Purpose <p>The biofidelity of anthropomorphic test devices directly affects the evaluation of safety performance of child restraint systems. The purpose is to enhance the biofidelity of Q3 child dummy by chest structure reconstruction for the accurate prediction of the child injuries during a frontal crash.</p> Methods <p>The finite element model of Q3 child dummy restrained in impact shield child restraint systems was validated through a frontal sled test. Based on the validated sled test simulation models, the comparative biofidelity analyses between Q3 model and PIPER 3-year-old human model were conducted by the quantitative kinematic and biomechanical analyses. The internal chest structure difference between Q3 and PIPER 3-year-old human model is discussed, and the absence of the heart, lungs, and great vessels in the Q3 dummy leads to the low biofidelity; therefore, the chest structure and cardiopulmonary model of Q3 dummy were reconstructed to enhance the biofidelity.</p> Results <p>In comparison to the original Q3 model, the chest deflection, head forward displacement, and neck bending angle of the reconstructed Q3 model increased by 38.5, 2.2, and 17%, respectively, and the upward displacement of the hip decreased by 49%. The head swing degree of the reconstructed Q3 model is dramatically reduced during the rebound process, and the injury assessment criteria of the head, chest, and pelvis can reach more than 95% of the level of the PIPER 3-year-old human model.</p> Conclusions <p>This study shows that the chest reconstruction can significantly improve the biofidelity of the Q3 dummy, and future study is recommended to optimize the spinal structures of the Q3 model for further enhancement of biofidelity.</p>