Research And Application For Clinical Pathway Mining

Currently, in order to improve the quality of internal management and external service in hospital, the research points in healthcare field reform are focusing on the establishment and management of standard clinical pathway. Existing medical clinical pathway mining measures mine the relationship between medical activities based on the flow control attributes in the event logs to get the standard clinical pathway. Since the measures analyze all the event logs, the clinical pathway mined by traditional clinical pathway mining measures is just the general treatment processes for the specific disease, which can't give out the appropriate treatment against the feature of different patients. Meanwhile, the measures for the detection of abnormal clinical pathway instances use the clustering methods to identify the outliers in medical data, in order to determine the abnormal clinical pathway instances. But it has less effect on the identification of abnormal instance groups which have similar medical data and difference with normal ones.In this paper, the concept of clinical pathway set is raised from the feature of medical data in the clinical instances. The clinical instances are divided by the distribution of the medical data from the event logs, and mined and merged into the clinical pathway set which can provide the targeted clinical pathway for the specific clinical instances. According to the feature of the abnormal clinical pathway instances, this paper gives the concept of outlier cluster and the detection algorithm based on the classification idea of clinical pathway set. The abnormal clinical pathway instance identification method proposed by this paper achieves the detection of collective abnormal instances.This research includes the following three aspects:(1) Against the single-path result for the clinical pathway mining, this paper designs the medical clinical pathway framework based on clinical pathway set, which adds the event log preprocess module in the classification of medical data based on the existing clinical pathway mining measures. The framework proposed by this thesis clarifies the difference during the clinical instances with different features, and enhances the relevance of clinical pathway. (2) According to the features of clinical pathway and the relative theory of process modeling, the clinical pathway set model based on process mining is proposed. This model chooses the medical data in the event logs as the classification standard and divides the clinical instances based on the distribution of the medical data. Then, the clinical pathways in group are mined by using process mining technology and the clinical pathway set is merged by these classified clinical pathways. Also, the meta-model for the clinical pathway set is given.(3) Abnormal medical clinical pathway instances identification method is proposed combined with consistency checking and outlier cluster detection based on medical data classification concept of clinical pathway set. The detection method uses the path consistency to check if the difference between clinical pathway instance and treatment processes similar with the clinical data characteristics exists. Outliers and outlier clusters in this data space are checked based on characteristic classification of clinical data through the medical data in the selected event logs to identify clinical pathway instances which exists treatment processes exceptions or medical data and operation exceptions.Finally, a plug-in is designed and developed to check medical clinical pathway set mining and abnormal clinical pathway instances based on ProM framework, which is verified by practical clinical instance data in hospital. The result shows that the clinical pathway set model proposed in this paper with process mining techniques provides an effective solution for clinical pathway mining measure and application in health information system.