Department of Computer Science, ETH Zürich, Zürich, 8092, Switzerland.
NEXUS Personalized Health Technologies, ETH Zürich, Schlieren, 8952, Switzerland.
Bioinformatics. 2024 Jun 28;40(Suppl 1):i247-i256. doi: 10.1093/bioinformatics/btae212.
Acute kidney injury (AKI) is a syndrome that affects a large fraction of all critically ill patients, and early diagnosis to receive adequate treatment is as imperative as it is challenging to make early. Consequently, machine learning approaches have been developed to predict AKI ahead of time. However, the prevalence of AKI is often underestimated in state-of-the-art approaches, as they rely on an AKI event annotation solely based on creatinine, ignoring urine output.
We construct and evaluate early warning systems for AKI in a multi-disciplinary ICU setting, using the complete KDIGO definition of AKI. We propose several variants of gradient-boosted decision tree (GBDT)-based models, including a novel time-stacking based approach. A state-of-the-art LSTM-based model previously proposed for AKI prediction is used as a comparison, which was not specifically evaluated in ICU settings yet.
We find that optimal performance is achieved by using GBDT with the time-based stacking technique (AUPRC = 65.7%, compared with the LSTM-based model's AUPRC = 62.6%), which is motivated by the high relevance of time since ICU admission for this task. Both models show mildly reduced performance in the limited training data setting, perform fairly across different subcohorts, and exhibit no issues in gender transfer.
Following the official KDIGO definition substantially increases the number of annotated AKI events. In our study GBDTs outperform LSTM models for AKI prediction. Generally, we find that both model types are robust in a variety of challenging settings arising for ICU data.
The code to reproduce the findings of our manuscript can be found at: https://github.com/ratschlab/AKI-EWS.
急性肾损伤 (AKI) 是一种影响很大一部分危重病患者的综合征，早期诊断并接受充分治疗至关重要，但早期诊断也极具挑战性。因此，已经开发了机器学习方法来提前预测 AKI。然而，由于这些方法依赖于仅基于肌酐的 AKI 事件注释，忽略了尿量，因此 AKI 的患病率在最先进的方法中经常被低估。
我们在多学科 ICU 环境中使用完整的 KDIGO AKI 定义构建和评估 AKI 的早期预警系统。我们提出了几种基于梯度提升决策树 (GBDT) 的模型变体，包括一种新的基于时间堆叠的方法。我们还使用了一种之前针对 AKI 预测提出的基于 LSTM 的最先进模型作为比较，该模型尚未在 ICU 环境中进行专门评估。
我们发现，通过使用基于 GBDT 的时间堆叠技术可以实现最佳性能（AUPRC=65.7%，而基于 LSTM 的模型的 AUPRC=62.6%），这是因为该任务与 ICU 入院后的时间高度相关。在有限的训练数据设置下，这两种模型的性能都略有下降，在不同的子队列中表现相当，并且不存在性别转移问题。
根据官方 KDIGO 定义，注释的 AKI 事件数量大大增加。在我们的研究中，GBDT 在 AKI 预测方面优于 LSTM 模型。一般来说，我们发现这两种模型类型在 ICU 数据中出现的各种具有挑战性的设置中都具有很强的稳健性。
重现我们论文结果的代码可在 https://github.com/ratschlab/AKI-EWS 上找到。
