Pharmaco-morbidity linking: a pilot study of its technical possibilities in The Netherlands

The feasibility has been studied of linking pharmacy-based dispensing data and hospital-based morbidity data by using as a linkage key a combination of gender and birth date of the patient and of his general practitioner (GP) code. Firstly, the records of one pharmacy (14,686 patients) and those of one hospital (8475 admissions in 1987) were linked on the basis of date of birth and gender. This resulted in 3,715 matching record pairs. To a random sample of 1,713 of these pairs the GP code was added. Subsequently these pairs were validated by reference to the original medical and pharmaceutical records with a procedure that guaranteed that only the name and address of the patients could be compared without having access to the clinical or pharmaceutical record of an individual. We were able to link records with a sensitivity of 91.4% (95%-confidence interval: 88.8-94.0) and a specificity of 96.0% (95%-confidence interval: 94.5-97.5). In conclusion, probabilistic linkage of pharmaceutical and medical data with the combination of gender, date of birth and GP code facilitates successful record linkage.