Research Of Expert Process And Artificial Intelligence Model Of Anemia Management In Hemodialysis Patients

Objective:Anemia is one of the most common complications in maintenance hemodialysis(MHD)patients,and the compliance rate of hemoglobin(Hb)in MHD patients worldwide is low.This study investigated the Hb compliance rate and iron metabolism indicators in MHD patients,constructed prediction model of Hb by machine learning,combined with renal anemia management process to achieve the purpose of standardized,individualized and accurate management of renal anemia in MHD patients.Methods:Forty-one dialysis centers were included.Laboratory data were collected to analysis Hb compliance and others.Then by referring to domestic and foreign guidelines,consensus,standard operating procedures,expert opinions to formulate the process.Finally,the medical records of MHD patients from 2021-01-01 to 2023-01-01 in Sichuan Provincial People’s Hospital were included.Randomly divided the data into training set(80%)and test set(20%),10 machine learning types were used to build the prediction model.The optimal model performance was assessed by internal validation,combined with the process to guide the anemia management.Results:7190 patients from 41 dialysis centers were enrolled,the Hb compliance rate was 34.1%(95%CI 33.0%-35.2%),and the compliance rate was different in different regions(p<0.0001);women’s compliance rate was lower than men,the difference was statistically significant(p=0.002);women had an increased risk of Hb non-conformity compared with men(OR=1.167,95%CI 1.057-1.290),and 18.9%(16.9%-20.9%)had absolute iron deficiency.Then,495 patients’medical records from Sichuan Provincial People’s Hospital were enrolled,10 machine learning types were used to construct the prediction model,in which Random Forest showed better prediction performance compared with other models,with RMSE of 9.41 and a coefficient of determination R~2of 0.65 in training set.Conclusion:The Hb compliance rate of MHD patients in Sichuan is low and there are regional differences.The females’compliance is lower than males,about 1/4 of MHD patients have absolute iron deficiency.The Random Forest prediction model can be used to predict Hb concentration,and in combination with the renal anemia management process,which can assist clinicians to make clinical decisions to achieve individualized and precise management of renal anemia in MHD patients.