Assessing severity of anaphylaxis: a data-driven comparison of 23 instruments

The severity of an allergic reaction can range from mild local symptoms to anaphylactic shock. To score this, a number of instruments have been developed, although heterogeneous in design and purpose. Severity scoring algorithms are therefore difficult to compare, but are frequently used beyond their initial purpose. Our objective was to compare the most used severity scoring instruments by a data-driven approach on both milder reactions and anaphylaxis.

All positive challenges to foods or drugs (n = 2828) including anaphylaxis (n = 616) at Odense University Hospital, Denmark from 1998 to 2016 were included and severity was scored according to Sampson5. Based on recommendations from an expert group, the symptoms and values from Sampson5 were for all reactions and anaphylaxis only translated and compared by kappa statistics with 22 instruments, ranging from 3 to 6 steps.

For milder reactions, there was a significant correlation between the number of steps in an instrument and the number of challenges that could be translated, whereas all instruments were good to identify food anaphylaxis. Some instruments scored reactions more severely than Sampson5, other scored them milder and some scored food and drug challenges differently. Instruments for hymenoptera reactions were difficult to apply on food and drug reactions, and thus distributed severity differently. Algorithms hampered the translation between instruments, and 7 instruments were poor concerning drug anaphylaxis, including the only instrument developed specifically for drug reactions.

The distributions of severity differed between the 23 instruments in both food and drug allergy, and thus rendering translation especially between scoring systems with 3 and 5 grades difficult. Fine-graded and simple instruments are preferred for comparison especially among milder reactions, and instruments applied to non-intended situations may not reflect a true severity picture.