Supplementary Material for: The Ability of Different Scoring Systems to Predict Mortality in Chronic Obstructive Pulmonary Disease Patients: A Prospective Cohort Study

Background: Chronic obstructive pulmonary disease (COPD) is a leading cause of mortality, therefore the prediction of mortality in COPD patients is crucial. In the current study, the abilities of different categorization systems to predict mortality in stable COPD patients from a prospective cohort were compared. Methods: The ability to predict mortality was compared in terms of discrimination by Harrell’s C (HC) index and calibration using graphical comparison among the GOLD (Global Initiative for Chronic Obstructive Lung Disease) 2011, GOLD 2017, GOLD grade, BODE (BMI, Airflow Obstruction, Dyspnea, Exercise), updated BODE, BODEx (BMI, Airflow Obstruction, Dyspnea, Exacerbation), e-BODE (Exacerbation and BODE), ADO (Age, Dyspnea, Airflow Obstruction), COPD prognostic index (CPI), and simplified/optimized B-AE-D (BMI, Acute Exacerbation, Dyspnea) indexes. Results: The study included 520 patients, of whom 63 died during a median 40-month follow-up period. Combined prediction systems exhibited higher discrimination properties than single predictors. The CPI exhibited the highest with a HC of 0.768, followed by the simplified B-AE-D (HC 0.761), ADO (HC 0.760), and optimized B-AE-D (HC 0.756). The BODE and its variants other than the ADO exhibited relatively lower HCs (0.656–0.705), and GOLD exhibited the lowest discrimination ability among the combined indices (HCs 0.628–0.637). Subjective symptom questionnaires such as the modified Medical Research Council (mMRC) scale (HC 0.693) and SGRQ (HC 0.679) exhibited the highest ability to predict mortality among the single indices. Conclusion: The ADO, simplified B-AE-D, optimized B-AE-D, and GOLD 2017 exhibited good calibration properties, but the CPI did not. The simplified and optimized B-AE-Ds and the ADO index had good discrimination and calibration properties for the prediction of mortality in stable COPD patients.背景：慢性阻塞性肺疾病（Chronic obstructive pulmonary disease, COPD）是主要致死性病因之一，因此对慢阻肺患者的死亡风险进行预测具有重要临床价值。本研究旨在对比不同分类预测系统在前瞻性队列的稳定期慢阻肺患者死亡风险预测中的效能。 方法：本研究以哈雷尔C（Harrell’s C, HC）指数评估区分度、通过图形化对比评估校准性能为评价框架，对比了以下多种模型对死亡风险的预测能力：2011版全球慢性阻塞性肺疾病防治创议（Global Initiative for Chronic Obstructive Lung Disease, GOLD）指南、2017版GOLD指南、GOLD分级、BODE指数（体重指数、气流受限、呼吸困难、运动能力）、改良版BODE指数、BODEx指数（体重指数、气流受限、呼吸困难、急性加重）、e-BODE指数（急性加重与BODE指数）、ADO指数（年龄、呼吸困难、气流受限）、慢阻肺预后指数（COPD prognostic index, CPI）以及简化/优化版B-AE-D指数（体重指数、急性加重、呼吸困难）。 结果：本研究共纳入520例患者，其中63例患者在中位随访40个月期间死亡。联合预测模型的区分度优于单一预测指标。其中CPI的区分度最高，HC指数为0.768，其次为简化版B-AE-D（HC=0.761）、ADO（HC=0.760）与优化版B-AE-D（HC=0.756）。除ADO外的BODE指数及其衍生模型的HC指数相对较低（0.656~0.705），而在所有联合型指标中，GOLD指南的区分能力最差（HC指数为0.628~0.637）。单一预测指标中，改良医学研究委员会（modified Medical Research Council, mMRC）呼吸困难量表（HC=0.693）与圣乔治呼吸问卷（SGRQ，HC=0.679）这类主观症状评估工具的死亡预测效能最优。 结论：ADO、简化版B-AE-D、优化版B-AE-D以及2017版GOLD指南具备良好的校准性能，但CPI则未表现出该特性。简化版与优化版B-AE-D以及ADO指数在稳定期慢阻肺患者的死亡风险预测中，同时具备优异的区分度与校准性能。