Prediction Of Patient Readmission Based On Medical Service Supply Chain

The mismatch between supply and demand of medical service resources is a prominent problem in China’s medical field,which indirectly makes it difficult for the supply of medical service to adapt to the change of demand.One of the specific manifestations is that the quantity,quality and utilization rate of medical service resources need to be improved.With the continuous improvement of medical information system and the rapid development of big data and artificial intelligence,medical informatization has become the development trend of medical service industry.It is feasible to use medical information to promote the supply and demand matching of medical service resources.Medical service supply chain is a typical pull service supply chain.In the process of medical service,patients are the service demanders and end consumers,and the medical service supply chain is driven by the needs of patients,so it becomes the source and foundation of the medical service supply chain.Therefore,identifying the needs of patients is conducive to promoting the effective operation of the medical service supply chain.Because the demand for medical service is changeable and uncertain,and the medical service supply chain is relatively weak in flexibility and flexibility compared with the product supply chain,it is extremely important to reduce the uncertainty of demand with certain means.Based on this,this study focuses on the patient demand side of the medical service supply chain,and predicts the risk of readmission by recording the electronic medical records and feedback information of patients’ diagnosis and treatment data,so as to assist hospital doctors in management and decision-making,so as to optimize medical services.In this study,more than 4500 inpatient records of COPD patients were obtained from a top three hospital in Guangzhou.After data preprocessing,univariate analysis and multivariate logistic regression analysis,12 statistically significant influencing factors were obtained and used as predictive variables.Logistic regression,support vector machine,random forest and BP neural network were used to establish the prediction model of readmission risk.The prediction accuracy of four different classification models was compared and analyzed.The results showed that random forest algorithm had the highest prediction accuracy,and the prediction accuracy of test samples reached 82.54%.The prediction accuracy of BP neural network was slightly higher than that of support vector machine.Logistic regression model had the worst prediction effect on the test sample set,and the prediction accuracy was only69.78%.At the same time,using random forest to rank the importance of characteristics,we found that leukocyte,eosinophil ratio,duration of disease,oxygen saturation,Charlson’s comorbidity index and length of hospital stay were the six important predictors of readmission.Based on the prediction of patients’ readmission in the medical service supply chain,this paper identifies the influencing factors of patients’ readmission through the risk prediction of readmission,reduces the uncertainty in the process of medical service,assists hospital doctors in demand management,service management and corresponding intervention diagnosis and treatment,reduces the consumption of medical service resources,and alleviates the mismatch between supply and demand of medical service resources To improve the benefit of medical service supply chain.