Supplementary Material for: Incorporation of chest computed tomography quantification to predict outcomes for patients on hemodialysis with COVID-19

Introduction: Patients undergoing maintenance hemodialysis are vulnerable to coronavirus disease 2019 (COVID-19), exhibiting a high risk of hospitalization and mortality. Thus, early identification and intervention are important to prevent disease progression in these patients.
Methods: This was a two-center retrospective observational study of patients on hemodialysis diagnosed with COVID‐19 at the Lingang and Xuhui campuses of Shanghai Sixth People's Hospital. Patients were randomized into the training (130) and validation cohorts (54), while 59 additional patients served as an independent external validation cohort. Artificial intelligence-based parameters of chest computed tomography (CT) were quantified, and a nomogram for patient outcomes at 14 and 28 days was created by screening quantitative CT measures, clinical data, and laboratory examination items, using univariate and multivariate Cox regression models.
Results: The median dialysis duration was 48 (IQR, 24-96) months. Age, diabetes mellitus, serum phosphorus level, lymphocyte count, and chest CT score were identified as independent prognostic indicators and included in the nomogram. The concordance index values were 0.865, 0.914, and 0.885 in the training, internal validation and external validation cohorts, respectively. Calibration plots showed good agreement between the expected and actual outcomes.
Conclusion: This is the first study in which a reliable nomogram was developed to predict short-term outcomes and survival probabilities in patients with COVID-19 on hemodialysis. This model may be helpful to clinicians in treating COVID-19, managing serum phosphorus, and adjusting the dialysis strategies for these vulnerable patients to prevent disease progression in the context of COVID-19 and continuous emergence of novel viruses.

引言：维持性血液透析患者易感染新型冠状病毒肺炎（coronavirus disease 2019, COVID-19），住院及死亡风险较高，因此早期识别与干预对阻止此类患者病情进展至关重要。
方法：本研究为一项双中心回顾性观察研究，纳入上海市第六人民医院临港及徐汇院区确诊感染COVID-19的血液透析患者。将患者随机分为训练队列（130例）与内部验证队列（54例），另纳入59例患者作为独立外部验证队列。对胸部计算机断层扫描（computed tomography, CT）的人工智能量化参数进行提取，通过单因素及多因素Cox回归模型筛选定量CT指标、临床资料与实验室检查项目，构建用于预测患者14天及28天转归的列线图（nomogram）。
结果：本研究纳入患者的透析中位时长为48（四分位间距，24~96）个月。经筛选，年龄、糖尿病、血清磷水平、淋巴细胞计数及胸部CT评分被确定为独立预后指标，并纳入列线图。训练队列、内部验证队列及外部验证队列的一致性指数（concordance index）分别为0.865、0.914及0.885。校准曲线显示模型预测结果与实际转归具有良好的一致性。
结论：本研究为首个针对血液透析合并COVID-19患者构建可预测短期转归与生存概率的可靠列线图的研究。在COVID-19疫情持续及新型病毒不断出现的背景下，该模型可辅助临床医师开展COVID-19诊疗、管理血清磷水平及调整透析策略，从而阻止此类高危患者的病情进展。