Large Scale Data Collection Systems - the Nass Experience 856071

The National Accident Sampling System (NASS) has grown from a pilot data collection system with 10 teams in 1978 to the current system of 50 teams using over 130 accident data researchers in 27 States. At the end of 1984, there were over 45,000 cases in the NASS files with each case including approximately 400 different data elements. This large system with widely separated collection sites requires that each team use specific operational procedures to assure the consistency, uniformity, and reliability of data. The overall goals of NASS are to create and maintain a network, of collection teams that use specific operational procedures to assure consistency, validity, and reliability of the resultant data file. Since NASS attempts to serve statistical and clinical needs, both creativity and compromise have been employed in developing the field capabilities and in maintaining the system so that useful data are acquired. It is important for the users of any data file to be aware of the definitions and limitations of the data, but it is especially important in a system such as NASS where such a large amount of data is gathered and the collection occurs at such widely separated locations under varying local conditions and regulations. Current procedures for the selection of cases, collection of data, and quality control are described. Sampling revisions that will be introduced in 1985 to enhance the file usefulness and significant field operational changes are also discussed. The paper concludes with the argument that NASS, or any similar system, must be concerned as much with the qualifications, training, and motivation of field personnel as with statistical design.