[Identification of adverse events in hospitalised influenza patients]

Objective: To test the inter-observer agreement in identifying adverse events (AE) in patients hospitalized by flu and undergoing precautionary isolation measures.

Methods: Historical cohort study, 50 patients undergoing isolation measures due to flu, and 50 patients without any isolation measures.

Results: The AE incidence ranges from 10 to 26% depending on the observer (26% [95%CI: 17.4%-34.60%], 10% [95%CI: 4.12%-15.88%], and 23% [95%CI: 14.75%-31.25%]). It was always lower in the cohort undergoing the isolation measures. This difference is statistically significant when the accurate definition of a case is applied. The agreement as regards the screening was good (higher than 76%; Kappa index between 0.29 and 0.81). The agreement as regards the accurate identification of AE related to care was lower (from 50 to 93.3%, Kappa index from 0.20 to 0.70).

Conclusions: Before performing an epidemiological study on AE, interobserver concordance must be analyzed to improve the accuracy of the results and the validity of the study. Studies have different levels of reliability. Kappa index shows high levels for the screening guide, but not for the identification of AE. Without a good methodology the results achieved, and thus the decisions made from them, cannot be guaranteed. Researchers have to be sure of the method used, which should be as close as possible to the optimal achievable.