A Predictive Model for the Diagnosis of Allergic Drug Reactions According to the Medical History.

Abstract

Quantification of the risk of an allergic drug reaction through the medical history is essential in clinical decision making. However, in normal clinical practice, this evaluation is generally entirely subjective. The objective of this study was to construct a mathematical model to predict the risk of allergic drug reactions using the data collected in the medical history. A total of 696 active principles, corresponding to 466 patients aged more than 14 years attending the Allergy Service of the University Hospital of Salamanca, were included. Simple binary logistic regression was used to determine associations between variables from the medical history and the final diagnosis, to construct a predictive model. Variables useful in predicting a final diagnosis of allergic drug reaction were age, sex, drug class, number of active principles, time to the reaction, number of doses, clinical presentation suggestive of allergic disease, and time to medical consultation. True adverse drug reactions were estimated to occur in 20% of active principles. However, possible allergic reactions could only be ruled out in 52.2%. The use of mathematical models could greatly improve the discriminatory capacity of the medical history. Both the overdiagnosis and underdiagnosis of allergic drug reactions should be considered a public health problem.