In this paper, a stiffness-based model is initially assessed for fatigue damage simulation of composite materials. The model is evaluated with three sets of experimental data. A residual strength model is coupled to the choice model and a modified model is developed. Numerical results show that the modified model leads to a noticeable improvement of accuracy in fatigue life prediction of composites under two-stage loadings. In the second step of the research, an uncertainty analysis is conducted to evaluate the reliability of the achieved results. The fatigue life and strength are assumed as random variables and uncertainty analysis is done by the Monte Carlo (MC) approach. Reliability variations of predicted residual fatigue lives are studied as well. The results demonstrate that the fatigue life dispersion noticeably decreases the reliability of predicted remaining fatigue cycles.