Homepage > MCMC

Data Mining and MCMC

Case Study. To improve drug safety it is important to develop methodologies detecting adverse drug events using postmarketing drug surveillance data. A strong association of drug and adverse reaction forms the basis for further epidemiological study and consequently for regulatory actions. Data mining helps clinical reviewers to sort through millions of potential drug-event combinations, and to prioritize reviews to undertake. Data mining concerning postmarketing drug surveillance is called medical signal detection.

Adverse event reporting system. Adverse event reporting system (AERS) is created to monitor a possible causal relationship between drug and event. The database contains the information about the entire list $ D$ of medical products and $ R$ of medical terms of adverse reaction. Each event is reported exactly once alone with the list of medical products prescribed to a patient at the point of event, say ``Rosinex & Ganclex,'' and the list of medical terms describing adverse events, say ``Nausea.''

Frequency of event. Each event is reported with the list $ A$ of drug names and $ B$ of adverse reactions, and the entire data are summarized in terms of the frequency of such events, denoted by $ N_{A,B}$. Note that a pair $ (A,B)$ is not necessarily labeled as a valid association of model. For example, an adverse event of ``Rosinex & Ganclex'' and ``Nausea'' is reported, but the drug combination of Rosinex and Ganclex may not be necessarily the cause of nausea.


© TTU Mathematics