Background and objectives

Sepsis is a life-threatening condition that is associated with increased mortality. Artificial intelligence tools can inform clinical decision making by flagging patients at risk of developing infection and subsequent sepsis. This systematic review aims to identify the optimal set of predictors used to train machine learning algorithms to predict the likelihood of an infection and subsequent sepsis.

Methods

This systematic review was registered in PROSPERO database (CRD42020158685). We conducted a systematic literature review across 3 large databases: Medline, Cumulative Index of Nursing and Allied Health Literature, and Embase. Quantitative primary research studies that focused on sepsis prediction associated with bacterial infection in adults in all care settings were eligible for inclusion.

Results

Seventeen articles met our inclusion criteria. We identified 194 predictors that were used to train machine learning algorithms, with 13 predictors used on average across all included studies. The most prevalent predictors included age, gender, smoking, alcohol intake, heart rate, blood pressure, lactate level, cardiovascular disease, endocrine disease, cancer, chronic kidney disease (eGFR<60 mL/min), white blood cell count, liver dysfunction, surgical approach (open or minimally invasive), and pre-operative haematocrit < 30 %. All included studies used artificial intelligence techniques, with average sensitivity 75.7 ± 17.88, and average specificity 63.08 ± 22.01.

Conclusion

The type of predictors influenced the predictive power and predictive timeframe of the developed machine learning algorithm. Predicting the likelihood of sepsis through artificial intelligence can help concentrate finite resources to those patients who are most at risk. Future studies should focus on developing more sensitive and specific algorithms.

中文翻译：

---

预防败血症；人工智能如何告知临床决策过程？系统评价

背景和目标

败血症是威胁生命的疾病，与死亡率增加有关。人工智能工具可以通过标记存在感染和继发败血症风险的患者，为临床决策提供依据。该系统综述旨在确定用于训练机器学习算法以预测感染和随后的败血症可能性的最佳预测因子集。

方法

该系统评价已在PROSPERO数据库（CRD42020158685）中注册。我们在3个大型数据库中进行了系统的文献综述：Medline，护理和相关健康文献的累积索引以及Embase。在所有护理环境中，以成年人细菌感染相关的败血症预测为重点的定量基础研究均符合纳入条件。

结果

符合我们纳入标准的文章有17篇。我们确定了194个用于训练机器学习算法的预测器，其中所有纳入研究平均使用了13个预测器。最普遍的预测因素包括年龄，性别，吸烟，饮酒，心率，血压，乳酸水平，心血管疾病，内分泌疾病，癌症，慢性肾脏病（eGFR <60 mL / min），白细胞计数，肝功能障碍，手术方式（开放或微创）和术前血细胞比容<30％。所有纳入的研究均使用人工智能技术，平均灵敏度为75.7±17.88，平均特异性为63.08±22.01。

结论

预测器的类型影响了已开发的机器学习算法的预测能力和预测时间范围。通过人工智能预测败血症的可能性可以帮助将有限的资源集中到那些风险最大的患者身上。未来的研究应集中在开发更敏感和更具体的算法上。