Dictionaries and Coding in Pharmacovigilance

For some reason, many professionals working in the field of drug safety do not find the topic of dictionaries exciting. Most would admit, however, that they are of critical importance. The purpose of the dictionary is to bring order to seeming chaos. They are intended to bring some discipline to the vast number of descriptive terms that health professionals and patients use for medical conditions, and to the enormous array of medicines that the former inflict on the latter. By abbreviating the original descriptions and reducing them to some form of code or standard terms, it is possible to record the data effectively and concisely on a computer database, to search for similar medical conditions associated with unique medicinal products and to present the information in summarized format or numerical tables. The characteristics of the dictionary exert a profound effect on the data. If there are too few terms in the dictionary, then compromises have to be made when coding the data. Details that have been reported may be lost, e.g. staphylococcal bronchopneumonia and acute exacerbation of chronic bronchitis might both just become ‘respiratory infection’. If the relationships within the dictionary are not completely valid, then a case reported as ‘psychological problems’ might be transformed into ‘psychotic’ in the database – which is enough to drive anyone mad. At the other end of the scale, it could be that the dictionary accurately reflects the facts, but does not group conditions appropriately. If grouping is not effective, then it might be difficult to find items in the database. For example, if pneumonitis is classed together with some similar conditions as a respiratory disorder, but pneumonias are grouped separately under the heading of infections, then this could result in failure to identify all cases relevant to a particular safety concern about the adverse effects of a drug on the lung. Another way in which the choice of dictionary could affect one’s view of the world is, paradoxically, by being too specific. If a dictionary included 25 different types of headache, then it might be difficult to answer the simple question of whether there were there more reports of headache with the beta-blocker ololol in patients receiving active drug or placebo in a comparative trial with 100 patients in each treatment arm. The answer might be that