Identifying patients with myasthenia for epidemiological research by linkage of automated registers

Abstract

Background: We validated a new method of identifying patients with incident myasthenia in automated Danish registers for the purpose of conducting epidemiological studies of the disorder.

Methods: For residents of a Danish county (population 484,862) in 1993-2008, we identified any hospital contacts coded for myasthenia in a nationwide patient register and any prescriptions for pyridostigmine in the county prescription register. Results from an acetylcholine receptor antibody register were linked to the data. We verified the diagnosis by a review of medical records.

Results: Subjects identified in the Patient Register (n = 83) were comparable with individuals found in the Prescription Register (n = 89) with regard to age and gender, but were more often seropositive (83.1 vs. 74.2%). Seropositivity increased to 91.6% by restricting the data to individuals recorded in both Patient and Prescription Registers (n = 71). We found that for subjects identified in both Patient and Prescription Registers the positive predictive value of the register diagnosis was 92.9% (95% confidence interval, CI, 84.3-97.7), the false-positive rate was low (2.8%), and the sensitivity was acceptable (81.2%; 95% CI 71.2-88.8).

Conclusions: Our data indicate that this novel approach of combining diagnosis register and prescription register information provides a feasible and valid method to trace incident myasthenia patients for population-based epidemiological studies.