Research On Time Window Prediction And Risk Scoring Method For Traumatic Hemorrhagic Shock

Smart healthcare is a combination of computer,big data,and clinical medicine.Based on the huge amount of data,it makes more precise and scientific decisions on patient treatment,triage,and prognosis,so as to reduce patient mortality and improve the quality of life.At present,China is vigorously promoting the "big health" of the whole people,data analysis,big data application and other methods have been widely used in various branches of medicine,resulting in great economic and social benefits.Traumatic hemolytic shock is one of the leading causes of death in trauma patients and is the largest preventable factor.Trauma patients often have complex and rapid development characteristics,and hemorrhagic shock in the early stages is difficult to identify.The mortality rate in patients once onset is extremely high.Therefore,it is of great significance to explore the key risk index that affects hemorrhagic shock,to predict and warn the patient’s outcome in advance,and to indicate the patient’s risk level,and to realize the "early identification,early diagnosis and early treatment" of patients with traumatic hemorrhagic shock.Based on the first aid database of the Chinese People’s Liberation Army General Hospital and the MIMIC III database of the Massachusetts Institute of Technology,this paper,under the guidance of a professional clinician,designs the exclusion criteria of the research experiment and extracts the medical index data of the relevant patients.Data was preprocessed by Python,and 13 key risk indicators and their feature weights were selected based on the combination of e Xtreme Gradient Boosting(XGBoost)algorithm and sequential forward search strategy,among which 5 vital signs(systolic blood pressure,diastolic blood pressure,heart rate,respiratory rate,temperature)accounted for the largest feature weight.Then the data is processed into a regular time series by the way of hourly slice,and reasonable time window parameters are set.Combined with the machine learning algorithm of decision tree,random forest,decision tree-based adaptive reinforcement(Adaboost)algorithm and e Xtreme Gradient Boosting(XGBoost),we build a time window prediction model.The key indicator data set and the full indicator data set brought into the first aid database respectively provide early prediction and early warning for traumatic hemorrhagic shock and external verification through the MIMIC III database.The results show that compared with the whole index set,the prediction effect of the key index set is not much different,and the combination of individual time parameters and algorithms is even better.In addition,the required data dimension and operation time are also greatly shortened.The closer the time is to the outcome,the more accurate the prediction effect will be.The prediction effect of XGBoost algorithm is better than the other three.At the same time,although the effect of external validation is not as good as internal validation,it can still prove that the generalization ability of time window prediction model is better.Finally,on the basis of above research,comprehensive consideration of the economic and time benefits of indicator collection and clinical application,select 5 vital indicators in key risk indicators,use the multi-factor Logistic regression method,combine with clinicians’ index classification consensus,and construct a simple trauma hemorrhagic shock risk scoring tool.Then compared with the traditional risk scoring method,and finally through the MIMIC III database for external verification.The results show that the data-driven risk scoring tool for traumatic hemorrhagic shock can effectively predict the outcome of patients with traumatic hemorrhagic shock and indicate the degree of risk,while the method can be adapted to the MIMIC III database,has a certain generalization ability,and has great significance in clinical application.