The purpose of our study was to evaluate the impact of shoulder belt use on motor vehicle crash ejection, morbidity and mortality. We analyzed motor vehicle crash records linked to hospital inpatient data for front seat occupants of passenger cars in Utah between 1994 and 1996 (n = 103,035). Stochastic simulations were used to adjust for possible seatbelt misclassification. There were 276 (0.3%) occupants coded as using only a shoulder belt. The adjusted odds of ejection for shoulder only belted occupants was higher compared to lap-shoulder belted (odds ratio (OR) = 18.9; 95% confidence interval (CI) = 15.1, 25.1) and lap only belted occupants (OR = 4.3; 95% Cl = 2.9, 7.7). There was no difference in the odds of ejection for an occupant using a shoulder belt only and an occupant using no seatbelt (OR = 1.1; 95% CI = 1.0, 1.3). Occupants using a shoulder belt only were more likely to sustain a fatal or hospitalizing injury than lap-shoulder belted (OR = 2.3; 95% Cl = 1.9, 3.0), and lap only belted occupants (OR = 1.8; 95% CI = 1.3, 2.7), while controlling for other covariates. Occupants using only a shoulder belt had the same odds of a fatal or hospitalizing injury as unbelted occupants (OR = 1.1; 95% Cl = 0.9, 1.4). Average hospital inpatient length of stay, charges and injury severity scores were similar for all restraint types. These results stress the need for the use of a lap belt in conjunction with the shoulder belt.