Equivalent continuum models estimating the deformability of complexly fractured rock masses may incur significant discrepancies. This paper proposes an equivalent discrete fracture network (E-DFN) method to proficiently capture the non-continuous characteristics of rock masses while significantly reducing computational cost compared with traditional discrete fracture network (DFN) models. A significance index of individual fractures is defined to quantitatively evaluate their impact on the elastic modulus. Accumulative significance indices are then generated to define E-DFNs containing representative quantities of fractures. Numerical simulations are conducted with E-DFN models using UDEC (universal distinct element code). The relationship between elastic moduli of complexly fractured rocks and models with E-DFNs is established by an extrapolation function obtained through regression analyses based on an assumption of the marginal diminishing effect of modulus reduction. The proposed method is validated by stochastic DFNs with different geometric configurations. The results show that reasonable estimation with small errors is achieved by the extrapolation of density-reduced E-DFN models. The E-DFN method strikes a balance between result reliability and computational intensity.