Mining consequence events in temporal health data

It is useful, sometimes crucial in medicine domain, to discover a temporal association or causal relationship among events. Such a mining problem is often challenging because 'consequence events' may not reliably occur after each trigger event of interest. This makes it difficult to apply existing temporal data mining techniques directly to real world problems. In this paper, we formalise the problem of mining consequence events of newly-introduced interventions. We combine the Before-After-Control-Impact (BACI) design with frequent pattern mining techniques to define an interestingness measure called consequency. We then propose a Multiple Occurrence of Target events Mining (MOTM) algorithm. MOTM is applied to the real world problem of monitoring the consequence effects of newly-marketed medicines in linked administrative health databases. The results for the case of the cholesterol lowering drug atorvastatin highlight the consequence events with lowest negative consequency values, which suggest replacement of existing therapies with the new one. The consequence events with highest consequency values are likely to be associated with adverse reactions of atorvastatin or treatments of cardiovascular (or associated) conditions. Sensitivity examination of MOTM on another drug further illustrates its effectiveness.