This paper describes an extensive study on thermal conductivity (TC) of polymer composites with randomly distributed hybrid fillers. Finite element method in combination with user-defined code is used to predict accurately the TC of these composites under many effects and effective parameters such as volume fractions (VFs) and TC ratios of fillers to that of the matrix. A literature review on the TC prediction models of hybrid-filler polymer composites is studied and discussed. The effects of particle distribution and particle size of hybrid filler are also taken into account and analyzed. It was found that these effects become important and affect significantly to the effective TC, particularly at high VF, high TC, and large particle size. Remarkably, a modified Hashin-Shtrikman model is first proposed based on an extensively numerical results. It can be widely utilized for predicting the TC of polymer composites with randomly distributed hybrid fillers accurately and effectively, regardless of non-spherical filler shape. (C) 2017 Elsevier Ltd. All rights reserved.