COTE 与 Charlson 合并症指数对慢阻肺急性加重患者死亡预测能力比较
该研究利用丹麦全国登记数据，纳入 2005–2018 年首次因慢阻肺急性加重入院的 87 271 例患者，比较 COTE、CCI 及更新权重 CCI（Quan）评分对 1 年死亡率的预测能力。结果显示三种评分预测能力相近，AUC 分别为 0.707、0.714 和 0.714。
要点
- 首次 AECOPD 入院患者中，COTE 与 CCI 评分对 1 年死亡率的预测能力总体相近。
- COTE、CCI 和 CCI（Quan）模型的 AUC 分别为 0.707、0.714 和 0.714。
- Brier 评分显示 CCI 模型仅略优于 COTE 模型，临床差异有限。
核心结论
- 在首次因 AECOPD 入院的患者队列中，COTE、CCI 与 CCI（Quan）评分对 1 年死亡率的预测能力相似。
研究设计与样本
- 基于丹麦全国登记数据的回顾性队列研究，纳入 2005–2018 年首次因 AECOPD 入院的全部患者共 87 271 例。
- 采用 logistic 回归并配合五折交叉验证比较 AUC 和 Youden 指数确定的最佳阈值。
- 进一步行 10 000 次 bootstrap 分析并通过 T 检验比较准确性分布，使用 Cox 回归比较随时间变化的 Brier 评分。
主要发现
- 研究共识别 87 271 例首次 AECOPD 患者，中位年龄 74 岁，IQR（66，81），男性占 48.4%。
- CCI 和 CCI（Quan）模型的 AUC 为 0.714，COTE 模型 AUC 为 0.707（p<0.001）。
- Brier 评分 CCI 为 0.1793，CCI（Quan）为 0.1789，COTE 为 0.1802，仅略微倾向于 CCI 模型。
关键数值与推荐
- CCI 和 CCI（Quan）模型 AUC 均为 0.714，COTE 模型 AUC 为 0.707（p<0.001）。
- Brier 评分：CCI 0.1793，CCI（Quan）0.1789，COTE 0.1802。
临床怎么用
- 对首次 AECOPD 住院患者进行合并症评估时，COTE 与 CCI 评分在预测 1 年死亡风险方面可视为具有相近的临床实用性。
局限性
- 研究为回顾性队列设计，基于登记数据，可能存在未测量的混杂因素。
- 研究对象限于丹麦首次 AECOPD 入院患者，结果外推至其他人群需谨慎。
以上解读依据论文摘要（全文未开放获取或许可不允许翻译转载）。
原文
Introduction The chronic obstructive pulmonary disease (COPD)-specific comorbidity test (COTE) aims to predict long-term mortality in patients with COPD, whereas the Charlson comorbidity index (CCI) score is used as a validated tool for predicting long-term mortality across medical conditions. Comparison of the two comorbidity scores’ ability to predict mortality has only been studied scarcely. Aims and objectives This study aimed to compare the predictive ability of the COTE and CCI scores including the updated CCI score with updated weights (CCI (Quan)) in patients with COPD following acute exacerbation of chronic obstructive pulmonary disease (AECOPD). Methods Using nationwide registry data, all Danish patients at first admission due to AECOPD during 2005–2018 were included in this retrospective cohort study. 1-year mortality was explored including age, sex and CCI or COTE comorbidity scores. Logistic regression was performed with fivefold cross-validation comparing area under the curve (AUC) and optimal threshold using Youden index. The logistic regression models were furthermore examined using 10 000 bootstrap analyses and comparing the distribution of accuracy using a T-test. Cox regression analysis was used to compare Brier scores over time for models including combinations of age, sex, COTE, CCI and CCI (Quan). Results A total of 87 271 individuals were identified with a first-time AECOPD during the study period. The median age was 74 IQR (66, 81) and 48.4% were males. The logistic regression revealed an AUC of the CCI and CCI (Quan) models of 0.714 and for the COTE model 0.707 (p<0.001). Comparison using T-test on 10 000 bootstrap samples revealed small differences with an accuracy of 0.7281 (95% CI 0.7241 to 0.7320) for the COTE model and 0.7275 (95% CI 0.7235 to 7314) for the CCI model (p<0.001). Brier scores were 0.1793 for CCI, 0.1789 for CCI (Quan) and 0.1802 for COTE, only slightly favouring the CCI model. Conclusions In a cohort of patients with first-time AECOPD, we found similar predictive abilities of 1-year mortality comparing the CCI, CCI (Quan) and COTE score.
引用
- BMJ Open Respiratory Research：Predictive ability of the COPD-specific comorbidity test (COTE) score versus the Charlson comorbidity index (CCI) in patients with COPD
供执业医师工作参考，不替代临床判断；数值与分级以指南原文为准。