Griffin Benjamin R, Mudireddy Avinash, Horne Benjamin D, Chonchol Michel, Goldstein Stuart L, Goto Michihiko, Matheny Michael E, Street W Nick, Vaughan-Sarrazin Mary, Jalal Diana I, Misurac Jason
Division of Nephrology, Department of Medicine, University of Iowa Carver College of Medicine, Iowa City, IA.
Center for Access & Delivery Research and Evaluation, Iowa City VAMC, Iowa City, IA.
Kidney Med. 2024 Oct 15;6(12):100918. doi: 10.1016/j.xkme.2024.100918. eCollection 2024 Dec.
Acute kidney injury (AKI) is a common complication among hospitalized adults, but AKI prediction and prevention among adults has proved challenging. We used machine learning to update the nephrotoxic injury negated by just-in time action (NINJA), a pediatric program that predicts nephrotoxic AKI, to improve accuracy among adults.
A retrospective cohort study.
Adults admitted for > 48 hours to the University of Iowa Hospital from 2017 to 2022.
A NINJA high-nephrotoxin exposure (≥3 nephrotoxins on 1 day or intravenous aminoglycoside or vancomycin for ≥3 days).
AKI within 48 hours of high-nephrotoxin exposure.
We collected 85 variables, including demographics, laboratory tests, vital signs, and medications. AKI was defined as a serum creatinine increase of ≥0.3 mg/dL. A gated recurrent unit (GRU)-based recurrent neural network (RNN) was trained on 85% of the data, and then tested on the remaining 15%. Model performance was evaluated with precision, recall, negative predictive value, and area under the curve. We used an artificial neural network to determine risk factor importance.
There were 14,480 patients, 18,180 admissions, and 37,300 high-nephrotoxin exposure events meeting inclusion criteria. In the testing cohort, 29% of exposures developed AKI within 48 hours. The RNN-GRU model predicted AKI with a precision of 0.60, reducing the number of false alerts from 2.5 to 0.7 per AKI case. Lowest hemoglobin, lowest blood pressure, and highest white blood cell count were the most important variables in the artificial neural network model. Acyclovir, piperacillin-tazobactam, calcineurin inhibitors, and angiotensin-converting enzyme inhibitor/angiotensin receptor blockers were the most important medications.
Clinical variables and medications were not exhaustive, drug levels or dosing were not incorporated, and Iowa's racial makeup may limit generalizability.
Our RNN-GRU model substantially reduced the number of false alerts for nephrotoxic AKI, which may facilitate NINJA translation to adult hospitals by providing more targeted intervention.
急性肾损伤（AKI）是住院成人患者常见的并发症，但事实证明，对成人进行AKI预测和预防具有挑战性。我们运用机器学习对“及时行动消除肾毒性损伤”（NINJA）进行更新，这是一个预测肾毒性AKI的儿科项目，旨在提高其在成人中的准确性。
一项回顾性队列研究。
2017年至2022年入住爱荷华大学医院且住院时间超过48小时的成人患者。
NINJA定义的高肾毒素暴露（一天内≥3种肾毒素或静脉注射氨基糖苷类药物或万古霉素≥3天）。
高肾毒素暴露后48小时内发生的AKI。
我们收集了85个变量，包括人口统计学、实验室检查、生命体征和用药情况。AKI定义为血清肌酐升高≥0.3mg/dL。基于门控循环单元（GRU）的循环神经网络（RNN）在85%的数据上进行训练，然后在其余15%的数据上进行测试。通过精确率、召回率、阴性预测值和曲线下面积评估模型性能。我们使用人工神经网络确定危险因素的重要性。
共有14480例患者、18180次入院以及37300次高肾毒素暴露事件符合纳入标准。在测试队列中，29%的暴露事件在48小时内发生了AKI。RNN-GRU模型预测AKI的精确率为0.60，将每例AKI病例的误报数量从2.5次减少至0.7次。最低血红蛋白、最低血压和最高白细胞计数是人工神经网络模型中最重要的变量。阿昔洛韦、哌拉西林-他唑巴坦、钙调神经磷酸酶抑制剂以及血管紧张素转换酶抑制剂/血管紧张素受体阻滞剂是最重要的药物。
临床变量和用药情况并不详尽，未纳入药物水平或剂量信息，且爱荷华州的种族构成可能会限制研究结果的普遍性。
我们的RNN-GRU模型大幅减少了肾毒性AKI的误报数量，通过提供更具针对性的干预措施，这可能有助于将NINJA推广至成人医院。
