| Objective The research discusses the application of data mining technology, which is described how to apply in two application examples.One is mining influence factors related in health care costs of hospitalization expenses management to data mining, in order to effectively control the cost of health care.Another is the use of Microsoft time series model to predict the future of the hospital outpatient workload, so that reasonable arrangements for the people, money, resources to provide a scientific basis.This not only from hospital managers angle to provide intelligent analysis method, also from the hospital researchers perspective provides new means and perspective.Materials and Methods This paper uses the Qingdao municipal hospital HIS system in2008to the first half of2011medical record data (including inpatient and outpatient) as a data source.The specific use of hospital patient information include name, gender, date of birth, age, date of admission, discharge date, ICD-10coding, hospital charges, length of hospital stay. Outpatient information include registration department, registration date, registration person-time.It makes the medical record data information to the data warehouse through data extraction, data changes, data cleaning. The specific process is data noise processing, error and empty data modifying by finding the original record, to guarantee the accuracy of the data. For data discretization, age field, days of hospitalization field and cost field belongs to continuous data, that is not conducive to the analysis of the data, then they need discrete process. The age is divided into age section,0-5,6-15,etal. The cost is divided into cost section,5000yuan the following,5000-10000yuan,10000yuan of above, length of hospital stay is divided into length of hospital stay section, the following15days,16-30days,31-45days,etal including9sections. The registration date of outpatient information is discretized in the year, quarter, monthly presented.On the established data warehouse, it processes data section, cutting, drilling, rotational analysis, using OLAP technology. Using of Microsoft decision tree, the first half of2011inpatient data for data mining, in all draw-out28723case records, the selected data were divided into two parts, wherein the training set was70%, the test set was30%, model by decision tree, and then test and compare on the test set, analyze the effect of hospitalization expenses related factors. The outpatient registration data from2008to the first half of2011are screened from the medical record information data warehouse, then using the Microsoft time series mining model prediction in2011July outpatient workload.Results Through the decision tree mining model, it finds correlation intensity size order for the cost of hospitalization:Department>diagnose> length of hospital stay>age> cost categories> treatment outcome. Using The Microsoft time series mining model, prediction in2011July outpatient workload shows Eastern Hospital rheumatology outpatients, Eastern hospital neurology outpatient, Eastern Hospital Outpatient Department of hepatobiliary surgery predictive value is very close to the actual value, other professional clinic practical value is in predicting deviation range, only the Eastern Hospital urological surgery predicted value deviates from the actual value, because of the replacement of the Department Director, bringing more patient.Conclusion the decision tree algorithm and prediction of registration information based on time sequence model, these can be well applied to the hospital management. The departments of hospital charges10000yuan higher proportion have Eastern Hospital cadre health care three, Eastern Hospital Department of Hematology, as hospitalization costs focus section, used to control the cost of health care. Time series mining model can better complete the task of predicting, for outpatient management departments to provide a reliable basis for arrangement of medical resources. |
PDF Full Text Request