Hu Biao, Yang Zhandong, Yuan Lu, Huang Supu, Huang Xiuli, Wang Chudong, Cai Jiaxin
Department of Radiology, the Second Affiliated Hospital of Guangzhou Medical University, Guangzhou 510260, China; Guangzhou Medical University, Guangzhou 511436, China.
Guangzhou Medical University, Guangzhou 511436, China.
Diabetes Res Clin Pract. 2025 Aug;226:112301. doi: 10.1016/j.diabres.2025.112301. Epub 2025 May 31.
The Stress Hyperglycemia Ratio (SHR) reflects stress-related hyperglycemia and is linked to poor outcomes in various diseases. This study explores the association between SHR and all-cause mortality in hyperlipidemia patients and its value in enhancing predictive models.
Data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) included 4,883 first-admission hyperlipidemia patients. Patients were grouped by SHR quartiles. Outcomes included in-hospital, ICU, and 28-day mortality. Cox regression, restricted cubic spline (RCS), and ROC curves were used. Seventeen machine learning models assessed SHR's predictive value, with LightGBM evaluating variable importance.
We extracted data from 4,883 first-admission hyperlipidemia patients aged 18-100 years (mean age 70 ± 12.1; 36.0 % female). SHR was independently associated with ICU mortality (HR = 1.29, 95 % CI 1.08-1.56), in-hospital mortality (HR = 1.24, 95 % CI 1.07-1.44), and 28-day mortality (HR = 1.32, 95 % CI 1.15-1.52). RCS analysis showed a linear association between SHR and mortality. Adding SHR to general models improved predictive accuracy.
In the study of first-admission hyperlipidemia patients, elevated SHR levels were significantly associated with in-hospital, ICU, and 28-day all-cause mortality. Incorporating SHR into existing models enhances their predictive ability, holding significant clinical value for identifying high-risk patients.
应激性高血糖比值（SHR）反映了与应激相关的高血糖，并且与多种疾病的不良预后相关。本研究探讨了SHR与高脂血症患者全因死亡率之间的关联及其在增强预测模型中的价值。
重症监护医学信息集市IV（MIMIC-IV）的数据包括4883例首次入院的高脂血症患者。患者按SHR四分位数分组。结局包括住院、重症监护病房（ICU）和28天死亡率。使用Cox回归、受限立方样条（RCS）和ROC曲线。17种机器学习模型评估了SHR的预测价值，LightGBM评估变量重要性。
我们从4883例年龄在18至100岁（平均年龄70±12.1岁；女性占36.0%）的首次入院高脂血症患者中提取了数据。SHR与ICU死亡率（HR = 1.29，95%CI 1.08 - 1.56）、住院死亡率（HR = 1.24，95%CI 1.07 - 1.44）和28天死亡率（HR = 1.32，95%CI 1.15 - 1.52）独立相关。RCS分析显示SHR与死亡率之间存在线性关联。将SHR添加到一般模型中可提高预测准确性。
在首次入院高脂血症患者的研究中，SHR水平升高与住院、ICU和28天全因死亡率显著相关。将SHR纳入现有模型可增强其预测能力，对识别高危患者具有重要的临床价值。
