Research And Application Of Sequence Pattern Mining In Clinical Path Discovery

Improving the efficiency of hospital management and increasing the transparency of patient medical treatment,standardizing the clinical behavior of medical staff and reducing the waste of medical resources is related to the vital interests of everyone.These are the hot issues that concern to the whole society.Clinical pathway is a means of clinical medical behavior management.It has been used for hospital management by many countries in the world,and they all have achieved good results.Our country starts the pilot and promotion work of the clinical path from 2009 and strives to put all the second-class and third-class hospitals into the clinical path management by the end of 2020.The current clinical pathways are developed by the experts of relevant field of medical based on the experience and expertise.It's time-consuming and laborious.The development of medical information technology makes the hospital accumulate massive clinical data.It promotes the research and application of large data technology in clinical medicine and also provides a new way for the discovery and development of clinical pathway.Data mining technology can make full use of the existing data information and provide reference and guidance for the formulating of clinical path.And this will make the development of clinical path more scientific and rational.In this paper,the problem of clinical path finding is studied based on frequent sequence pattern mining.The main work is listed as follows:(1)A frequent sequence pattern mining algorithm CPM-PC(Clinical Pathways Mining with Prefix Constraint)with prefix constraint is proposed.The clinical path-way is a set of standardized treatment regimens for a disease.It is actually a series of inspection,treatment and care behavior in accordance with the order of time sequence of a sequence.So the problem of clinical path mining is turned into frequent sequence pat-tern mining problem.In addition,the patient's complete course of treatment is always initiated by some specific medical behavior,for example,surgery is usually starts from the anesthesia.Therefore,the paper puts forward the concept of clinical path "prefix set",which is a collection of clinical items at the beginning of a disease treatment.Based on this,a CPM-PC algorithm for sequence pattern mining for clinical path dis-covery is proposed.The algorithm takes full account of the role of the "prefix set" of the clinical path,and can extract the meaningful sequence pattern more efficiently.(2)An evaluation model of clinical path based on attribute is proposed.As the mining of clinical path will produce a large number of candidate sequence patterns,how to choose these models,there is no relevant research.Most of the existed studies on clinical pathways evaluation focus on the analysis of the results of the clinical pathways that have been piloted.This approach cannot be applied to the selection of clinical pathways.Therefore,on the basis of summarizing the clinical path assessment methods at home and abroad,the evaluation indexes are analyzed.Three basic attributes of the clinical path are selected,the length of hospital stay LOS,the total cost of medical care C and the drug proportion P and an attribute-based clinical path assessment model is proposed.The model focuses more on the properties of the clinical path itself.The merits of the clinical path were evaluated by calculating the weighted evaluation values of the attributes.The model evaluates the merits of clinical pathways by calculating the value of attribute-based weighted evaluation what is denoted by ABWE.Finally,the real data set of a hospital in Fuyang City is used on the experiment.The analysis of the experimental results shows that the CPM-PC algorithm has better per-formance than the traditional sequence pattern mining algorithm in mining the clinical path.