Many of the common social, economic, occupational and environmental health concerns of today are complex and multi-faceted. Census of Population, vital statistics, cancer morbidity, radiation exposure and work histories are important national data sources because of their comprehensiveness and inclusiveness over time. Record linkage is the process of bringing together two or more data sources relating to the same individual, family or entity. This paper examines some practical methods recently developed to improve the searching, comparison and grouping of records in a typical generalized linkage process. The data sources used to illustrate this are from linkages using the Canadian Birth Data Base (from 1985), the Canadian Cancer Data Base (from 1969) and the Canadian Mortality Data Base (from 1950). In particular, the procedures developed for a birth-death linkage have been used to improve:
a) the quality of the data files themselves (e.g. by identifying duplicate and missing registrations),
b) the accuracy of carrying out generalized linkages,
c) the analytical value of the vital statistics data (e.g. for a study of maternal education and fetal and infant mortality) and d) the development of new health indicators (e.g. using birth weight, gestational age and age at death variables).
The increasing capacity of users to incorporate statistical information in their decision-making and research challenges statistical agencies to improve the quality and breadth of their information. This increasing demand is also accompanied by a rapidly changing technological environment and the increasing skill of the workforce for bringing together and analyzing data. For example, vital statistics, census, administrative and survey data are used for a multiplicity of purposes, and data are integrated from a number of different sources. One important tool to achieve the integration of these data is by computerized record linkage. Record linkage is the bringing together of two or more records, usually in order to match up those records relating to the same individual, family, event or entity. This paper will examine some of the practical methods developed at Statistics Canada to improve the searching, comparison and grouping of records in a typical generalized linkage process. Some background is also provided as to why record linkage of health data is needed, and some of the current initiatives that are furthering this work in Canada.