(P-307) Evaluation of Availability of Information on Risk Factors in Reports of Adverse Drug Reactions in a Pharmacovigilance Database

Background: It is unknown how much information on risk factors is currently available in the reports on adverse drug reactions (ADRs) in the database of the Dutch spontaneous reporting system (SRS). The availability of this information in a SRS is important, because it can result in better risk group identification.

Objectives: Providing insight in the information on risk factors in reports in the database for the Dutch SRS.

Methods: A gap analysis was performed through a qualitative and quantitative analysis of information on risk factors for developing ADRs. Quantitative research investigated four risk factors (renal function, hepatic function, frailty and comorbidities) for developing serious ADRs related to NSAIDs and antithrombotics in the Dutch SRS since 1991. Qualitative research consisted of interviews (n=6) with pharmacovigilance (PV) assessors on the current processes involved in coding, follow-up questions and analysing information on risk factors.

Results: Quantitative research showed that information on risk factors in the reports was scarce; for renal impairment related to ADRs associated with NSAIDs the highest availability was found (16.3%). 53.6% of the investigated reports on antithrombotics and 30.2% on NSAIDs were lacking information on all four investigated risk factors. Less than 50% of the available information on risk factors in the reports was coded in MedDRA® terms. The qualitative interviews showed that the focus of assessors was on exploring new signals, instead of risk group assessment for existing ADRs. Information on risk factors was not actively pursued by assessors through follow-up. Work pressure and the absence of clear instructions on coding risk factors could lead to less thorough coding.

Conclusions: This gap analysis showed information on risk factors in reports is currently sub-optimal to conduct thorough risk group analyses. Optimalization of reporting forms, coding practices among assessors and work-instructions might help to improve the coding of risk-factors in the database. Future steps could include the use of machine learning techniques to retrieve information on risk-factors from unstructured data fields. Further automation could help the process but needs to be investigated more.