内科

目的分析重症监护室（ICU）终末期肾病（ESRD）患者的临床特征，并探讨其短期与远期死亡的影响因素。方法基于MIMIC-Ⅳ数据库，筛选出院诊断包含ESRD相关国际疾病分类编码的患者，收集其人口学特征、合并症及相关评分、入院情况及疾病严重程度评分、实验室指标、治疗措施、结局指标。采用Lasso-Cox回归模型进行变量初筛，采用多因素Cox比例风险回归模型分别确定ICU ESRD患者30 d及1年内死亡的独立影响因素，并绘制受试者操作特征（ROC）曲线评估联合预测模型的预测效能。结果共纳入1 185例ICU ESRD患者，年龄66.5（56.1，76.0）岁，简明急性生理学评分Ⅱ44.0（34.0，52.0）分，序贯器官衰竭评估（SOFA）评分8.0（6.0，10.0）分，ICU住院时间2.9（1.8，5.3） d，总住院时间9.3（5.7，16.8） d；ICU内、院内、30 d内、90 d内及1年内死亡率分别为8.4%、13.9%、16.5%、24.3%及39.7%。多因素Cox比例风险回归分析结果显示，年龄增大、合并脑血管疾病、合并恶性肿瘤、血磷水平升高、白细胞计数升高、SOFA评分升高均是ICU ESRD患者30 d内死亡的独立危险因素（均HR>1且P<0.05），血肌酐水平升高是其独立保护因素（HR<1，P<0.05）；年龄增大、经急诊科入院、查尔森合并症指数（CCI）升高、血磷水平升高、白细胞计数升高、血红蛋白水平升高、SOFA评分升高均是ICU ESRD患者1年内死亡的独立危险因素（均HR>1且P<0.05），血肌酐水平升高、血清白蛋白水平升高均是其独立保护因素（均HR<1且P<0.05）。基于上述独立影响因素构建的联合预测模型预测30 d及1年内死亡的ROC曲线下面积分别为0.755及0.760。结论 ICU ESRD患者远期死亡风险高。年龄、血磷水平、血肌酐水平、白细胞计数及SOFA评分是该人群短期与远期死亡的共同独立影响因素；合并脑血管疾病与恶性肿瘤是30 d内死亡的独立影响因素，而经急诊科入院、CCI、血红蛋白水平及血清白蛋白水平是1年内死亡的独立影响因素。基于上述因素构建的联合预测模型具有良好预测效能。

Objective To analyze the clinical characteristics of patients with end‑stage renal disease (ESRD) in the intensive care unit (ICU), and to explore the influencing factors for their short‑ and long‑term mortality. Methods Based on the MIMIC‑Ⅳ database, patients with discharge diagnoses including ESRD‑related International Classification of Diseases codes were screened out, whose data on demographic characteristics, comorbidities and related scores, admission status and disease severity scores, laboratory indicators, therapeutic interventions, and outcome indicators were collected. Variables were initially screened using the Lasso‑Cox regression model, and the multivariate Cox proportional hazards regression model was used to identify independent influencing factors for 30‑day and 1‑year mortality in ICU patients with ESRD, respectively. Receiver operating characteristic (ROC) curves were plotted to evaluate the predictive performance of the composite prediction models. Results A total of 1 185 ICU patients with ESRD were enrolled, with an age of 66.5 (56.1, 76.0) years, Simplified Acute Physiology Score Ⅱ of 44.0 (34.0, 52.0), Sequential Organ Failure Assessment (SOFA) score of 8.0 (6.0, 10.0), ICU length of stay of 2.9 (1.8, 5.3) days, and total hospital length of stay of 9.3 (5.7, 16.8) days, and their in‑ICU, in‑hospital, 30‑day, 90‑day, and 1‑year mortality rates were 8.4%, 13.9%, 16.5%, 24.3%, and 39.7%, respectively. Multivariate Cox proportional hazards regression analysis showed that older age, complicated with cerebrovascular disease, complicated with malignant tumor, elevated serum phosphorus level, elevated white blood cell count, and higher SOFA score were independent risk factors for 30‑day mortality in ICU patients with ESRD (all HR>1 and all P<0.05), whereas the elevated serum creatinine level was an independent protective factor (HR<1, P<0.05); older age, admission via the emergency department, higher Charlson Comorbidity Index (CCI), elevated serum phosphorus level, elevated white blood cell count, elevated hemoglobin level, and higher SOFA score were independent risk factors for 1‑year mortality in ICU patients with ESRD (all HR>1 and all P<0.05), whereas the elevated serum creatinine level and elevated serum albumin level were independent protective factors (all HR<1 and all P<0.05). The areas under the ROC curves for 30‑day and 1‑year mortality were 0.755 and 0.760, respectively, for the two composite prediction models constructed from the aforementioned independent influencing factors. Conclusion ICU patients with ESRD carry a high risk of long‑term mortality. Age, serum phosphorus level, serum creatinine level, white blood cell count, and SOFA score are common independent influencing factors for both short‑term and long‑term mortality in this population; comorbid cerebrovascular disease and malignant tumor are independent risk factors for 30‑day mortality, whereas admission via emergency department, CCI, hemoglobin level, and serum albumin level are independent influencing factors for 1‑year mortality. The composite prediction models constructed from the aforementioned factors show favorable predictive performance.

ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref
ref