Keywords: data linking, probabilistic record linking, client tracking
 
 

Discussed here are issues related to the development and application of a probabilistic record linking system for New York State?s Office of Alcoholism and Substance Abuse Services (OASAS) Client Data System (CDS). With over 2.4 million client transactions recorded since 1991, the OASAS CDS incorporates a wealth of administrative and epidemiological data. The usefulness of this data is limited, however, because most treatment programs utilize unrelated client identifiers that vary in scope and accuracy. It is anticipated that a standardized, probabilistically-based, system of identifiers will afford a comprehensive system-wide modeling of client flow and treatment utilization. Given that client anonymity must be preserved, such identifiers are being developed from fragments of the client?s personal data such as (1) gender, (2) date of birth, (3) last four digits of social security number, and (4) first two characters of last name. Various likelihood ratio approaches will be discussed and evaluated. This research is supported by grant 5-H79-T1112237 from the Center for Substance Abuse Treatment.