Evaluation of the reproducibility and validity of the use of the causality assessment method for poisonings designed and applied by the French poison control centres.

Introduction: French poison control centres have developed and use a standardised causality assessment method based on a decision tree for poisoning cases involving a wide range of xenobiotics, which includes five ascending levels (I0-I4) and six determinants. This study was designed to evaluate inter-rater reliability and the validity of using this method.

Methods: A reference group of toxicologists identified five categories of cases - based on the route of exposure with two circumstances for the oral route - recorded in the French National Database of Poisonings. The reference group assessed by consensus the level of causality of each randomly selected case as the reference level. Toxicologists from poison centres, not belonging to the reference group, were selected as raters. Inter-rater reliability was assessed using weighted Light's kappa. The results of raters were compared against the reference to test the validity of the method. A subgroup analysis of inter-rater reliability was also performed according to rater experience, case category, and causality determinant.

Results: Nineteen raters reviewed the 86 cases selected by the reference group. Kappa was equal to 0.55 (moderate agreement). Sensitivity was 0.90 and specificity was 0.62 when comparing I0 versus I1-I4 classes. The agreement between raters increased with experience except for the most experienced group. Ocular route of exposure had the highest kappa (0.70) among the five case categories. Kappa for the causality determinants varied from 0.31 (exposure) to 0.54 (bibliographical references).

Discussion: The causality assessment of poisonings was carried out on real-life cases. Our results are close to those of studies on causality methods in pharmacovigilance and nutrivigilance, despite a wider scope of application.

Conclusion: Causality assessment method employed by French poison control centres is useful, although coding of some determinants should be improved. Further refinement of the causality assessment method will also further enhance its utility.