Professor Clinical Pharmacology, Pharmacy, and Environmental Medicine, Department of Public Health, University of Southern Denmark. Odense C, Denmark
Background: The indication codes (INDO) in the Danish National Prescription Registry are rarely used. Data on indications would hold considerable value in register-based pharmacoepidemiological research. However, the validity of this variable has not yet been formally studied.
Objectives: To assess the validity of the indication codes recorded in the Danish National Prescription Registry.
Methods: This validation study was based on data extracted from the software systems of five Danish community pharmacies on 80,814 prescriptions redeemed in the randomly selected period from 4 to 16 February in 2019 and 2020. We obtained data from the two available systems PharmaNet and C2. C2 was, however, overrepresented (three of five pharmacies) compared to its actual use (about one in four pharmacies). We excluded prescriptions for magistral drugs and prescriptions with multiple daily dispensings, the latter to ensure accurate linkage to the Danish National Prescription Registry. The indication information contained in the label texts from the pharmacy software systems was used as the “gold standard” against which the corresponding indication codes recorded in the Danish National Prescription Registry were compared.
Results: The proportion of prescriptions with recorded indication codes in the Danish National Prescription Registry was 82% (66,164) but varied between pharmacy software systems, being lower for C2 (77%) than PharmaNet (91%). Correcting for the overrepresentation of C2 data in our sample, the estimated proportion of prescriptions containing an indication code is thus ≈88%. Almost 100% (66,158 of 66,164) of the prescriptions with recorded indication codes in the Danish National Prescription Registry had correctly recorded indication codes. For a random sample of 500 of the prescriptions with absent indication codes in the Danish National Prescription Registry, only 3.2% (n=16) also contained no indication information in the label texts from the pharmacy software systems. Identification of the eight most frequently recorded indication codes in the Danish National Prescription Registry for selected drugs/drug classes revealed the presence of non-specific indication codes (ranging from 5.6 to 36%).
Conclusions: The indication codes recorded in the Danish National Prescription Registry are valid and thus usable in register-based pharmacoepidemiological research. However, researchers should be aware of about 12% missingness and that recorded indication codes may be non-specific.
