Scholarlink

<p>Venous thromboembolism (VTE), including deep vein thrombosis (DVT) and pulmonary embolism (PE), is a significant complication in surgical patients. Artificial intelligence (AI) and machine learning (ML) may enhance VTE risk stratification by leveraging novel parameters in electronic medical records (EMRs). We aim to systematically review the role of AI and ML in stratifying the risk of postoperative VTE in surgical patients. A systematic literature search was conducted in PubMed, Scopus, and CINAHL Complete through May 24, 2024, and registered in PROSPERO (CRD420250522393). Eligible studies included primary research on postoperative adult patients undergoing any surgery that reported VTE risk assessment using the area under the receiver operating characteristic curve (AUC). Risk of bias was assessed, and the Wilcoxon signed-rank test was used to compare AUCs. 34 studies met the inclusion criteria, of which 22 directly compared AI models with non-AI models and were used for analysis. AI models showed higher discrimination than non-AI comparators within the same cohorts, with a median ΔAUC of + 0.10 (IQR 0.03–0.21; Wilcoxon signed-rank test, <i>n</i> = 22, <i>p</i> &lt; 0.001). 31 studies were found to have a high risk of bias in model development, with only 12 reporting calibration metrics. Studies were limited to the United States (17) and China (17), with substantial surgical heterogeneity. Across head-to-head studies, AI models consistently showed higher discrimination than their non-AI comparators within the same cohorts. Future research should prioritize external validation, standardized reporting of calibration metrics, and generalizability.</p> Graphical abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Artificial intelligence for venous thromboembolism risk stratification in surgical patients: a systematic review

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月6日

查看全文
  • Kavin Shah,
  • Michael Gadelrab,
  • Emily A. Brennan,
  • Maggie L. Westfal,
  • Colleen A. Donahue,
  • John Del Gaizo,
  • Arman Kilic,
  • Thomas Curran

摘要

Venous thromboembolism (VTE), including deep vein thrombosis (DVT) and pulmonary embolism (PE), is a significant complication in surgical patients. Artificial intelligence (AI) and machine learning (ML) may enhance VTE risk stratification by leveraging novel parameters in electronic medical records (EMRs). We aim to systematically review the role of AI and ML in stratifying the risk of postoperative VTE in surgical patients. A systematic literature search was conducted in PubMed, Scopus, and CINAHL Complete through May 24, 2024, and registered in PROSPERO (CRD420250522393). Eligible studies included primary research on postoperative adult patients undergoing any surgery that reported VTE risk assessment using the area under the receiver operating characteristic curve (AUC). Risk of bias was assessed, and the Wilcoxon signed-rank test was used to compare AUCs. 34 studies met the inclusion criteria, of which 22 directly compared AI models with non-AI models and were used for analysis. AI models showed higher discrimination than non-AI comparators within the same cohorts, with a median ΔAUC of + 0.10 (IQR 0.03–0.21; Wilcoxon signed-rank test, n = 22, p < 0.001). 31 studies were found to have a high risk of bias in model development, with only 12 reporting calibration metrics. Studies were limited to the United States (17) and China (17), with substantial surgical heterogeneity. Across head-to-head studies, AI models consistently showed higher discrimination than their non-AI comparators within the same cohorts. Future research should prioritize external validation, standardized reporting of calibration metrics, and generalizability.

Graphical abstract