The condition of tree crowns is an important indicator of tree and forest health. Crown conditions have been evaluated during inventories of the US Forest Service Forest Inventory and Analysis (FIA) program since 1999. In this study, remeasured data from 55,013 trees on 2616 FIA plots in the eastern USA were used to assess the probability of survival among various tree species using the suite of FIA crown condition variables. Logistic regression procedures were employed to develop models for predicting tree survival. Results of the regression analyses indicated that crown dieback was the most important crown condition variable for predicting tree survival for all species combined and for many of the 15 individual species in the study. The logistic models were generally successful in representing recent tree mortality responses to multiyear infestations of beech bark disease and hemlock woolly adelgid. Although our models are only applicable to trees growing in a forest setting, the utility of models that predict impending tree mortality goes beyond forest inventory or traditional forestry growth and yield models and includes any application where managers need to assess tree health or predict tree mortality including urban forest, recreation, wildlife, and pest management.