BP neural network based analysis of factors influencing hospitalization expenses in TCM hospitals / 中华医院管理杂志

ABSTRACT

Objective To control medical expenses in a rational and effective way by analyzing the main factors influencing hospitalization expenses in TCM hospitals. Methods Based on ICD 10 codes, the top 50 disease entities causing hospitalization, as had been accumulated over the years, were extracted from the hospitalization record data warehouses of the national TCM supervision networks using nationally representative medical record databases, the main factors influencing hospitalization expenses were identified using multi layer BP neural network, and further analyses were made by disease entities and hospital grades. Results When disease entity mix analyses were conducted, it was found that the variables most influencing hospitalization expenses/days were hospital code and disease entity while variables specific to TCM had little influence on the expenses. The results were also true of analyses made by hospital grades and disease entities. Conclusion Under the current circumstances in China, the influence on hospitalization expenses resulting from the difference between hospitals is far greater than any other factor; it is even greater than that resulting from the difference between disease entities. It is therefore imperative to pay attention to the extensiveness and coverage of data sources when conducting research on the control of medical expenses.