Study On Length Of Stay And Factors Influencing Hospital Costs Based On Multilevel Linear Models

ObjectiveThe present study aims to analyze the basic characteristics of length of stay and hospital costs of patients from a sample of hospitals as well as the cost structure;to investigate the factors influencing hospital costs of inpatients from a sample of hospitals in various cities and cross-level interactions between the level-1 and level-2 explanatory variables;to provide scientific evidence for shortening of the average length of stay,improvement of hospital efficiency and control of hospital costs based on the effect of factors influencing hospital costs.Data and MethodsMultistage random sampling was used to seclect 14 grade-A tertiary hospitals from 7 cities in the eastern,central and western regions of a province in this study.200 medical records of inpatient with type 2 diabetes,pneumonia,cataract,inguinal hernia,cesarean section,vocal cord polyp and lumbar disc herniation were randomly selected from each sample hospital for the year 2014-2017,and a total of 12843 medical records were finally included.Summary the theory and methodology of multilevel models and their application to health services via literature analysis,together with the research status and progress of factors influencing hospital costs;conduct descriptive analysis of basic characteristics of length of stay and hospital costs of inpatients and the cost structure;apply univariate analysis of factors influencing hospital costs;use multilevel models to explore factors influencing hospital costs and identify whether there are cross-level interactions between different levels of influence factors.ResultsThe median of length of stay(ALOS)of inpatients with type 2 diabetes,pneumonia,cataract,inguinal hernia,cesarean section,vocal cord polyp and lumbar disc herniation were 9 days,9 days,4 days,5 days,6 days,6 days and 12 days,and the median of hospital costs were 6832.15 yuan,6780.26 yuan,4643.02 yuan,6353.9 yuan,8425.10 yaun,6562.40 yuan and 15690.16 yuan,respectively.Drug fees occupy the largest part of hospital costs of type 2 diabetes and pneumonia(37.02%and 47.07%,respectively),followed by diagnostic fees(34.99%and 27.69%,respectively).Treatment fees have the largest proportion of hospital costs of cataract,inguinal hernia,cesarean section and vocal cord polyp(47.82%,41.90%,43.79%and 48.14%),while the principal hospital costs of lumbar disc herniation are consumable material fees(40.00%).Intraclass correlation coefficients(ICC)of hospital costs for patients with type 2 diabetes,pneumonia,cataract,inguinal hernia,cesarean section,vocal cord polyp and lumbar disc herniation are 0.0417,0.1524,0.3847,0.1837,0.5030,0.3204 and 0.1391,respectively,which are statistically significant(P<0.1).In the two-level models for hospital costs of type 2 diabetes,the slope of variable medical insurance is specified as random,and cross-level interactions are identified between GDP per capita and medical insurance(P<0.05);in the model for hospital costs of pneumonia,all level-1 slope are fixed with level-2 variable not statistically significant(P<0.05);in the two-level models for hospital costs of cataract,the slope of variable length of stay,number of beds and medical insurance are specified as random,and cross-level interactions are identified between GDP per capita and length of stay,number of beds,presence or absence of surgery,medical insurance(P<0.05);in the two-level models for hospital costs of inguinal hernia,the slope of variable length of stay,number of beds,presence or absence of surgery and medical insurance are specified as random,and cross-level interactions are identified between GDP per capita and number of beds,medical insurance(P<0.01);in the two-level models for hospital costs of cesarean section,the slope of variable length of stay,number of beds,presence or absence of surgery and medical insurance are specified as random,and cross-level interactions are identified between GDP per capita and length of stay,number of beds,presence or absence of surgery and medical insurance(P<0.05);in the two-level models for hospital costs of vocal cord polyp,the slope of variable length of stay and presence or absence of surgery are specified as random without cross-level interactions(P<0.01);in the two-level models for hospital costs of lumbar disc herniation,the slope of variable length of stay,presence or absence of surgery and medical insurance are specified as random,and cross-level interactions are identified between GDP per capita and presence or absence of surgery,medical insurance(P<0.05).Conclusions? This study represents significant differences in length of stay and hospital costs for patients with several pecific diseases from sample hospitals in different cities,as well as the roughly same cost structure.? The hospital cost data has a hierarchical aggregation,which is available for multilevel models.? Economic development,length of stay,number of beds,presence or absence of surgery and medical insurance are important factors influencing hospital costs,while the effects of length of stay,number of beds,presence or absence of surgery,and medical insurance on hospital costs are influenced by economic development.