DIP grouping study based on decision tree model: a case study of patients with premature rupture of membranes

Objective:To explore the grouping of diagnosis-intervention packet(DIP)and cost standard of hospitalization expenses in patients with premature rupture of membranes (PROM), so as to provide a basis for optimizing the scheme of the disease category grouping and improving the level of hospital refined management.Methods:The first page information of 1051 medical records of discharged patients with premature rupture of membranes from two secondary hospitals in a district of Shenzhen from 2019 to 2021 was collected, and the influencing factors of hospitalization expenses were analyzed by single factor analysis and multiple linear regression.The chi-square automatic interactive detection (CHAID) decision tree algorithm was used to establish the case-mix scheme of premature rupture of membranes and calculate the expenses of each group.Results:The classification node was roughly grouped by surgical operation.Based on the results of multiple linear regression, the length of hospital stay (t=14.465, P< 0.001) and the payment method (t=-9.166, P< 0.001) were included as the main influencing factors of hospitalization expenses into the sub-group classification node tobuild the DIP group decision tree model, and finally formed a total of 9 DIP subgroups and corresponding cost standards with strong inter-group homogeneity(CV of each group was less than 0.3)and great inter-group heterogeneity (χ²=519.346, P< 0.001).Conclusion:The case combinations after grouping based on the decision tree model and comprehensive consideration of the treatment operation and individual characteristics of patients is in line with the clinical diagnosis and treatment practice, and the grouping is reasonable, which can provide reference for the relevant departments to optimize the detailed grouping method of the disease; DIP subgroups can provide decision-making basis for hospital fund allocation and construction of performance evaluation system.