Abstract：
OBJECTIVE: This study aims to develop an accurate and simplified scoring system based on the national early warning score (NEWS) to predict the mortality of intensive care unit (ICU) patients.
METHODS: The information of patients was retrieved from the Medical Information Mart for Intensive Care (MIMIC)-III and -IV databases. The Modified National Early Warning Score (MNEWS) of the patients was calculated. The discrimination ability of the MNEWS, acute physiology and chronic health scoring system II (APACHE II), and original NEWS systems in predicting patients\' mortality was evaluated using area under the receiver operating characteristic (AUROC) analysis. The DeLong test was used to estimate the receiver operating characteristic curve. The Hosmer-Lemeshow goodness-of-fit test was then applied to evaluate the calibration of the MNEWS.
RESULTS: In total, 7275 ICU patients from the MIMIC-III and -IV databases were included in the derivation cohort and 1507 ICU patients from Xi\'an Medical University were included in the validation cohort. In the derivation cohort, the nonsurvivors had significantly higher MNEWSs than the survivors (12.5 ± 3.4 vs 8.8 ± 3.4, P < 0.05). MNEWS and APACHE II both had a better performance than the NEWS in predicting hospital mortality and 90-day mortality. The optimal cutoff of MNEWS was 11. Patients with an MNEWS ≥ 11 had significantly shorter survival than those having an MNEWS of <11. Furthermore, MNEWS had a high calibration ability in predicting hospital mortality of ICU patients (χ2 = 6.534 and P = 0.588) by the Hosmer-Lemeshow test. This finding was confirmed in the validation cohort.
CONCLUSIONS: MNEWS is a simple and accurate scoring system for evaluating the severity and predicting the outcomes of ICU patients.
METHODS: The information of patients was retrieved from the Medical Information Mart for Intensive Care (MIMIC)-III and -IV databases. The Modified National Early Warning Score (MNEWS) of the patients was calculated. The discrimination ability of the MNEWS, acute physiology and chronic health scoring system II (APACHE II), and original NEWS systems in predicting patients\' mortality was evaluated using area under the receiver operating characteristic (AUROC) analysis. The DeLong test was used to estimate the receiver operating characteristic curve. The Hosmer-Lemeshow goodness-of-fit test was then applied to evaluate the calibration of the MNEWS.
RESULTS: In total, 7275 ICU patients from the MIMIC-III and -IV databases were included in the derivation cohort and 1507 ICU patients from Xi\'an Medical University were included in the validation cohort. In the derivation cohort, the nonsurvivors had significantly higher MNEWSs than the survivors (12.5 ± 3.4 vs 8.8 ± 3.4, P < 0.05). MNEWS and APACHE II both had a better performance than the NEWS in predicting hospital mortality and 90-day mortality. The optimal cutoff of MNEWS was 11. Patients with an MNEWS ≥ 11 had significantly shorter survival than those having an MNEWS of <11. Furthermore, MNEWS had a high calibration ability in predicting hospital mortality of ICU patients (χ2 = 6.534 and P = 0.588) by the Hosmer-Lemeshow test. This finding was confirmed in the validation cohort.
CONCLUSIONS: MNEWS is a simple and accurate scoring system for evaluating the severity and predicting the outcomes of ICU patients.
摘要：
目的：本研究旨在开发一种基于国家早期预警评分（NEWS）的准确，简化的评分系统，以预测重症监护病房（ICU）患者的死亡率。
方法：从重症监护医学信息集市（MIMIC）-III和-IV数据库中检索患者的信息。计算患者的改良国家早期预警评分(MNEWS)。新闻的辨别能力，急性生理和慢性健康评分系统II（APACHEII），和原始NEWS系统在预测患者死亡率方面使用受试者工作特征下面积(AUROC)分析进行评估。DeLong检验用于估计接收器工作特性曲线。然后应用Hosmer-Lemeshow拟合优度测试来评估MNEWS的校准。
结果：总计，来自MIMIC-III和-IV数据库的7275名ICU患者被纳入推导队列，来自西安医科大学的1507名ICU患者被纳入验证队列。在派生队列中，非幸存者的MNEWSs明显高于幸存者（12.5±3.4vs8.8±3.4，P<0.05）。MNEWS和APACHEII在预测医院死亡率和90天死亡率方面都比NEWS更好。MNEWS的最佳截止值为11。MNEWS≥11的患者的生存期明显短于MNEWS<11的患者。此外，通过Hosmer-Lemeshow检验，MNEWS在预测ICU患者的住院死亡率方面具有很高的校准能力（χ2=6.534和P=0.588）。这一发现在验证队列中得到证实。
结论：MNEWS是一种简单而准确的评分系统，用于评估ICU患者的严重程度和预测预后。
方法：从重症监护医学信息集市（MIMIC）-III和-IV数据库中检索患者的信息。计算患者的改良国家早期预警评分(MNEWS)。新闻的辨别能力，急性生理和慢性健康评分系统II（APACHEII），和原始NEWS系统在预测患者死亡率方面使用受试者工作特征下面积(AUROC)分析进行评估。DeLong检验用于估计接收器工作特性曲线。然后应用Hosmer-Lemeshow拟合优度测试来评估MNEWS的校准。
结果：总计，来自MIMIC-III和-IV数据库的7275名ICU患者被纳入推导队列，来自西安医科大学的1507名ICU患者被纳入验证队列。在派生队列中，非幸存者的MNEWSs明显高于幸存者（12.5±3.4vs8.8±3.4，P<0.05）。MNEWS和APACHEII在预测医院死亡率和90天死亡率方面都比NEWS更好。MNEWS的最佳截止值为11。MNEWS≥11的患者的生存期明显短于MNEWS<11的患者。此外，通过Hosmer-Lemeshow检验，MNEWS在预测ICU患者的住院死亡率方面具有很高的校准能力（χ2=6.534和P=0.588）。这一发现在验证队列中得到证实。
结论：MNEWS是一种简单而准确的评分系统，用于评估ICU患者的严重程度和预测预后。
