In this paper, we present a physics-based compact model of thermal resistance in Resistive Random-Access Memory (RRAM) devices considering (a) thermal properties of electrode materials, (b) temperature-induced variations in material thermal conductivity, and (c) parasitic heat losses through the oxide surrounding the filament. Fully coupled numerical simulation data obtained from the self-consistent solution of the drift-diffusion equation (for vacancy migration), carrier continuity equation (for electronic conduction), and the Fourier heat diffusion equation (accounting for electro-thermal heating) has been used for developing and validating the proposed model. Excellent agreement is observed between the model and the numerical simulation results. The proposed model is validated with experimentally measured conductive filament temperature obtained from two different temperature sensing methods. Finally, we present a modeling framework to estimate the potential thermal-cross talk in RRAM crossbar arrays. Our results illustrate that a power-dependent thermal resistance model significantly improves the accuracy in estimating the electro-thermal effects in RRAMs.