Design And Implementation Of Clinical Path Analysis System Based On Medical Big Data

With the arrival of the deep ageing society,and beneath the "universal health insurance" environment,it will become the new normal of the medical insurance fund what is less income and more expenses.Currently,for the medical insurance fund,there is not much room to increase revenue.Cost control based on the clinical pathway has become a necessary path for medical insurance system reform.This thesis is mainly based on the Spark on Yarn distributed platform,and incorporates machine learning clustering algorithms,open source middleware Sqoop,big data analysis technology,regular expressions,mathematical statistics,probability distribution and JavaEE framework and other technologies.Combined with the core content of the national clinical path,the key technologies,involved in empirical clinical pathways analysis,were studied.A set of clinical path analysis system solutions which are based on medical data migration,medical data cleaning,medical data analysis and application services was proposed.The operations,mainly including data cleaning and data analysis,are performed on medical data to obtain clinical pathways that conform to empirical rules,by using techniques such as big data analysis and machine learning algorithms and other technologies.The clinical pathways is referred to as the empirical clinical pathways.This thesis firstly analyzes the current status of the development of clinical pathways,and proposes the goals and contents of itself.Secondly,it studies the theories and technologies involved in the research.Thirdly,it defines the functional and non-functional requirements of the system.Fourthly,the system's overall architecture,technical route,function modules,and database are designed.Fifthly,the key modules of the system are designed and implemented in detail;a medical data migration's method based on Sqoop is proposed;a medical data cleaning's and analysis' s scheme is designed and implemented based on the Spark on Yarn platform;at the same time,the empirical clinical pathways data service interaction interface is developed based on the SpringMVC framework.The main work of this thesis is as follows.1.Combining the actual needs of medical data migration,based on the open source middleware Sqoop,which is expanded the data types of the data sources it supports,and modify its ORM(Object Relational Mapping)module to support data mapping,standardized processing,and character encoding conversion.2.The discharge diagnosis text of medical data was studied,and its feature was summarized.The method of extracting the ICD-10 disease's code from the discharge diagnosis text was designed and implemented.Analyzing the feature of refund items in medical data,a method for processing refund items is designed and implemented.3.Based on the parallel computing feature of Spark,the parallel processing and design of medical data cleaning and analysis are designed and implemented,which improves the overall performance of the system.4.Based on the one-dimensional index sample of the analysis unit,its distribution conforms to the Gaussian distribution.Combined with mathematical statistics and probability distribution theory,a center-outward expansion algorithm based on frequency histogram is proposed.Based on this algorithm,the empirical range analysis of the one-dimensional index of the clinical pathways was completed.5.The empirical range,which belongs to relevant indicators of clinical pathways,is analyzed by implementing the machine learning clustering algorithm DBSCAN.