Discussion
Our study evaluated the adjudication processes for the causality assessment of ADR and the agreement between WHO-UMC. Data suggested slight interrater agreement for WHO-UMC, according to Lands and Koch scale.[12] Previous studies have reported interrater agreement ranging from fair to substantial agreement in the ADR causality assessment involving inpatients.[3, 13] In hospitalized patients with high complexity of clinical condition, this classification is particularly difficult, since many factors are involved in the scenario, such as polypharmacy, multiple comorbidities, different healthcare teams with diverse clinical experience involved in patient care.
WHO-UMC system is based on standardized criteria, demanding minimal expertise in pharmacovigilance process, although clinical practice is an important requisite for judges. This dependence on the individual judgement and different judges’ background could also explain the expected weak reproducibility and slight interrater agreement, as cited by Khan et al., [4] and demonstrated by our findings. Nevertheless, from the perspective of patient safety, the evaluation performed by a multiprofessional team could enrich the assessment of ADR given the diverse experiences and subjectivity. In the field of pharmacovigilance, the search for consensus among professionals is highly desirable to improve the quality of the process instead of prioritizing their individual view.
In pharmacovigilance, most cases are considered only suspected, classified as probable or possible, with rarely definite or unlikely cases.[14] Our study demonstrated this aspect, considering that all 90 cases evaluated by Naranjo algorithm were classified in these categories and 93% of ADR were categorized similarly by WHO-UMC system. The assessment using WHO-UMC is not dependent on a score, allowing a discreet distribution among the categories when compared with Naranjo algorithm. This method has been shown to be easy to assess the causality of ADR, however, the influence of confounding variables was not considered.
Considering preventability, our results revealed about half of cases were considered definitely preventable or probably preventable, in consonance with a meta-analysis conducted by Hakkarainen et al. [15] To our knowledge, there is no previous study designed to compare the causality assessment and preventability of ADR developed by multiprofessional teams. We should address some limitations of our study, including fixed sample size, small number of judges and intrinsic interrater variability that hindered to obtain homogeneous results in this investigation. Besides, we did not compare how much better for clinical practice would be the consensus of three professionals from different areas than a consensus of three professionals from the same area.
Finally, to date, a great number of preventable ADR reinforce the need for improvements in pharmacovigilance strategies in hospital settings. Variability in the assessment of ADR could result in heterogeneous decisions in drug safety. Our results suggest that the involvement of interdisciplinary teams could improve the quality of the monitoring process of ADR. Further studies focused on standard definitions and multiprofessional consensual assessment should also be considered as a strategy to seek for improvements in patient safety.