Zhongguo cuzhong zazhi (Apr 2023)

无基础代谢性疾病急性缺血性卒中患者短期营养不良预测模型构建与验证 Construction and Verification of Short-term Malnutrition Prediction Model for Patients with Acute Ischemic Stroke without Basal Metabolic Diseases

  • 张兰, 沈晓芳, 金瑾, 徐吉, 张静

DOI
https://doi.org/10.3969/j.issn.1673-5765.2023.04.008
Journal volume & issue
Vol. 18, no. 4
pp. 428 – 433

Abstract

Read online

目的 分析无基础代谢性疾病急性缺血性卒中(acute ischemic stroke,AIS)患者短期营养不良的影响因素,并构建AIS短期营养不良的临床预测模型。 方法 回顾性选取2019年3月—2021年6月在苏州市第九人民医院神经内科治疗的无基础代谢性疾病AIS患者作为建模队列,以患者入院2周时微型营养评定简表(mini-nutritional assessment short-form,MNA-SF)、BMI和白蛋白(albumin,Alb)指标作为综合评估,即MNA-SF评分<8分且BMI<18.5 kg/m2、Alb<35.0 g/L为营养不良,以此为依据划分营养不良组和无营养不良组。比较两组的人口学特征及入院时临床资料,将单因素分析具有统计学意义的变量纳入多因素logistic回归分析,基于多因素分析出的影响指标构建预测模型。用ROC曲线评价其区分度,用拟合优度检验评价其校准度。以2021年8月—2022年6月在苏州市第九人民医院神经内科治疗的无基础代谢性疾病AIS患者作为验证队列验证模型的效能。 结果 建模队列共纳入924例无基础代谢性疾病AIS患者,平均(55.0±14.3)岁,男性544例(58.9%)。入院2周时的营养评估发现72例(7.8%)营养不良。多因素分析显示,高龄(OR 2.059,95%CI 1.132~3.743,P=0.018)、饮酒史(OR 1.747,95%CI 1.076~2.835,P=0.024)、入院时有营养风险(OR 2.951,95%CI 1.485~5.859,P=0.002)、院内无营养支持(OR 1.870,95%CI 1.065~3.284,P=0.029)、入院才藤氏分级低(OR 0.226,95%CI 0.079~0.652,P=0.006)、入院NIHSS评分高(OR 1.556,95%CI 1.057~2.289,P=0.025)均为无基础代谢性疾病AIS患者短期营养不良的独立影响因素。根据影响因素得出预测模型方程:logit(P)=0.722×年龄+0.558×饮酒史+1.082×入院时营养风险+0.626×院内无营养支持-1.483×才藤氏分级+0.442×NIHSS评分+0.479-31.187。该模型预测患者短期营养不良的AUC为0.863(95%CI 0.811~0.914),最大约登指数(0.703)对应的灵敏度和特异度分别为87.50%、83.90%;拟合优度检验χ2=2.754,P=0.498。验证队列共纳入126例无基础代谢性疾病AIS患者,采用模型对其营养不良进行预测得出的灵敏度为86.96%、特异度为83.50%、准确率为84.13%。 结论 高龄、饮酒史、入院时有营养风险、院内无营养支持、才藤氏分级低和入院NIHSS评分高是无基础代谢性疾病AIS患者短期营养不良的影响因素,经验证基于上述指标构建临床预测模型具有良好的预测效能。 Objective To analyze the influencing factors of short-term malnutrition in patients with acute ischemic stroke (AIS) without basal metabolism disease, and construct a clinical prediction model of short-term malnutrition in AIS. Methods The AIS patients without basal metabolism diseases from Department of Neurology of Suzhou Ninth People’s Hospital from March 2019 to June 2021 were retrospectively included as the modeling cohort. The mini-nutritional assessment short-form (MNA-SF) , BMI and albumin (Alb) were used to evaluate for nutrition condition at 2 weeks after admission. The malnutrition was defined as the MNA-SF score < 8, BMI < 18.5 kg/m2 and Alb < 35.0 g/L, then the patients were divided into malnutrition group and non-malnutrition group. A prediction model was constructed based on the variables screened by multivariate logistic regression analysis. The discrimination was evaluated by ROC curve, and the calibration was evaluated by goodness of fit test. AIS patients without underlying metabolic diseases in Department of Neurology of Suzhou Ninth People’s Hospital from August 2021 to June 2022 were selected as the validation cohort of the model. Results The modeling cohort included 924 AIS patients, with the mean age of 55.0±14.3 years and 544 males (58.9%) .72 cases (7.8%) of malnutrition were screened at 2 weeks after admission. Multivariate analysis showed that advanced age (OR 2.059, 95%CI 1.132-3.743, P=0.018) , drinking history (OR 1.747, 95%CI 1.076-2.835, P=0.024) , nutritional risk at admission (OR 2.951, 95%CI 1.485-5.859, P=0.002) , no nutritional support in hospital (OR 1.870, 95%CI 1.065-3.284, P=0.029) , low admission Caiteng's grade (OR 0.226, 95% CI 0.079-0.652, P=0.006) , and high admission NIHSS score (OR 1.556, 95%CI 1.057-2.289, P=0.025) were independent influencing factors of short-term malnutrition in AIS patients without underlying metabolic diseases. According to the influencing factors, the prediction model equation is: logit (P) = 0.722 × age + 0.558 × drinking history + 1.082 × nutritional risk at admission + 0.626 × no nutritional support in hospital-1.483 × Caiteng grade + 0.442 × NIHSS score + 0.479-31.187. The AUC of this model for predicting short-term malnutrition was 0.863 (95% CI 0.811-0.914) . The sensitivity and specificity of the maximum Youden index (0.703) were 87.50% and 83.90%, respectively. Goodness of fit test χ2=2.754, P=0.498. The validation cohort included 126 AIS patients without underlying metabolic diseases. The sensitivity, specificity and accuracy of the validation cohort-based model for predicting malnutrition were 86.96%, 83.50% and 84.13%, respectively. Conclusions Advanced age, drinking history, nutritional risk at admission, no nutritional support in hospital, low Caiteng grade and high NIHSS score at admission were the influencing factors of short-term malnutrition in AIS patients without underlying metabolic diseases. The clinical prediction model based on the above indicators has good prediction performance.

Keywords