Construction of risk prediction model for poor prognosis in uremic patients with intracerebral hemorrhage

WANG Chun; LIU Wei; YIN Jianhua; LI Kui; SONG Yan; LI Fan

doi:10.3969/j.issn.1671-315x.2023.07.002

Abstract：Objective To build a risk prediction model for poor prognosis of uremic patients with intracerebral hemorrhage based on SMOTE algorithm and Logistic analysis.Methods Totally,239 uremic patients with intracerebral hemorrhage admitted to a hospital from July 2019 to January 2021 were selected as the study subjects,and were divided into poor prognosis group and good prognosis group according to the prognosis of patients.The clinical data of the two groups of patients were collected,and the risk factors of poor prognosis were screened by univariate and binary Logistic regression analysis,and a Logistic regression model(P1)was established.At the same time,the data set was im-proved based on SMOTE algorithm,and a prediction model(P2)was constructed,and the prediction efficiency of those models was compared.Results Thirty-four of 239 uremic patients with intracerebral hemorrhage had poor prognosis,with an incidence of 14.23%.Age≥60 years old,mean arterial pressure≥120 mm Hg,hemoglobin＜90 g/L,albumin＜35 g/L,excessive intracerebral hemorrhage and hemorrhage breaking into the ventricle were independent risk factors for poor prognosis of uremic patients with intracerebral hemorrhage(P＜0.01 or P＜0.05).ROC curve showed that the area under ROC curve of P2 model was 0.904(95%CI:0.884,0.924),which was significantly higher than that of P1 model 0.806(95%CI:0.750,0.854),and the specificity of P2 model was significantly higher than that of P1 model(P＜0.05).Conclusion Age,mean ar-terial pressure,hemoglobin,albumin,excessive intracerebral hemorrhage and hemorrhage breaking into the ventricle are risk factors for poor progno-sis in patients with uremia complicated by intracerebral hemorrhage.The prediction model based on SMOTE algorithm has high prediction efficiency.