Font Size: a A A

Study On Early Warning Model Of Death In Fever With Thrombocytopenia Syndrome Based On Machine Learning

Posted on:2024-06-27Degree:MasterType:Thesis
Country:ChinaCandidate:L Z ZhangFull Text:PDF
GTID:2544307082470114Subject:Internal Medicine
Abstract/Summary:PDF Full Text Request
Objective In this study,by studying the first laboratory test data of patients with fever with thrombocytopenia syndrome after admission,the risk of death in SFTS patients was established for early prediction.Methods From April 2019 to November 2021,the research objects were selected as patients hospitalized in the Department of Infectious Diseases at Chaohu Hospital Affiliated to Anhui Medical University,who were suffering from both fever and thrombocytopenia.Standard,123 patients were included.The patients were split into two categories based on their outcome: the survival group and the death group.Within 72 hours of being admitted,the initial laboratory examination data,including blood routine,blood biochemistry,and coagulation function,was gathered.Statistical analysis was used to study the high-risk influencing factors of patients in the death group.The prediction model of SFTS death patients was established by Logistic regression analysis and random forest algorithm,the receiver operating characteristic curve was drawn,and the area under the curve was calculated.The index with high model contribution.Compare the AUC values of the two models to select the model with better performance.Results Through statistical analysis,we can get age(P value = 0.000),Lactate Dehydrogenase(LDH)(P value = 0.044),Creatine Kinase(CK)(P value =0.029),Creatine Kinase MB(CK-MB)(P value = 0.043),Creatinine(CREA)(P value =0.0033),Activated Partial Thromboplastin Time(APTT)(P value = 0.015),D-D(P value= 0.001)and other indicators increase indicates that There was a statistically significant increase in the risk of death among patients.Through Logistic regression analysis,it can be concluded that APTT,D-D,CREA,and CK-MB are all above the reference line,and have predictive value for the death of patients with this disease.AUC(APTT)is0.686,AUC(D-D)is 0.684,and AUC(CREA)value was 0.678,and AUC(CK-MB)value was 0.577.A joint prediction model was established through four indicators,and its AUC value was 0.703,all of which had predictive value,and the prediction performance of the comprehensive analysis model was better than that of single factor prediction.The death prediction model of SFTS patients established by random forest model has an AUC value of 0.82,which has predictive value,and its predictive performance is better than that of the model established by Logistic regression analysis.Further relative importance analysis and display shows that the contribution of D-D,CREA,APTT,and CK-MB to the random forest model ranks high,and the contribution is relatively close.Conclusion Analyzing the results of statistical analysis,S increases with age,LDH,CK,CK-MB,CREA,APTT,D-D and other indicators increase,and the risk of death in SFTS patients increases,and the results are statistically significant.Logistic regression analysis suggested that CK-MB,CREA,APTT,and D-D had predictive value for SFTS death patients,and the comprehensive predictive value of the four indexes was higher than that of single index.The prediction model established by random forest algorithm has higher prediction accuracy than Logistic regression analysis,and D-dimer has the highest contribution to the prediction model,followed by CREA,APTT,and CK-MB.
Keywords/Search Tags:Severe Fever with Thrombocytopenia Syndrome, Mortality Prediction Model, Random Forest, Machine Learning
PDF Full Text Request
Related items