基于条件概率的临床诊疗事件打包算法研究

Clinical-event packing method based on conditional probability

为了从过往的医疗数据中得到清晰可理解的过程模型,并将其合理应用于医疗决策的制定以及临床路径的改善,提出一种全新的临床诊疗过程挖掘方案,通过将相关临床诊疗事件进行合理打包来减少作为过程挖掘算法输入的事件个数,从而简化挖掘到的临床诊疗过程模型。针对临床诊疗事件打包,提出一种基于条件概率的打包算法,该算法将条件概率作为衡量事件之间关联程度的标准,并将关联程度达到一定程度的事件进行打包。实验结果表明,所提出的临床诊疗过程挖掘方案确实能够得到清晰可理解的过程模型,所提出的打包算法能够在更高容忍度的基础上得到更加精确、合理的结果。

To obtain the understandable clinical process model and apply to medical decision making, an innovation clinical processes scheme that reducing the number of different clinical events in billing data by putting correlate events into clinical-event-packages as new units of log event for further mining was proposed. Aiming at the clinicalevent package, a packing strategy based on conditional probability named CEPM to construct clinical-event-packages with correlate events was presented. The experiment results showed that the packing clinical events was a good way of generating more comprehensible clinical processes and the proposed packing method could generate packages with better accuracy and tolerance according to medical practitioners.