A Probabilistic Approach to the Patient Identification Problem

Abstract Patient identification, viz, the association of a person with a hospital medical record number is the foundation of all hospital information systems. An advanced empirically based algorithm has been utilized at the University of California, San Francisco Hospitals and Clinics for patient identification since 1978. A new algorithm based on a probability model has been designed and tested and found to provide a meaningful enhancement over the previous highly successful system. The algorithm provides the system user with a probability estimate of a match between the information provided by a person presenting to the hospital and records contained in the patient identification system file. This probability based algorithm has the advantage of not only providing weights for matching specific variables in a record, such as last name, first name, date of birth, etc., but has the ability to provide quantitative differentiation within a single variable, such as last name, even when an identical match is found.