An automated procedure for determining patient numbers from episode of care records

Records of in-patient stays which may contain multiple entries for the same patient are linked. These provide a profile of all in-patient hospital contacts for the same patient and statistics from which prevalence measures can be estimated. The linkage procedure is designed to link Scottish Morbidity Records for patients of the Greater Glasgow Health Board. It is suited to in-patient records generated from a relatively small number of hospitals (25-30). The records are selected by diagnosis. A suite of computer programs automates the method, which was initially investigated manually. Two types of linkage are performed. Type 1 links on the unique identifiers of hospital group code and case reference number to link records from the same hospital or group of hospitals. Type 2 linkage is designed to link patient stays in different hospitals or groups and links records with identical surname and date of birth and then identical surname and full unit postcode. The accuracy and efficiency of the automated procedure was estimated by linking records for 6 diagnoses (lung cancer, ischaemic heart disease, cerebrovascular disease, infectious diseases, road traffic accidents and head injuries). The gross difference between the actual number of patients and the number after linkage was between 0.6 and 2.1%. The net difference was within 1% for all diagnoses except lung cancer with a difference of 1.3%. These results compare favourably with other linkage procedures.