Distribution And Prediction Of Patients’ Length Of Stay

The length of stay is one of factors for the hospital to allocate ward beds reasonably,and affects the hospital’s operating speed,medical level and work quality.Hospital managers have been plagued by the long hospitalization period and low bed turnover rates.In order to improve the utilization of medical resources and relieve the economic pressure of patients in hospitalization,it is necessary to deeply study the length of stay.Based on the hospitalization data of rectal cancer patients in the oncology department of a domestic third-class hospital for more than five years,this paper studies the distribution and prediction of patients’ length of stay in the hospital,and provides some references for hospital administrators in the scheduling of medical resources.In reality,the distribution of patients’ length of stay is asymmetric.The average length of stay cannot reflect the characteristics of the distribution of actual length of stay.Decisions made by hospital administrators based on the average length of hospital stay may lead to unreasonable bed allocation,resulting in unnecessary losses.Therefore,it is important to find a suitable distribution model to fit the length of stay.This paper uses GMM to fit the empirical probability distribution of length of stay.The parameters of the model were estimated by EM algorithm,and statistics such as AIC are used as a measure of the fitting effect.The empirical analysis results show that GMM is better than the lognormal distribution,gamma distribution and Weibull distribution in fitting the distribution of length of stay,which provides a new idea for the study of the distribution of length of stay.In addition,the study of distribution fitting only uses the length of stay data to study its distribution,and it cannot reflect what factors affect the length of stay.Therefore,this paper also applies the current popular machine learning technology to the study of length of stay.XGBoost algorithm is used to establish a prediction model of length of stay.Several characteristic variables affecting length of stay are analyzed,and then predict whether the length of stay will be extended based on this.The empirical analysis compares this model with the traditional CART decision tree model and logistic regression model.The results prove that the model based on the XGBoost algorithm has higher accuracy and reliability,which provides a reference for the prediction of length of stay.In terms of the research on the distribution of length of stay,this paper proposes a method of fitting the distribution of length of stay with GMM,which effectively solves the problem of poor fitting effect of traditional single distribution in the past.In terms of the prediction of the length of stay,this paper introduces the XGBoost algorithm,and establishes a prediction model based on the XGBoost algorithm,which shows the characteristic variables that affect the length of stay,and the predicted effect is also improved compared with the traditional prediction models.