From safety reports to safety data. Exploring variability in coding of reported adverse drug reactions into standardized medical language.
Abstract
Objective: The responsibility of surveillance of medication safety after the medicinal product enters the market lies in the field of pharmacovigilance. This task hugely relies on the spontaneous adverse drug reaction reports submitted by health care professionals and the patients. The Medical Dictionary for Regulatory Activities (MedDRA) provides a standardized terminology for coding of these reports. Evidence indicates that there is a degree of inter-coder variability, especially when the information is ambiguous. This study examines the MedDRA codes assigned by Norwegian pharmacovigilance officers to a selection of ambiguous coding tasks and investigates the degree and type of variation among the chosen codes. Additionally, the study explores the reasoning and the thinking strategies underlying pharmacovigilance officers’ choice of codes and evaluates their training experience through qualitative analysis.
Methods: The research question was addressed through a combination of survey-based cross-sectional study and focus groups interviews. Participants were recruited from Norwegian Medical Products Agency, Regional Medicine Information Centre, and Norwegian Institute of Public Health by contacting the head of relevant departments in each organization. 12 coding tasks carrying varying degree of ambiguity were selected from the Norwegian Pharmacovigilance registries and included in the survey. The results of the survey were described with three approaches: 1. Based on correct or incorrect answers; 2. Based on differences in the MedDRA hierarchy at PT, HLT, or SOC level; 3. based on the type of coding variations, which were categorized into two domains of quantity and specificity. Focus group interviews were conducted, and results were analyzed by Reflexive Thematic Analysis (RTA).
Results: 26 participants took part in the survey (79% of the total population), and 8 participants were interviewed in three focus groups. Coding tasks had different degrees of variation, from 100% correct answers to 15% (4 out of 26 participants). The majority of the incorrect LLTs (70%) fell under Level 1 category where two terms have different PTs but are still connected to the same hierarchical tree further up in the MedDRA. 21% of the incorrect LLTs were categorized as Level 2, and 9% were in Level 3. In total, 88 LLTs were omitted as coders did not always code all the elements of the verbatims. Analysis of the interviews revealed 4 themes showing the 1. Challenges experienced with MedDRA, 2. Coding strategies, 3. Resources used when coding, and 4. Evaluation of their MedDRA-training.
Conclusion: Our findings revealed that ambiguity in ADR verbatims can cause discrepancy in coding. Ambiguous verbatims can result in more general MedDRA terms, loss of data and more assumption-based coding. Strategies adopted to code ambiguous verbatims include, but are not limited to: contextual thinking, consulting colleagues, contacting the reporter or simply selecting less specific terms.