Design And Development Of Medicare Costs Reduction System

With the aging of the population and development of medical technology,the expenditure of medical insurance rises continuously,the national medical insurance fund faces enormous challenges and the task of reducing medical costs is urgent.In order to solve this problem,there are three points in the following:?1?Using a rule-learning approach to check medical expenses.Rule-learning approach learns a set of rules on a supervised data set and then the learned rules are reviewed by the medical insurance experts as the auditing rules for checking the medical expenses uploaded by medical institutions.The illegal data discovered by machine will be submitted to auditing department and inpatient department as well as birth department in social security agency for further confirmation.Medicare fund of illegal bills will not be reimbursed,which can reduce the medical costs directly.In the experiment,the performance of ID3-based decision tree learning algorithm and the CN2-based sequential covering algorithm are compared on the dataset.The result shows that the CN2-based sequential covering algorithm perform better than ID3-based decision tree learning algorithm,whose precision is 70.04%,recall is 60.36,₁ is 64.84%.Therefor the CN2-based sequential covering algorithm is used to learn a rule set in the system.?2?Using random forest algorithm to predict the treatment result of inpatients.The treatment result of inpatients predicted by classifiers is used to calculate the effective treatment rate of each doctor within the scope of easy-cure diseases,to find out doctors whose effective treatment rate is abnormally low and submit information of these doctors to auditing department in social security agency for further processing.Six classification models including k-NN,naive bayes,logistic regression,decision tree,SVM and random forest are used to predict the treatment outcome of five diseases including thyroid nodules,left renal cyst,cervical spondylopathy,fibroid and benign prostatic hyperplasia with calculi.The experimental result shows that average precision of random forest is 74.93%,ranking first,average recall is 72.44%,ranking second,average₁ is 73.35%,ranking first.Therefore the random forest model is used to predict the treatment outcome of inpatients.?3?Developing a medical costs reduction system.Two approaches introduced above are used to check and count the medical expenses data uploaded by the medical institution.This system reduces the workload of social security agency's auditors,and provides various statistical reports for the departments such as the inpatient department,the industrial injury department,and the birth department.The medical costs reduction system proposed in this paper runs stably in Yantai City and Weihai City,Shandong Province for two years,significant effect of reducing medical costs has been achieved,is praised by the leader of social security agency.