The research reported in this paper is part of a larger effort to achieve better signal-to-noise ratio, hence accuracy, in pharmacovigilance applications. The relatively low frequency of occurrence of adverse drug reactions leads to weak causal relations between the reaction and any measured signal. We hypothesize that by grouping related signals, we can enhance detection rate and suppress false alarm rate.
Objective
ICD-9 codes are commonly used to identify disease cohorts and are often found to be less than adequate. Data available in structured databasesFlab test results, medications etc.Fcan supplement the diagnosis codes. In this study, we describe an automated method that uses these related data items, and no additional manual annotations to more accurately identify patient cohorts.