Division of Infectious Diseases & International Health, Department of Medicine, University of Virginia School of Medicine, Charlottesville, VA, USA; Office of Hospital Epidemiology/Infection Prevention & Control, University of Virginia School of Medicine, Charlottesville, VA, USA.
Department of Microbiology, Immunology, and Cancer Biology, University of Virginia, Charlottesville, VA, USA.
EBioMedicine. 2024 Aug;106:105244. doi: 10.1016/j.ebiom.2024.105244. Epub 2024 Jul 17.
Clostridioides difficile infection results in life-threatening short-term outcomes and the potential for subsequent recurrent infection. Predicting these outcomes at diagnosis, when important clinical decisions need to be made, has proven to be a difficult task.
52 clinical features from existing models or the literature were collected retrospectively within ±48 h of diagnosis among 1660 inpatient infections. A modified desirability of outcome ranking (DOOR) was designed to encompass clinically-important severe events attributable to the acute infection (intensive care transfer due to sepsis, shock, colectomy/ileostomy, mortality) and/or 60-day recurrence. A deep neural network was constructed and interpreted using SHapley Additive exPlanations (SHAP). High-importance features were used to train a reduced, shallow network and performance was compared to existing conventional models (7 severity, 7 recurrence; after summing DOOR probabilities to align with conventional binary outputs) using area under the ROC curve (AUROC) and DeLong tests.
The full (52-feature) model achieved an out-of-sample AUROC 0.823 for severity and 0.678 for recurrence. SHAP identified 13 unique, highly-important features (age, hypotension, initial treatment, onset, PCR cycle threshold, number of prior episodes, antibiotic exposure, fever, hypotension, pressors, leukocytosis, creatinine, lactate) that were used to train a reduced model, which performed similarly to the full model (severity AUROC difference P = 0.130; recurrence P = 0.426) and significantly better than the top severity model (reduced model predicting severity 0.837, ATLAS 0.749; P = 0.001). The reduced model also outperformed the top recurrence model, but this was not statistically-significant (reduced model recurrence AUROC 0.653, IDSA Recurrence Risk Criteria 0.595; P = 0.196). The final, reduced model was deployed as a web application with real-time SHAP explanations.
Our final model outperformed existing severity and recurrence models; however, it requires external validation. A DOOR output allows specific clinical questions to be asked with explainable predictions that can be feasibly implemented with limited computing resources.
National Institutes of Health-Institute of Allergy and Infectious Diseases.
艰难梭菌感染会导致危及生命的短期结局，并有可能随后再次感染。在需要做出重要临床决策时，预测这些结局证明是一项艰巨的任务。
在 1660 例住院感染患者中，我们在诊断后±48 小时内回顾性地收集了来自现有模型或文献中的 52 个临床特征。设计了一种改良的结局偏好度评分（DOOR），以涵盖与急性感染相关的临床重要严重事件（因败血症、休克、结肠切除术/回肠造口术、死亡率而转入重症监护病房）和/或 60 天复发。使用 SHapley Additive exPlanations（SHAP）构建并解释了一个深度神经网络。使用高重要性特征训练一个简化的浅层网络，并使用 ROC 曲线下面积（AUROC）和 DeLong 检验比较其与现有传统模型（7 个严重程度，7 个复发；通过将 DOOR 概率相加以与传统的二进制输出对齐）的性能。
全（52 个特征）模型的严重程度和复发的外部样本 AUROC 分别为 0.823 和 0.678。SHAP 确定了 13 个独特的、高度重要的特征（年龄、低血压、初始治疗、发病、PCR 循环阈值、既往发作次数、抗生素暴露、发热、低血压、升压药、白细胞增多症、肌酐、乳酸），这些特征用于训练简化模型，该模型的性能与全模型相似（严重程度 AUROC 差异 P=0.130；复发 P=0.426），且显著优于最佳严重程度模型（简化模型预测严重程度 0.837，ATLAS 0.749；P=0.001）。简化模型也优于最佳复发模型，但这在统计学上并不显著（简化模型复发 AUROC 为 0.653，IDSA 复发风险标准为 0.595；P=0.196）。最终的简化模型作为一个具有实时 SHAP 解释的网络应用程序进行了部署。
我们的最终模型优于现有的严重程度和复发模型；然而，它需要外部验证。DOOR 输出允许提出具体的临床问题，并提供可解释的预测，这些预测可以在有限的计算资源下得以实现。
美国国立卫生研究院过敏与传染病研究所。
