For the calculation of incidence rates based on population or exposition measures a most accurate estimation of the number of fatally and non-fatally injured persons per year is needed. In Switzerland accident data is collected by the police and gathered at the Swiss federal statistical office (BFS). As a consequence, accidents not registered by police are missing in the official statistics resulting in an underestimation of incidence rates. And also a second problem arises: Statistical analyses would be only unbiased, if the registration probability of an event is by chance (Missing At Random-assumption). But the registration process by the police is obviously biased by three important factors: travel mode, type of accident and injury severity. Non-collision or fixed object collisions of cyclists for example are less equally registered than those of other road users. Also, the registration probability increases with injury severity. In 1990 the Swiss council for accident prevention (bfu) conducted a study to estimate the number of unknown cases in road accidents. Data of the registered accidents and statistics from the so called “UVG”, an accident insurance, mandatory for all employees in Switzerland, were compared and the number of unknown cases estimated. For subsequent years the results of the study were extrapolated. One of the study problems was that a large part of the Swiss population is not covered by the “UVG” statistic, e.g. children and retired persons. A second problem for a comparison was the discrepancy in the definition of injury severity: for the police data the rough KABCOU rating was used while for the UVG-statistic injury severity was defined by a categorization of insurance payments. In the meantime, an annually updated hospital discharge statistic is available for Switzerland, covering all hospital treatments. Road accident fatalities are covered by mortality statistics. So relying on four data sources, police registered data, “UVG” statistic, mortality statistic and hospital discharge data, the calculation of the number of unknown cases is updated. The paper discusses the usability of the new data source and methodological problems. (Author/publisher)
Abstract