ORIGINAL ARTICLES原创文章

FDA-Authorized AI/ML Tool for Sepsis Prediction: Development and Validation

A. Bhargava and Others

Abstract

BACKGROUND

Sepsis is a life-threatening condition that demands prompt treatment for improved patient outcomes. Its heterogenous presentation makes early detection challenging, highlighting the need for effective risk assessment tools. Artificial intelligence (AI) models could potentially identify patients with sepsis, but none have previously been authorized by the U.S. Food and Drug Administration (FDA) for commercial use. This study outlines the development and validation of the Sepsis ImmunoScore, the first FDA-authorized AI-based software designed to identify patients at risk of sepsis.

METHODS

In this prospective study, we enrolled adult patients (18+ years of age) suspected of infection, as indicated by a blood culture order, from five U.S. institutions between April 2017 and July 2022. The participants were divided into an algorithm development cohort (n=2366), an internal validation cohort (n=393), and an external validation cohort (n=698). The primary end point was sepsis presence (as defined by Sepsis-3) within 24 hours of test initiation. Secondary end points included length of hospital stay, intensive care unit (ICU) admission within 24 hours, mechanical ventilation use within 24 hours, vasopressor use within 24 hours, and in-hospital mortality.

RESULTS

The Sepsis ImmunoScore demonstrated high diagnostic accuracy, with an area under the curve of 0.85 (0.83 to 0.87) in the derivation cohort, 0.80 (0.74 to 0.86) in internal validation, and 0.81 (0.77 to 0.86) in external validation. The scores were categorized into four sepsis risk levels with corresponding likelihood ratios: low (0.1), medium (0.5), high (2.1), and very high (8.3). These risk categories also predicted in-hospital mortality: low (0.0%), medium (1.9%), high (8.7%), and very high (18.2%) in the external validation cohort. Similar trends were observed for other metrics, such as length of hospital stay, ICU utilization, mechanical ventilation, and vasopressor use.

CONCLUSIONS

The Sepsis ImmunoScore demonstrated high accuracy for the identification and prediction of sepsis and critical illness metrics that could enable prompt identification of patients at high risk of sepsis and adverse outcomes, potentially improving clinical decision-making and patient outcomes. (Funded by the Defense Threat Reduction Agency and others.)

DOI: 10.1056/AIoa2400867

全文链接：https://ai.nejm.org/doi/abs/10.1056/AIoa2400867

FDA授权的人工智能/机器学习工具用于败血症预测：开发和验证

A. Bhargava 等人

摘要： 背景： 败血症是一种威胁生命的状况，需要及时的治疗以改善患者结果。其异质性表现使得早期检测具有挑战性，这突出了有效风险评估工具的需求。人工智能（AI）模型有可能识别出败血症患者，但以前没有被美国食品药品监督管理局（FDA）授权用于商业用途。这项研究概述了Sepsis ImmunoScore的开发和验证，这是第一个FDA授权的基于AI的软件，旨在识别败血症风险患者。 方法： 在这项前瞻性研究中，我们从2017年4月至2022年7月，从五个美国机构招募了怀疑感染的成年患者（18岁以上），如血液培养订单所示。参与者被分为算法开发队列（n=2366）、内部验证队列（n=393）和外部验证队列（n=698）。主要终点是在测试开始后24小时内出现败血症（根据Sepsis-3定义）。次要终点包括住院时间、24小时内ICU入院、24小时内机械通气使用、24小时内使用血管活性药物和住院死亡率。 结果： Sepsis ImmunoScore显示出高诊断准确性，在衍生队列中曲线下面积为0.85（0.83至0.87），在内部验证中为0.80（0.74至0.86），在外部验证中为0.81（0.77至0.86）。分数被分为四个败血症风险水平，对应的可能性比率分别为：低（0.1）、中等（0.5）、高（2.1）和非常高（8.3）。这些风险类别还预测了住院死亡率：低（0.0%）、中等（1.9%）、高（8.7%）和非常高（18.2%）在外部验证队列中。对于其他指标，如住院时间、ICU使用、机械通气和血管活性药物使用，也观察到了类似的趋势。 结论： Sepsis ImmunoScore在识别和预测败血症和危重疾病指标方面显示出高准确性，这可能使医生能够及时识别出败血症和不良结果高风险的患者，从而可能改善临床决策和患者结果。（由国防威胁降低局和其他人资助。）

NEJM AI, Volume 1 No. 12 December 2024

译文来自于AI工具Kimi