The use of computerized record linkage for linking vital, ill- health and other registration records into individual life histories for medical and health research have long been emphasized. Many of the record linkage methodologies are utilizing the discriminating power of common personal identifying information available on those documents as matching criteria. These identifiers are not unique in a population and are subject to a multiplicity of changes, errors and omissions in reporting and recording. This causes many factors which contribute to varying linkage results among the methodological developments and applications. The need for accurate and effective linkage results for scientific study is essential and this need underscores the importance of developing research in linkage methodology. The proposed research will provide a comparative study of two probabilistic methods, one developed in Hawaii and the other in Canada. This research stresses an experimental approach under which test data representing specific error conditions will be used. Since true links will be known in the test data causes responsible for failure to accept true links and to reject false links can be determined. Linkage problems such as multiple links, weighting for partial agreement and cross-comparison, and name aliases will also be studied. The results will provide a quantitative basis for refinement and improvement of linkage methodology.
