Katja Biering Leth-Møller, Tea Skaaby, Flemming Madsen, Janne Petersen
Pharmacoepidemiol Drug Saf. 2020 Sep 23. doi: 10.1002/pds.5120. Online ahead of print.
The prevalence of allergic rhinitis has been increasing, with more than one in five people being affected. The objective of this study was to evaluate the validity of 13 different Danish prescription algorithms and hospital data to identify people with allergic rhinitis.
This study included 10 653 Danish adults in two time periods. Investigators used a positive serum-specific IgE and self-reported nasal symptoms as the primary gold-standard of allergic rhinitis. The secondary gold standard of allergic rhinitis was self-reported physician diagnosis. They calculated sensitivity, specificity, positive predictive value (PPV), negative predictive value, and corresponding 95 % confidence intervals for each register-based algorithm in the two periods.
All algorithms had a low sensitivity, irrespective of the definition of allergic rhinitis or period. The highest positive predictive values were achieved for algorithms that required antihistamines and intranasal corticosteroids, with a value of 0.69 (0.62 – 0.75) and a corresponding sensitivity of 0.10 (0.09 – 0.12) for the primary gold standard of allergic rhinitis.
In conclusion, due to the low use of prescription medication among those with allergic rhinitis, sensitivity was low (≤ 0.40) for all algorithms irrespectively of the definition of allergic rhinitis. Algorithms based on antihistamines and intranasal corticosteroids granted the highest PPVs. Nevertheless, PPVs were still moderate, due to low sensitivity, when applying a strict gold standard (sIgE and nasal symptoms). Studies using administrative data must consider how to reliably identify allergic rhinitis, for example, using different data sources, and how a potential misclassification will
impact their results.