Research On Algorithm Of Age Segmentation And Individual Feature Selection In Diagnosis-Related Groups

Diagnosis-related groups(DRGs)is a case mix method.As an important medical management tool,it can be used not only for payment management,but also for budget management and quality management,as well as for coordinating the relationship among health insurance bureagues,hospitals and patients.DRGs is based on the data on the first page of the medical record and based on the principle of similar clinical processes and similar resource consumption,and divide the cases into different DRGs.In the case of many versions of DRGs groupers,different groupers need to be distinguished by grouping details.Individual characteristics are one of the details in the grouping process of DRGs.Since individual characteristics will be affected by factors such as region,disease spectrum,economic development level,etc.,DRGs should be grouped with dynamic processing methods for individual characteristics.It is not appropriate to homogenize individual characteristics and adopt static processing methods.Therefore,the DRGs algorithm that can adjust the partitioning process according to the data is very important for the research of DRGs.First,in view of the problem of homogeneity when using age to divide DRGs,an age segmentation algorithm based on least squares polynomial fitting and an age segmentation algorithm for DRGs based on ordered sample clustering are proposed.Among them,the age segmentation algorithm based on least squares polynomial fitting uses the least squares method to fit the relationship between age and hospitalization expenses,and it is expressed in the form of polynomial.Then the extreme points of the fitting polynomial are calculated and the age is segmented by the extreme points.The DRGs age segmentation algorithm based on ordered sample clustering uses the segmentation point with the smallest difference in hospitalization costs within the same age group and the largest difference in hospitalization costs between different age groups to segment the age.Then,aiming at the problem of feature selection when using individual characteristics for division in DRGs,a feature selection algorithm based on Pearson’s correlation coefficient and random forest is proposed.Pearson Correlation Coefficient is used to screen the influencing factors related to the target variable.Then,within the range of relevant influencing factors,random forest is used to evaluate the feature importance of attributes,and the attributes for DRGs grouping are selected according to the degree of importance.Finally,taking the first page data of medical records from a secondary hospital in a certain area as an example,from the rationality of the algorithm and the impact on the number of unsatisfied in-group consistency groups,the age segmentation algorithm based on least squares polynomial fitting and the DRGs age segmentation algorithm for ordered sample clustering is analyzed,and verify the accuracy of the feature selection algorithm based on Pearson correlation coefficient and random forest.