Background Canada has a Road Safety Vision of having the safest roads in the world, yet vehicle crashes have remained the leading cause of death of Canadian children for a number of years.

Objectives Determine the influence of high rates of non-participation on the estimates for correct use of safety seats for child occupants in vehicles. Examine the impact of three different criteria for determining correct safety seat use on the estimates of correct use of safety seats for children in Canada.

Methods A national child seat safety survey was conducted in 200 randomly selected sites across Canada that included both naturalistic observation of child seat safety use at intersections and a detailed vehicle inspection in nearby parking lots. Non-participation in the detailed parking lot study was high. This study reports on statistical methods for managing high rates of non-response and compared estimates of correct use using three different criteria.

Results and conclusions Results revealed that high non-participation rates introduced bias into the raw estimates of correct safety seat use. Correct use estimates also varied substantially depending on which criterion (more stringent or less stringent) for correct use was applied in the analysis. When child age was the only criterion for correct use, estimates were higher than when more stringent criteria of child height and weight were applied to estimate rates of correct use. This study identifies the importance of managing high rates of non-response in safety seat observation studies using statistical techniques. Stringent criteria for correct use may provide more accurate estimates of the correct use of safety seats. Studies of child seat use in vehicles (using voluntary participation) may benefit from the use of naturalistic observation to capture non-participants' use of child occupant restraints, as it may more accurately estimate the rates of correct use in populations.