Predicting The Recurrence Of Hypertensive Intracerebral Hemorrhage: A Risk Scoring Model Derived And Validated Based On Clinical Data

Background : This study aims to develop and validate the risk scoring model to predict the RHCH within one year after the initial HICH,and to facilitate preemptive clinical intervention for the prevention of secondary hemorrhage.Methods:Patient gender,age,blood pressure,Glasgow Coma Scale(GCS),location of cerebral hemorrhage,surgery,past medical history,blood biochemical parameters,and Glasgow Outcome Scale(GOS)were analyzed and evaluated to determine the independent predictors of RHCH within one year after the initial HICH through logistic regression analysis.A risk scoring model was constructed by assigning coefficients to each predictor and validated it in another independent cohort.The accuracy of the model was then assessed by the area under the receiver operating characteristic curve(AUC),and the calibration ability of the model was assessed by the Hosmer-Lemshow test.Result:Of 520 patients in the derivation cohort,38 patients developed RHCH during follow-up within 1 year after discharge.Independent risk factors of RHCH include age > 60 years(OR = 2.46,95% confidence interval CI 1.12-5.44),stage 3hypertension at admission(OR = 4.30,95% CI 2.84-12.65),GCS 9-12 points at admission(OR = 3.16,95% CI 1.28-7.79),GCS 3-8 points at discharge(OR = 9.69,95% CI 1.65-57.01),history of cerebral ischemic stroke(OR = 3.34,95% CI1.18-9.49),history of smoking(OR = 3.52,95% CI 1.53-8.13),history of alcoholism(OR = 2.84,95% CI 1.33-6.08),plasma homocysteine(Hcy)≥10umol/L(OR = 2.42,95% CI 1.35-6.81).The recurrence rates for the low-(0-13points),intermediate-(14-26 points),and high-risk(27-39 points)groups were 1.73%,6.11% and 57.14%(p<0.001),respectively.The corresponding rates in the validation cohort,where 10/107(9.35%)developed RHCH,were 3.45%,7.14%and71.43%(p<0.001).The risk scoring model exhibited good-excellent discrimination inboth derivation and validation cohoets,with AUC of 0.802 versus 0.863.The model also showed good calibration ability(Hosmer-Lemshow p of the two cohorts were0.532 versus 0.724).Conclusion:This model will help identify high-risk groups for RHCH in order to facilitate and improve preemptive clinical intervention.