259 DEVELOPMENT AND TESTING OF AN ALGORITHM FOR RELIABLE CLINICAL IDENTIFICATION OF ADVERSE DRUG REACTIONS

Abstract

Despite widespread epidemiological attention to surveillance for adverse drug reactions (ADR's), the results of surveys remain difficult to interpret because identification of ADR's has been a non-reproducible act of unspecified, subjective judgment.

To improve scientific precision in the diagnosis of ADR's, we have developed an algorithm that provides detailed operational criteria for ranking the probability of causation when an ADR is suspected. The algorithm provides a scoring system for six axes of decision strategy: previous general experience with the drug, alternative etiologic candidates, timing, drug levels, and the effects of dechallenge and rechallenge. The sum of the scores is ordinally partitioned to rate the candidate ADR as definite, probable, possible, or unlikely.

The reproducibility and validity of the algorithmic scores were tested with a clinical spectrum of 30 cases of suspected ADR's. Three physicians independently using the algorithm agreed completely in 20 cases; were within one ordinal category of each other in the remaining 10; and scored between 0.911 and 0.956 in the Cicchetti index of pair-wise agreement. The algorithm scorers' consensus agreed with ‘truth’, as defined by the implicit judgments of 2 experts, in 26 of 30 cases.

The algorithm provides a formal, valid, and reproducible method of identifying the likelihood of an ADR, and can allow accurate assessment of this important, complex clinical phenomenon.

Log in or create a free account to read this content

Gain free access to this article, as well as selected content from this journal and more on nature.com

Access through your institution

or

Sign in or create an account

Continue with Google

Continue with ORCiD