Development and Validation of a Machine Learning Algorithm to Identify Anaphylaxis in US Administrative Claims Data

This presentation reports the first predictive model algorithm to define anaphylaxis in administrative claims databases. While prior literature suggests anaphylaxis is difficult to identify accurately in claims databases, machine learning techniques used in this study yielded an algorithm that achieved a substantially higher positive predictive value than prior algorithms while retaining a similar number of cases (true positives). The positive predictive value of this anaphylaxis algorithm was 94% and the sensitivity was 92%. This validated algorithm could be considered in future safety studies using similar claims data to reduce potential outcome misclassification.