In frontal crashes, drivers restrained by a seat belt and airbag are at elevated risk of serious thoracic injury compared with most other body regions. Hybrid III (HIII) sternum deflection has some ability to predict injury, but its utility is limited by several considerations. The Test device for Human Occupant Restraint (THOR) was developed to address limitations of HIII, but it has not been sufficiently validated under combined loading from a belt and airbag. Thirty-five crash tests were conducted with a THOR in the driver seat. Logistic regression was used to assess the ability of THOR metrics to predict injury outcomes in 57 real-world crashes involving matched vehicle designs. Results showed R_max was inversely related to Abbreviated Injury Scale (AIS) ≥3 injury outcome, with a 4 mm increase associated with an injury odds reduction of 48% (p = 0.04). By contrast, increasing shoulder-belt load was estimated to increase the odds of both AIS≥2 and AIS≥3 injury, with both effects significant at alpha = 0.05. Additionally, several THOR metrics suggesting greater airbag loading were associated with higher R_max values in the test data but reduced field injury risk. The biofidelity of THOR under combined restraint loading should be further investigated.