Global warming is expected to alter the frequency and/or magnitude of extreme precipitation events. Such changes could have substantial ecological, economic, and sociological consequences. However, climate models in general do not correctly reproduce the frequency and intensity distribution of precipitation, especially at the regional scale. In this study, gridded data from a dense network of surface precipitation gauges and a global atmospheric analysis at a coarser scale are combined to develop a diagnostic framework for the large-scale meteorological conditions (i.e. flow features, moisture supply) that dominate during extreme precipitation. Such diagnostic framework is first evaluated with the late 20th century simulations from an ensemble of climate models participating in the Coupled Model Intercomparison Project Phase 5 (CMIP5), and is found to produce more consistent (and less uncertain) total and interannaul number of extreme days with the observations than the model-based precipitation over the south-central United States and the Western United States examined in this study. The framework is then applied to the CMIP5 multi-model projections for two radiative forcing scenarios (Representative Concentration Pathways 4.5 and 8.5) to assess the potential future changes in the probability of precipitation extremes over the same study regions. We further analyze the accompanying circulation features and their changes that may be responsible for shifts in extreme precipitation in response to changed climate. The results from this study may guide hazardous weather watches and help society develop adaptive strategies for preventing catastrophic losses.