The Research And Application Of Data Mining Technology In The Hospital Medical Insurance Cost Analysis

With the traditional database technology based on relational structure widely used in the fields of information age, all kinds of complex database systems have been build. It greatly enhance the working efficiency by changing the manual operation to computers. Because the database and all kinds of information system have been used for a long period of time, there are more and more business data in the database, and it is still increasing with the deeply use of business applications. With the increased competitiveness of society, people can not satisfy with simply using the function of data processing, they want to change the direction to deep analysis and application. However they can not just use the database function of analysis and ability of business system to discover the rules and knowledge in the huge amounts of data without the powerful data mining and analysis tools. In order to meet the growing need of deep analysis and application, the technology of data mining and data warehouse come into being. They can discover the knowledge and experience in deep level and get the helpful information about management. Now the technology is widely used in the financial services industry (such as banking, insurance), retail trade (supermarkets), and other commercial areas of the telecommunications industry, but the application in the medical field is still in the initial stage. In the article we made a bold attempt to research in the field, using a technology of data mining to get the meaningful rules set in the hospital medical insurance cost analysis.Many domestic hospitals have set up their own hospital information system(HIS), and some large hospitals even have set up a integrated information system including HIS,PACS and LIS. There are not only a large number of patients basic information but also cost information,clinical physiological information and the image picture information in the business database. While medicare and other new medical way have emerged to occupy an important position, it becomes a very important task how to discover the knowledge and experience in the medical treatment,discipline construction,decision-making management etc. It becomes an inevitable trend that we use data mining and data warehouse technology to analyse the historical data, which should be a new topic faced by the construction of hospital information.With the deepening reform of medical treatment, strongly popularizing medical insurance coverage, and striving to improve the level of society medical insurance, the medicare patients in hospitals in the proportion of patients is higher and the medicare income in the proportion of whole hospital income is the same. Now the problems faced by the hospital medicare management are as follows: How to reasonably control the rising medical costs of medicare patients and give the them perfect medical service at the same time; how to establish the scientific and rational index of medicare and allocate the resources returned by medicare insurance; how to arouse the enthusiasm in hospital department to actively implementing insurance department policy.In our research we extracted all types of information about medicare patients over the years, established the data warehouse based on the theme of cost analysis, and introduced a series of data preparation process in the data of business such as data extraction, data cleaning, data transfer, and data loading. We established the star model based on the theme of cost analysis and set up the fact table which associated with cost. We also set up many kinds of analysis dimensions such as time,section,doctor,diagnosis,address,sex,age,identity.The data source in our research are from different structure system: HIS and PJ3. So we have some problems that maybe we will describe the same data with different expression or we may find the data missing and data duplication. If we plan to finish the job of data cleaning by manual filter it can not be done. The data preparation is one of the most important job in the process of establishing the data mining system, and it will spend you much time and energy. More important, data cleaning is the key in the data preparation. It is responsible for putting the data extracted from business database 'clean' to the data warehouse, Whether it will be done successfully directly affect the quality and efficiency of data mining, and of course affect the result.In response to this research, we designed a automatic way of data cleaning. The main job is to clean the data from different structure database and the data with the double meaning or nonstandard. By the record of the track we draw a conclusion that the auto data cleaning has solved the following problems:1. Filtering the medicare record automatically which lack of patient number in hospital.2. Filtering the medicare record automatically which do not match between patient number and patient name.3. Filtering the medicare record automatically which do not match between patient ID and patient number.Filtering the medicare record automatically which lack of department information and transfer information.The system can filter the wrong record, speed up the time of data cleansing and ensure the quality of the data mining by the method of filtering automatically.By analyzing the historical data about medicare patient the subject use Apriori algorithm in association rules to mine the relations between medicare cost and associated factors such as identity,department,doctor,diagnosis,days in hospital,region. We draw a conclusion about the rule set in medical insurance cost analysis, and we can see the nearest association with medicare cost are the combination between days in hospital and specific discharge department by using the rule set. It is very clearly that 5 rules anterior to the rule set are the most important factors which will affect the medicare cost. So we should give more concern about the medicare patients who have the factors. Only in this way we can avoid the situation that the patient with sufficient medicare spend much more money on medicare quota. At the same time we can mobilize the enthusiasm of sections, allocate the medical resources rationally and protect the medicare compensatory income.It makes clear that we can use the data mining and data warehouse technology to find the rule set which associated closely with medicare cost management through the research project, provide a useful reference and guidance to the medicare cost management, assist in drawing up a dynamic medicare quota, speed up the rate of analysis feedback and trace back to the analysis data timely. Comparing with the traditional technology the data mining technology has the advantages of flexible, rapid, and strong analysis ability. It is very wise to apply the data mining technology to medicare management and it will meet the change and new challenges of medicare management.It has made beneficial exploration in the application that domestic medical institutions use the data mining technology to analyze the medicare data in this article. We provide the relevant technical, process and data reference for the new medical management decision-way. Furthermore, we have made useful attempts at application of data mining in the medical information construction.