Analysis On Hospitalization Expense Of Acute Myocardial Infarction Patients On Data Minging

Objective The aim of this study is to describe the basic composition of hospitalization costs for acute myocardial infarction(AMI),to construct a classification model for hospitalization expenses for patients with AMI using a variety of data mining tools,to analyze the main influencing factors affecting the hospitalization cost of AMI,and to find an effective method to control the hospitalization cost of AMI.Methods The data were obtained from inpatients from the Cardiovascular Hospital of Xiamen University from January 1,2017 to December 31,2017.The patients were diagnosed with AMI and a total of 925 samples were selected after data cleaning.This study provides a descriptive analysis of hospitalization costs and their composition in patients with AMI,and a univariate analysis of factors such as gender,age,whether surgery and medical payment methods.The hospitalization expenses were discretized by K-means clustering through python3.7.1 software.Support vector machine,artificial neural network,decision tree and random forest algorithm were used to fit the model to analyze the influencing factors of hospitalization expenses for patients with AMI.Results Analysis of the composition of hospitalization expenses for AMI showed that among the individual expenses,the composition ratio is:consumables fee(53.82%),treatment fee(19.0%),diagnosis fee(14.75%),drug fee(8.2%),and medical service fee(4.1%).Univariate analysis showed that hospitalization costs were higher in men,patients undergoing surgery,and those with AMI between the ages of 40-60 and 60-80 years.Hospitalization costs increase with increasing hospital stays,and hospitalization costs are higher when patients with AMI are accompanied by complications.The ordinal logistic regression analysis showed that gender,length of hospital stay,and whether surgery or not had a significant effect on hospitalization costs for patients.It can be seen from the OR value that,excluding other factors,compared with women with AMI,the risk of hospitalization for male patients with AMI is more likely to increase by one grade,the latter being 1.93 times of the former;for each additional hospitalization day,the risk of hospitalization costs for patients with AMI increased by 1.23 times;the risk of hospitalization costs for patients with AMI increased by 210.61 times compared with non-surgical patients.In this study,the hospitalization expenses for patients with AMI were divided into three categories.The first type of cluster center was 20,251.95 yuan,and the hospitalization cost range was 1533-412111 yuan.The second type of cluster center was 57,976.39 yuan.and the range is 41477-89321 yuan.The third type of cluster center is 117865.86 yuan,and the hospitalization expenses range is 89576-196373 yuan.Using the discretized hospitalization expenses as the target variable and the univariate analysis results as the predictive factors to construct the AMI hospitalization cost prediction model,the results show that the model with the highest prediction accuracy rate is the support vector machine(99.64%),and then is random forest(99.25%),artificial neural network(93.0%),logistic regression(73.4%)and decision tree(72.33%).Data mining algorithms are better predictive than traditional regression models.Conclusion In this study,consumables accounted for a higher proportion of hospitalization expenses,mainly because percutaneous coronary intervention(PCI)was the main method for the treatment of AMI and its application was increasingly widespread,including high stent costs.Therefore,it is critical to strengthen the management of high-value consumables and reduce the cost of stents.The longer the length of hospital stay,the higher the hospitalization cost.The long-term hospitalization will continue to produce the cost of diagnosis.,examination and medical services.Reasonable control of the hospital stay will help to alleviate the economic burden of the disease.The cost of hospitalization is also related to factors such as comorbidities.When patients have multiple comorbidities,the condition is often more complicated and serious.Strengthening the health education of chronic diseases helps to reduce the economic burden of the disease.In addition,from the perspective of model classification effects.In the five models of logistic regression,support vector machine,artificial neural network,decision tree and random forest,support vector machine and random forest model have higher predictive ability,and data mining algorithm has better prediction effect than traditional regression model.