Quality epidemiological studies require linked records to differentiate persons from events.
Medical records in the 1960's were mainly linked manually by visually comparing personal data, such as in the Western Australian Mental Health Services Registry.
In 1973, Australia's Hospital and Allied Services Advisory Council (HASAC) studied numerous ways of creating a uniform personal identifier using personal data / attributes which were readily available to the patient and healthcare provider without recalling a medical record number or finding a membership card. Using portions of the last name, first name, gender, and date of birth; only 17 out of 696 000 records could not be automatically linked. A tie-breaker, such as place of birth was recommended.
In 1993, the California Health Information for Policy Project's Interagency Working Party, working with 10 different state and local databases, suggested the use of the HASAC algorithm, but used mother's maiden name instead of place of birth to prevent duplicates requiring manual adjudication. This recommendation however could not be implemented practically because many organisations whose records did not include mother's maiden name found the cost of finding this information prohibitive.
In 2003, an attempt was made to further modify the HASAC criteria using only computerised available data to link just under 2 million records for the Kaiser Permanente Southern California Immunisation registry (KITS). As series of 26 computerised algorithms was developed which resulted in a 99.8% matching rate.
This study resulted in a cost-effected way of routinely linking records from large data bases with minimal manual input.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.