Background An episode of acute ulcerative colitis (UC) represents an important watershed moment in a patient’s disease course. Foreknowledge of a patient's likely response to intravenous corticosteroid therapy has significant clinical utility. Using a large prospectively collected acute UC patient database and machine learning-based techniques we aimed to derive and validate a personalised algorithm for identifying patients at high risk of corticosteroid therapy failure from variables available at hospital presentation. Methods A prospectively collected database of 600 consecutive presentations of acute UC was collated at a single referral centre between 1996 and 2022. An AIC-based Elastic Net model was used to select variables on the 419 earliest presentations of acute UC (1996-2017). Two risk-scoring algorithms, with and without utilising additional endoscopic variables, were constructed using logistic regression models. These risk scores were then validated on a separate cohort of 181 acute UC presentations (2018-2022). Results The partial risk of rescue (ROR) score included the admission indices of oral corticosteroid treatment; bowel frequency ≥6/24 hours; albumin; CRP ≥12mg/ml and log10CRP. The full ROR score incorporates the same variables with the addition of the Mayo endoscopic subscore and disease extent. The ROC AUCs in the validation cohort were 0.76 (95% CI: 0.69-0.83) and 0.78 (95% CI: 0.71-0.85) for the partial and full ROR scores, respectively. When incomplete cases were excluded, the full ROR score validation cohort ROC AUC increased from 0.78 to 0.80. Conclusion These pragmatic personalised risk scores (available at www.severecolitis.com) have comparably strong performance characteristics and usability enabling the identification of individuals at high risk of corticosteroid treatment failure before or after endoscopic assessment. These patients may be suitable for consideration of early treatment escalation or screening for participation in clinical trials.

中文翻译：

---

P743 预测急性溃疡性结肠炎静脉注射皮质类固醇失败风险的个性化算法

背景 急性溃疡性结肠炎（UC）的发作是患者病程中的一个重要分水岭。预知患者对静脉注射皮质类固醇治疗的可能反应具有重要的临床实用性。使用大型前瞻性收集的急性 UC 患者数据库和基于机器学习的技术，我们旨在推导并验证个性化算法，用于根据医院就诊时提供的变量识别皮质类固醇治疗失败高风险的患者。方法 1996 年至 2022 年间，在一个转诊中心对前瞻性收集的 600 例连续急性 UC 病例数据库进行了整理。基于 AIC 的弹性网络模型用于选择 419 例最早的急性 UC 病例（1996-2017 年）的变量。使用逻辑回归模型构建了两种风险评分算法，使用和不使用额外的内窥镜变量。然后，这些风险评分在由 181 例急性 UC 表现组成的单独队列中进行了验证（2018-2022 年）。结果部分救援风险（ROR）评分包括口服皮质类固醇治疗入院指标；排便次数≥6/24小时；白蛋白; CRP ≥12mg/ml 和 log10CRP。完整的 ROR 评分包含相同的变量，并添加了 Mayo 内窥镜子评分和疾病程度。验证队列中部分 ROR 评分和完整 ROR 评分的 ROC AUC 分别为 0.76 (95% CI: 0.69-0.83) 和 0.78 (95% CI: 0.71-0.85)。当排除不完整的病例时，完整的 ROR 评分验证队列 ROC AUC 从 0.78 增加到 0.80。结论 这些实用的个性化风险评分（可在 www.severecolitis.com 上获取）具有相对较强的性能特征和可用性，能够在内窥镜评估之前或之后识别皮质类固醇治疗失败高风险的个体。这些患者可能适合考虑早期治疗升级或筛选参与临床试验。