Study On Hospitalization Expenses Of Ischemic Heart Disease Based On BP Neural Network Model

BackgroundIschemic heart disease is one of the major diseases that seriously threaten human health.In recent years,studies have shown that the number of inpatients and treatment costs for ischemic heart disease in China continues to grow at a high rate,which places a heavy burden on social public health management,medical insurance funds,patients and their families.The study of hospitalization expenses of ischemic heart disease can help to control the rapid increase of treatment costs of ischemic heart disease,relieve the pressure of social health insurance funds,and promote the rational allocation of medical resources.It helps to solve public health problems related to cardiovascular diseases.At present,medical expenses research mostly uses traditional linear analysis model which has strict application conditions and cannot be used to solve nonlinear correlation problems.The BP neural network model is currently the most widely used artificial neural network model.The model has the ability of self-learning,self-adaptive,fault-tolerant and nonlinear fitting.It is especially good at solving nonlinear correlation problems.It has been widely used in various fields of science and technology and has achieved many outstanding results.ObjectiveExploring the application of BP neural network in the study of hospitalization expenses and providing new ideas for hospitalization expenses research.Finding out the important influencing factors of hospitalization expenses of ischemic heart disease and providing reference for controlling the hospitalization expenses of ischemic heart disease.MethodsIn this study,patients with ischemic heart disease who were discharged from a Three-A level hospital in 2016 were selected as subjects.The demographic information,hospitalization information,medical information and cost information were collected from the first page of inpatient medical records.The BP neural network model was constructed by SPSS23.0 and the variables were screened according to the standardized importance of sensitivity analysis.ResultsA total of 1877 patients with ischemic heart disease were included in the study.The BP neural network modeling results of one hidden layer and two hidden layers 100 times respectively showed that the BP neural network fitting ability of one hidden layer and two hidden layers is not significantly different.The 100-times modeling results of BP neural network of one hidden layer were as follows: The number of hidden layer neurons was 10.The average relative error of training set was 0.349.The average relative error of test set was 0.363.The 100-times modeling results of BP neural network of two hidden layer were as follows: The number of the first hidden layer neurons was 10.The number of the second hidden layer neurons was 8.The average relative error of training set was 0.352.The average relative error of test set was 0.367.The results of variable selection based on standardization importance in the modeling process of BP neural network were in good agreement with the results of variable selection using Forward: LR method in Logistic regression analysis(Kappa=0.643,P=0.008).The one hidden layer BP neural network model was established after the admission season,sex,number of hospitalizations,admission route,admission date,marriage,drug allergy and admission time were excluded by variable selection.The relative error of the model training set was 0.349,the relative error of the test set was 0.343,and the model fitting accuracy has no obvious change compared with that before the elimination of variables.According to the sensitivity analysis,hospital stay,decision to operate,types of ischemic heart disease,age and rescue situation were the important influencing factors of hospitalization expenses of ischemic heart disease,and the importances were 0.397,0.187,0.161,0.086 and 0.058 respectively.ConclusionBP neural network can be used for study on hospitalization expenses;Variable selection can be carried out based on the standardization importance of sensitivity analysis in BP neural network modeling.Hospital stay,decision to operate,types of ischemic heart disease,age and rescue situation were the important influencing factors of hospitalization expenses of ischemic heart disease.