The process of linking individual-level information from different sources has become an important tool in public health and medical research. In many cases, the records to be linked do not include a common unique identifier. The methods and software that are available to link records in the absence of a common unique identifier offer researchers limited options in terms of linkage methodology, practical application, and cost. The current proposal is to develop new methods for record linkage and to incorporate these into a software package that is designed to meet the needs of the research community. Specifically, we will investigate the use of classification and regression trees (CART) for record linkage and compare it to currently available methods. To our knowledge, the use of CART for record linkage has not been investigated previously.
Record linkage is used by a broad spectrum of health care organizations, health and medical researchers, and government health agencies. The proposed linkage system would potentially be used by all of the above for research on health and health care services and in the provision of health and medical care.
| Leiss, Jack K; Giles, Denise; Sullivan, Kristin M et al. (2010) U.S. Maternally linked birth records may be biased for Hispanics and other population groups. Ann Epidemiol 20:23-31 |
| Leiss, Jack K (2007) A new method for measuring misclassification of maternal sets in maternally linked birth records: true and false linkage proportions. Matern Child Health J 11:293-300 |
