【摘要】 目的 分析急性脑梗死(ACI)患者发生早期神经功能恶化(END)的危险因素，构建预测患者发生END的临床模型。方法收集2023年12月至2025年1月在神经内科住院的210例ACI患者的临床数据，结合患者入院时的美国国立卫生院神经功能缺损评分(NIHSS)，使用LASSO回归和Logistic回归分析END发生的危险因素，并建立列线图模型，使用Bootstrap法进行内部验证。另收集2025年2月至6月63例ACI患者的临床数据，对模型进行再次验证。结果 与非END组相比，END组患者入院时合并高血压(P=0.015)、高脂血症(P<0.001)、是否颈动脉狭窄率大于50%(P<0.001)之间差异有统计学意义。LASSO回归和多因素Logistic回归分析结果显示:高脂血症(OR=1.96,95%CI:1.40～6.26,P=0.004)，颈动脉狭窄大于50%(OR=2.01,95%CI:1.61～5.59,P<0.001)、入院时NIHSS评分(OR=2.44,95%CI:1.40～4.24,P=0.002)、PLT(OR=1.04,95%CI:1.01～1.06,P=0.006)、CRP(OR=1.02,95%CI:1.01～1.03,P<0.001)、HCY(OR=1.36,95%CI:1.03～1.81,P=0.032)和NSE(OR=1.37,95%CI:1.02～1.85,P=0.038)是END的危险因素，而PT(OR=0.34,95%CI:0.15～0.74,P=0.007)是END的保护性因素。联合上述指标建立列线图模型的Hosmer-Lemeshow检验P值为0.111,ROC曲线下面积(AUC)为0.889,95%CI:0.840～0.930，敏感性为87.0%，特异性为79.0%。此外，Bootstrap内部验证和验证组的AUC分别为0.890(95%CI:0.847～0.934)和0.814(95%CI:0.736～0.946)。校准曲线和决策曲线分析结果均显示该列线图模型具有良好的性能。结论 构建的模型可以有效预测END的发生，为临床医生早期识别高风险患者提供依据。更多还原

【Abstract】 Objective To analyze the risk factors associated with early neurological deterioration( END) in patients with acute cerebral infarction( ACI) and to develop a clinical predictive model for predicting the occurrence of END. Methods Clinical data from210 patients admitted to the neurology department between December 2023 and January 2025 were retrospectively collected. Based on admission National Institutes of Health Stroke Scale( NIHSS) scores,risk factors for END were analyzed using LASSO regression and multivariate Logistic regression,and a predictive nomogram model was subsequently developed. Internal validation was performed using the Bootstrap method. An additional cohort of 63 ACI patients admitted from February to June 2025 was used for external validation.Results Compared with the non-END group,the END group exhibited a statistically significant higher prevalence of hypertension( P= 0.015),hyperlipidemia( P<0.001),and carotid stenosis exceeding 50%( P<0.001) at baseline. LASSO and multivariable Logistic regression analyses identified hyperlipidemia( OR = 1. 96,95% CI: 1. 40 to 6. 26,P = 0. 004),carotid stenosis > 50%( OR =2.01,95%CI: 1.61 to 5.59,P<0.001),NIHSS score at admission( OR = 2.44,95%CI: 1.40 to 4.24,P = 0.002),PLT( OR = 1.04,95%CI: 1.01 to 1.06,P = 0.006),CRP( OR = 1.02,95%CI: 1.01 to 1.03,P<0.001),homocysteine( HCY)( OR = 1.36,95%CI:1.03 to 1.81,P = 0.032),and neuron-specific enolase( NSE)( OR = 1.37,95%CI: 1.02 to 1.85,P = 0.038) as independent risk factors for END,whereas prothrombin time( PT) was identified as a protective factor( OR = 0.34,95%CI: 0.15 to 0.74,P = 0.007). The nomogram model incorporating these variables demonstrated good calibration,with a Hosmer-Lemeshow test P value of 0.111 and area under the ROC curve( AUC) of 0.889( 95%CI: 0.840 to 0.930),sensitivity of 87.0%,and specificity of 79.0%. Bootstrap internal validation yielded an AUC of 0. 890( 95% CI: 0. 847 to 0. 934),while the external validation cohort achieved an AUC of 0. 814( 95%CI: 0.736 to 0.946). The results of both calibration and decision curve analyses demonstrated the robust clinical applicability.Conclusion The model developed in this study demonstrates effective predictive capability for the occurrence of END,offering a reliable foundation for early clinical identification of high-risk patients.更多还原