Abstract
BACKGROUND: The interface between acute hospital care and residential aged care has long been recognised as an important issue in aged care services research in Australia. However, existing national data provide very poor information on the movements of clients between the two sectors. Nevertheless, there are national data sets which separately contain data on individuals' hospital episodes and stays in residential aged care, so that linking the two data sets-if feasible-would provide a valuable resource for examining relationships between the two sectors. As neither name nor common person identifiers are available on the data sets, other information needs to be used to link events relating to inter-sector movement. METHODS: Event-based matching using limited demographic data in conjunction with event dates to match events in two data sets provides a possible method for linking related events. The authors develop a statistical model for examining the likely prevalence of false matches, and consequently the number of true matches, among achieved matches when using anonymous event-based record linkage to identify transition events. RESULTS: Theoretical analysis shows that for event-based matching the prevalence of false matches among achieved matches (a) declines as the events of interest become rarer, (b) declines as the number of matches increases, and (c) increases with the size of the population within which matching is taking place. The method also facilitates the examination of the trade-off between false matches and missed matches when relaxing or tightening linkage criteria. CONCLUSION: Event-based record linkage is a method for linking related transition events using event dates and basic demographic variables (other than name or person identifier). The likely extent of false links among achieved links depends on the two event rates, the match rate and population size. Knowing these, it is possible to gauge whether, for a particular study, event-based linkage could provide a useful tool for examining movements. Analysis shows that there is a range of circumstances in which event-based record linkage could be applied to two event-level databases to generate a linked database useful for transition analysis.