape, this metric was used to compare two different landscape scenarios in 2007: (1) an 'observed' landscape reflecting fuel conditions as a result of actual wildfire management strategies; and (2) a 'treated' landscape that reflects hypothetical fuels and vegetation assuming suppressed ignitions in 2007 had been allowed to burn. First, wildfire spread and behavior for suppressed ignitions in 2007 were retrospectively simulated using FARSITE. Hypothetical fuels layers were created for each retrospectively simulated fire by modifying the observed pre-fire fuels conditions within the simulated perimeter based on modeled burn severity. The observed and hypothetical fuels layers were then used as inputs in FSim, a large wildfire modeling system commonly used in quantitative wildfire risk analyses. Results suggest that larger treated areas arising from ignitions closer to the wilderness boundary had the greatest effect on reducing the likelihood of wilderness fire escape. Fire and fuels managers can use escape probability information during strategic decision-making and pre-season planning to allow natural fires to burn absent of suppression, as well as to evaluate the effectiveness of different risk-mitigation strategies based on how the strategies affect future opportunities to allow natural ignitions to burn.