Inferring glaucoma status from prescriptions, diagnoses, and operations data: A Danish nationwide study

Abstract

Purpose To assess a new method for inferring glaucoma status using prescriptions data. Methods The study population comprised all individuals living in Denmark in the period 1995 to 2018 and included 6,930,571 individuals. We used information from The National Prescription Registry on claimed prescriptions as the basis for our study (N = 223,592). We inferred glaucoma status using data on claimed prescriptions, in-hospital ICD-10 diagnoses, and in-hospital glaucoma surgeries. We infer glaucoma status in three ways using the prescription pattern: glaucoma inferred by (i) the use of a first claimed prescription, (ii) the use of a second claimed prescription with a gap of at least 90 days, and (iii) the use of a third claimed prescription for glaucoma medication, again with a gap of at least 90 days between prescriptions. Furthermore, we compared the results with alternative indications for glaucoma, namely in-hospital ICD-10-diagnosed glaucoma and in-hospital glaucoma surgery. Results We first determined that glaucoma status could be inferred from claimed prescription data and found that a single claimed prescription was highly correlated with the more restricted composite measure of glaucoma (R2 = 0.80, p <0.0001), with a kappa coefficient of 80%. Focusing on individuals with a confirmed in-hospital glaucoma diagnosis, we found a high sensitivity of 88% using anti-glaucomatous prescriptions as a surrogate marker for primary open-angle glaucoma (POAG). We then derived several descriptive insights. The prevalence of glaucoma increased during the period from 1996 to 2018, while the incidence was constant. We also found a decreasing trend in the ratio of the number of people diagnosed annually in hospitals to the number of people filling prescriptions. This indicated a relative increase in the number of patients treated or managed in the secondary sector. Finally, using data on diagnoses and claimed prescriptions, we found that the proportion of total noncompliant patients, i.e., patients who do not claim their prescription at any time in the study period (two decades) was at most 11.8%. This share is calculated on the basis of diagnosed patients who did not have surgery. The results was not sensitive to the glaucoma inference rule. Conclusion Anti-glaucomatous medicine prescriptions can be used to infer glaucoma status, with useful implications for epidemiological research. The sensitivity is particularly high for primary open-angle glaucoma (POAG).