Assessment of the impact of prevention actions carried out over 20 years by textual analysis of 77 cases of forklift accidents

Objective: Between years 2000 and 2019, INRS conducted several studies aimed at reducing the number and severity of accidents at work caused by forklift trucks during a lateral tipover. At the end of this work, the question arose of the impact of this work on the accident rate during this period. The traditional indicator of monitoring the number of these accidents, collected from the statistical data of the Cnam (National Health Insurance Fund), proved to be unsuitable due to a change in nomenclature in 2013.
Method: Epicea is a database administered by the INRS, which contains information on all fatal accidents, and some of serious accidents. It is fed by the written investigations of the inspectors of the Carsat (Caisses d'assurance retraite et de la sant au travail), following an accident. A textual analysis by hierarchical top-down classification (Alceste algorithm designed by Max Reinert) was then experimented as an alternative to the exploitation of numerical data, by applying it to the accounts of accidents, on the one hand, and to the prevention measures written by the inspectors, on the other. After partitioning the available texts into elementary context units (ECU), classes of sentence segments were identified for each type of account.
Results: The monitoring, over time, of two specific classes relating to the technical aspects and the circulation of the truck, made it possible to show that their presence decreased in the texts from the years 2012 - 2013. The hypothesis was then formulated of a possible impact of the recommendations, publications and results obtained by the INRS on the accident rate. If the proposed analysis method is confirmed for other accident situations, this approach could be a fruitful alternative for monitoring the impact of prevention actions when precise figures are lacking.