摘要

研究目的

神经内科重症监护病房（Neurological Care Unit, NCU）患者通常免疫功能下降，病情危急，且接受多种侵入性操作机会多，是多重耐药菌（Multi Drug Resistance Organisms, MDRO）感染的高危人群。MDRO感染会导致患者住院时间增加、死亡风险上升，同时给患者和社会带来更大的经济压力。因此，早期识别NCU患者MDRO感染风险，并采取针对性预防措施，对于改善患者预后、降低医疗成本具有重要意义。

本研究旨在了解和掌握神经内科监护病房多重耐药菌感染风险的影响因素，构建NCU患者MDRO感染风险预测模型，并进行内部外部多重验证，以期实现早期识别高危患者、优化医疗资源配置、指导个体化治疗、促进医院感染控制的目标。

研究方法  

采用回顾性研究方法选取中国人民解放军总医院第一医学中心2020年10月- 2022年9月NCU收治的患者396例，采用随机抽样的方法选取276例作为建模组，120例作为内部验证组，另外纳入2022年10月-2023年3月入住该院的127例患者及 2022年1月-2023年12月入住新疆军区总医院 NCU的213例患者分别作为外部时间验证组及时空验证组。采用Logistic回归模型筛选独立危险因素，基于分析结果建立列线图预测模型，并通过内部及外部验证评估其效能。

研究结果

单因素分析结果显示，年龄、婚育情况、Lovett肌力分级、洼田饮水试验分级、格拉斯哥昏迷量表（Glasgow Coma Scale, GCS）评分、高血压、糖尿病、既往史种类数、低蛋白血症、是否使用抗生素及使用抗生素天数、使用呼吸机、尿管、中心静脉置管、胃管、使用约束带、体位、饮食方式、感官功能、住院天数、压疮危险因素诺顿评分、跌倒坠床风险评分、静脉外渗风险评分、导管滑脱风险评分、营养筛查风险评分等方面与MDRO感染发生具有相关性。多因素 logistic 回归分析显示，GCS评分、Lovett 肌力分级、高血压、低蛋白血症、压疮危险因素诺顿评分、住院天数等因子差异具有显著意义（P值均<0.05），基于此构建列线图模型。建模组与内部验证组的ROC曲线下面积（Area Under the Curve, AUC）分别为 0.951（95% Confidence Interval, 95% CI：0.920~0.982）、0.930（95% CI：0.879~0.980），两组模型的敏感度、特异度、约登指数分别为：0.922、 0.898、0.820 和 0.802、0.917、0.719 。HosmerLemeshow 检验结果显示模型的校准度良好（P>0.05）。校准曲线显示预测值与实际观测值一致性较高。经外部时序验证，模型总体预测准确率达83.5%。（95% CI：0.758~0.894），时空验证组预测准确率为73.7%（95% CI：0.672~0.794）。

研究结论

该预测模型展现出优异的性能指标，能够有效识别NCU患者发生MDRO感染的风险，为临床决策提供科学依据。通过实施精准化防控策略，可显著降低MDRO感染风险，优化患者临床结局，对提升医疗质量与公共卫生安全具有双重价值。

Abstract

Objectives

Patients in the Neurological Care Unit (NCU) usually have weakened immune function, critical conditions, and many opportunities to undergo various invasive procedures.They are at high risk of infection by multiple Drug Resistance Organisms (MDRO).MDRO infection can lead to an increase in the hospital stay of patients, an elevated risk of death, and bring greater economic pressure to patients and society at the same time.Therefore, early identification of the risk of MDRO infection in patients with NCU and the adoption of targeted preventive measures are of great significance for improving the prognosis of patients and reducing medical costs.

This study aims to understand and master the influencing factors of the risk of multi-drug resistant bacteria infection in the neurology intensive care unit, construct a risk prediction model for MDRO infection in patients with NCU, and conduct multiple internal and external confirmations, with the expectation of achieving the goals of early identification of high-risk patients, optimization of medical resource allocation, guidance of individualized treatment, and promotion of hospital infection control.

Methods

A retrospective study was conducted to select 396 patients admitted to the NCU of a tertiary hospital in Beijing from October 2020 to September 2022. By random sampling, 276 cases were selected as the modeling group and 120 cases as the internal validation group.In addition, 127 patients who were admitted to the hospital from October 2022 to March 2023 and 213 patients who were admitted to the NCU of a tertiary hospital in Xinjiang from January 2022 to December 2023 were included as the external time verification group and the spatio-temporal verification group respectively.The Logistic regression model was used to screen for independent risk factors. A nomogram prediction model was established based on the analysis results, and its efficacy was evaluated through internal and external validation.

Results 

The results of the univariate analysis show thatAge, marital and reproductive status, Lovett muscle strength classification, Wada drinking water test classification, Glasgow Coma ScaleGCS score, hypertension, diabetes, number of types of previous medical history, hypoproteinemia, whether antibiotics were used and the number of days of antibiotic use, use of ventilator, urinary catheter, central venous catheterization, gastric tube, use of restraints, body position, dietary pattern, sensory function, length of hospital stay, NORTON score for risk factors of pressure ulcers, risk score for fall and bed drop, risk score for venous extravasation, catheter slippageThe risk score, nutritional screening risk score and other aspects are correlated with the occurrence of MDRO infection.Multivariate logistic regression analysis showed that there were significant differences in factors such as GCS score, Lovett muscle strength grade, hypertension, hypoproteinemia, NORTON score for pressure ulcer risk factors, and length of hospital stay (all P values <0.05). Based on this, a nomogram model was constructed.the areas Under the ROC Curve (Area Under the Curve, AUC) of the modeling group and the internal validation group were 0.951 (95% Confidence Interval, 95% CI: 0.920-0.982) and 0.930 (95% CI:(0.879-0.980), the sensitivity, specificity and Youden index of the two groups of models were: 0.922, 0.898, 0.820 and 0.802, 0.917, 0.719, respectively.The Hosmer Lemeshow test results showed that the calibration of the model was good (P>0.05).The calibration curve shows that the predicted values are in high consistency with the actual observed values.Verified by external time series, the overall prediction accuracy rate of the model reaches 83.5%.(95% CI: 0.758-0.894), the prediction accuracy rate of the spatio-temporal verification group was 73.7% (95% CI: 0.672-0.794).

Conclusion 

This prediction model demonstrates excellent performance indicators and can effectively identify the risk of MDRO infection in patients with NCU, providing a scientific basis for clinical decision-making.By implementing precise prevention and control strategies, the risk of MDRO infection can be significantly reduced, the clinical outcomes of patients can be optimized, and it has dual value for improving medical quality and public health security.