A Study On The In-hospital Factors Associated With Length Of Stay Among Trauma Patients Based On The Bayesian Network Model

Objective: In China,trauma is the fifth leading cause of death for the population and the first cause of death for people under 45 years of age,posing a heavy disease and socioeconomic burden.Length of stay(LOS)is often used to measure the outcome of trauma patients and is directly linked to healthcare costs,bed turnover and social benefits.The purpose of this study was to explore the in-hospital factors associated with length of stay of trauma patients and the relationships among them;to quantify the probability of long LOS under different conditions and to assess the key factors,so as to provide a scientific basis for research on interventions to improve the effectiveness of trauma emergency care and reduce the disease burden and socio-economic burden caused by trauma.Methods: The trauma electronic medical records on the Yidu Cloud medical big data intelligence platform of Chongqing Medical University was collected,and the final effective sample size was 1202 cases.Descriptive statistical analysis was conducted on the demographic characteristics,trauma-related characteristics and care-related factors of trauma patients;the in-hospital associated factors of the long LOS of trauma patients were explored by using chi-square test and multivariate logistic regression model;Bayesian network model was constructed to analyze the causal relationship among the factors and to evaluate the key factor by probabilistic reasoning.Results: A total of 1202 trauma cases were included,with a male predominance(68.64%).The main cause of trauma was car accidents(32.86%);the most common injury site was head and neck(38.02%);the majority(95.34%)of the trauma mechanism was blunt;nearly one-third(29.70%)of the cases were multiple injuries.The median stay time in the emergency department was 20.68 minutes,and 48.17% of patients with the time over 30 minutes;the median time from admission to operation was 6.23 hours,and there were 430 patients(52.12%)with the time over 24 hours;the median LOS for trauma patients was 14 days,and there were 360 patients(29.95%)with a long LOS.The logistic regression model showed that upper age(15-45: OR=6.079,46-64: OR=9.190,65-: OR=6.338),blunt injury(OR=3.986),serious injury(OR=2.244),transfer(OR=5.078),consultation(OR=1.817)and operation(OR=2.593)were risk factors for long LOS(P<0.05).The Bayesian network model consisted of 11 nodes and 13 directed edges.Advanced age,serious injury,transfer,operation and admission to operation time over 24 hours were all direct in-hospital influencing factors for long LOS,and all of them led to an increased probability of long LOS.Men,blunt injury,multiple injury,imaging examination and consultations are indirect in-hospital influences on long LOS and will ultimately lead to an increased probability of long LOS by influencing other factors.Controlling for trauma patients with constant injury factors,transfer,consultation,and admission to operation time over 24 hours will significantly increase the probability of long LOS.Conclusions: The application of the Bayesian network model for the study of factors influencing the long LOS of trauma patients can visualize the relationships among the factors and predict effectively the probability of long LOS under different conditions,which is of great significance to shorten the LOS of trauma patients and reduce the burden of trauma disease,and also shows that the Bayesian network model has good application value in this field.Transfer,consultation and admission to operation time over than 24 hours are the key factors leading to long LOS and have important interventional value.Trauma teams and trauma centers(including multidisciplinary medical staff and emergency equipment)can be established,which can reduce the occurrence of consultation and transfer,shorten the admission to operation time,and ultimately reduce the probability of long LOS.