A recent study of U.S. crash data has shown that children 0-23 months of age in forward-facing child restraint systems (FFCRS) are 76% more likely to be seriously injured in comparison to children in rear-facing child restraint systems (RFCRS). Motivated by the epidemiological data, seven sled tests of dummies in child seats were performed at the University of Virginia using a crash pulse similar to FMVSS 213 test conditions. The tests showed an advantage for RFCRS; however, real-world crashes include a great deal of variability among factors that may affect the relative performance of FFCRS and RFCRS. Therefore, this research developed MADYMO computational models of these tests and varied several real-world parameters. These models used ellipsoid models of Q-series child dummies and facet surface models of American- and Swedish- style convertible child restraints (CRS). The dummy-seat models in FFCRS and RFCRS orientations have been validated against sled test data, and then used to simulate several real-world conditions.These simulations show that the advantage of RFCRS was maintained under conditions of oblique collisions, pre-impact braking, and loose internal belts, but not for an out-of-position case with a gap between the child's back and child restraint. In general, this research indicates that children can benefit from being seated in rear facing restraints past the age of one, and that larger child restraints should be made available in the US to accommodate these larger children. The result also indicates that further improvements in RFCRS should address the out-of-position situation to realize further reductions in injury.