the epidemiology of less severe injury include routinely collected information from emergency medical services, police and fire departments, and poison control centers.
The utility of existing data can be enhanced significantly by linkages across jurisdictions, which overcome the limitations of separate databases and go far in developing comprehensive information about an event, its circumstances, the occurrence and severity of the injury, the type and cost of treatment received, the outcome in terms of both mortality and morbidity, and the administrative or legal outcome. NHTSA has fostered development of linked databases by funding several states to develop Crash Outcome Data Evaluation Systems (CODES) (Johnson and Walker, 1996). CODES were initially designed to develop comprehensive data for determining the impact of safety-belt and motorcycle helmet use on the incidence and severity of injuries, health care costs, and outcome. The implementation of CODES required the linkage of police crash reports with death certificate or medical examiner data and health care data (including emergency medical services data, emergency department data, hospital discharge data, and occasionally data from the insurance claims). CODES databases are now being used to address a variety of motor vehicle injury research and evaluation questions at the local, state, and national levels. Similar linkages will be needed for other types of injuries.
Significant barriers exist to successful linkage, however, and are related to (1) limited access to databases (in some cases relevant data are collected but not computerized and, if computerized, are not readily available because of data release policies, concerns about confidentiality, and interagency politics); (2) high costs and limited resources for developing and maintaining databases; and (3) technical difficulties. When the databases to be linked use similar unique identifiers, linkage is relatively easy. However, for confidentiality reasons, most databases do not contain unique identifiers. Probabilistic matching software has been developed and used to link databases when unique identifiers are not available and to deal with inevitable discrepancies related to spelling, data entry errors, or similar problems (Johnson and Walker, 1996).
In summary, although significant advances have been made in the development of effective strategies for injury surveillance, there is still much to be accomplished. The development of information to monitor trends in nonfatal injury (nationally and locally), to determine the place of occurrence of injuries, to determine trends and mechanisms for intentional injury deaths, to identify new or