脑卒中住院患者跌倒影响因素分析及Nomogram预测模型构建OA

Objective To investigate the factors influencing falls in hospitalized stroke patients and to construct a Nomogram prediction model based on these risk factors,in order to provide a reference for preventing falls in this population.Methods A total of 120 hospitalized patients with stroke admitted to Nanjing Hospital of Chinese Medicine,from September 2021 to September 2024 were enrolled.Patients were divided into a fall group(n=39)and a non-fall group(n=81)according to whether a fall occurred during hospitalization.Clinical data of the two groups were compared,and logistic regression models were used to analyze the factors influencing falls in hospitalized stroke patients.A Nomogram prediction model was con-structed based on these influencing factors.The predictive performance of the model was evaluated using the receiver operat-ing characteristic(ROC)curve.The Bootstrap method,calibration curve,Kolmogorov-Smirnov test,and decision curve a-nalysis(DCA)were adopted to assess the model's discriminative ability,calibration,goodness-of-fit,and clinical benefit.Results Among the 120 patients,39 experienced a fall,with an incidence rate of 32.50%.Multivariate logistic regression analysis showed that a higher Activities of Daily Living(ADL)score(OR=0.925,95%CI:0.874-0.979)was a protec-tive factor against falls in hospitalized stroke patients.In contrast,a higher National Institutes of Health Stroke Scale(NIH-SS)score(OR=1.117,95%CI:1.045-1.194),hemiplegia(OR=1.575,95%CI:1.315-1.886),a history of falls(OR=1.467,95%CI:1.049-2.051),and classification via the Johns Hopkins Fall Risk Assessment Tool(JHFRAT)as high fall risk(OR=2.158,95%CI:1.345-3.460)were independent risk factors influencing falls in these patients(P<0.05).ROC curve analysis results indicated that the area under the curve(AUC)of the Nomogram prediction model for predicting fall risk in hospitalized stroke patients was 0.924(95%CI:0.901-0.953),with a sensitivity of 91.96%and specificity of 89.75%.After 800 bootstrap resamples,the discriminative ability of the model decreased by only 0.019,and the internally validated concordance index(C-index)was 0.887.The calibration curve demonstrated good agreement be-tween the fall risk predicted by the Nomogram model and the actual observed incidence,with a Brier score of 0.171.The Kolmogorov-Smirnov goodness-of-fit test showed that the model was well-calibrated(χ2=3.818,R2=0.921,P>0.05).DCA results demonstrated that within the threshold probability range of 8%to 91%,the net benefit of the predictive model was higher than the extreme scenarios of intervening in all patients or intervening in none.Conclusion The incidence of falls in hospitalized stroke patients remains relatively high and is influenced by multiple factors,including activities of daily liv-ing,degree of neurological deficit,hemiplegia,and history of falls.The Nomogram prediction model constructed based on these influencing factors can provide a reliable reference for assessing patient fall risk,and implementing targeted interven-tions based on the fall risk assessment results is expected to yield favorable clinical benefits.