Pediatric anthropometric test devices (ATDs) are a key tool for developing motor vehicle safety systems. Such ATDs are limited by the adult-derived biofidelity data used to guide their design. A novel force deflection sensor (FDS) for determining applied force and chest deflection during cardiopulmonary resuscitation (CPR) has been developed, but is subject to certain errors associated with deformation of the mattress and backboard upon which the patient lies during CPR. The purpose of this paper is to describe a numerical method for compensating for mattress deformation errors, and to compare the compensated chest deflections with actual measured chest deflection. A series of CPR simulations with manikins on hospital beds and stretchers was performed with the FDS sensor in place and subsequently processed with the mattress compensation technique. Average error between mattress compensated and measured maximum chest deflection was lowest on the stretcher with stretcher pad (0.3 ± 0.9 to 1.5 ± 0.5 mm) compared to two different hospital beds (0.3 ± 1.6 to 5.9 ± 1.6 mm). The methods described and evaluated herein provide a promising approach to obtain thoracic biomechanical data from live children, with the end goal to supply enhanced thoracic biofidelity requirements for development of future pediatric ATDs.