Investigating the Detection of Adverse Drug Events in a UK General Practice Electronic Health-Care Database

Abstract

Data-mining techniques have frequently been developed for Spontaneous reporting databases. These techniques aim to find adverse drug events accurately and efficiently. Spontaneous reporting databases are prone to missing information, under reporting and incorrect entries. This often results in a detection lag or prevents the detection of some adverse drug events. These limitations do not occur in electronic healthcare databases. In this paper, existing methods developed for spontaneous reporting databases are implemented on both a spontaneous reporting database and a general practice electronic health-care database and compared. The results suggests that the application of existing methods to the general practice database may help find signals that have gone undetected when using the spontaneous reporting system database. In addition the general practice database provides far more supplementary information, that if incorporated in analysis could provide a wealth of information for identifying adverse events more accurately.

Incorporating Spontaneous Reporting System Data to Aid Causal Inference in Longitudinal Healthcare Data

“…In the later work, the authors proposed generating attributes based on the nine Bradford Hill causality considerations [6] that are often used by epidemiologists when manually determining causality between drugs and health outcomes. Training a classifier to distinguish between causal and non-causal relationships using five of the Bradford Hill causality consideration proposed attributes lead to a lower false positive rate that previously obtained using unsupervised methods [5] and was suitable for causal inference with big data. Unfortunately the false positive rate was still higher than desired, motivating further development of the idea by incorporating more of the Bradford Hill causality considerations.…”

“…Examples include creating constrained Bayesian networks [4] or creating features based on domain expertise in causal inference [5]. In the later work, the authors proposed generating attributes based on the nine Bradford Hill causality considerations [6] that are often used by epidemiologists when manually determining causality between drugs and health outcomes.…”

“…In [5] the authors incorporate causal inference domain expertise to extract features that could be used as input into training a classifier to identifying causal relationships between drugs and health outcomes. The features were extracted based on Bradford Hill's causality considerations [6].…”

“…The extracted features corresponded to the drug and health outcome relationship's association strength, temporality, specificity, biological gradient and experimentation. This framework considering these five Bradford Hill considerations resulted in AUC values ranging between 0.883-937 [5]. The analogy consideration was not used to create features, but was indirectly incorporated by applying a supervised learning technique.…”

Incorporating Spontaneous Reporting System Data to Aid Causal Inference in Longitudinal Healthcare Data

“…In previously work, researchers have investigated using more advanced supervised data mining methods to identify causality in longitudinal observational databases. Examples include creating constrained Bayesian networks [4] or creating features based on domain expertise in causal inference [5]. In the later work, the authors proposed generating attributes based on the nine Bradford Hill causality considerations [6] that are often used by epidemiologists when manually determining causality between drugs and health outcomes.…”