Scholarlink

Background <p>The Hypotension Prediction Index (HPI) is a machine learning algorithm that was designed to detect alterations in the cardiovascular physiological control mechanisms that may lead to hypotension through multidimensional analyses of the physiological interactions of core haemodynamic parameters. Hypotension is a physiological condition where the central perfusion pressure in the cardiovascular system is too low to ensure adequate flow regulation of local organs and its threshold is the pressure at the lower limit of autoregulation. As such, hypotension can occur in a wide pressure range that can exceed considerably the conventional MAP threshold of 65&#xa0;mmHg and can vary between individuals as well as within the same individual. We hypothesize that HPI is an indicator of haemodynamic instability. We used direct physiological signs of haemodynamic instability and hypoperfusion from the cardiovascular haemodynamic parameters available on the HPI technology. We conducted a retrospective analysis to investigate the performance of HPI as an indicator of haemodynamic instability related to hypotension and hypoperfusion.</p> Methods <p>A dataset of 1,683 cardiac and non-cardiac surgical patients were analyzed, including 871 patients monitored with invasive arterial line and 812 monitored with non-invasive finger cuff. There were 19,685 and 21,097 HPI alerts at the threshold of 85 and 50, respectively. Haemodynamic parameters used to define instability were MAP &lt; 65&#xa0;mmHg, SVV ≥ 13%, CI ≤ 2 L·min<sup>−1</sup>·m<sup>−2</sup>, and SVR ≤ 800 dyn·s·cm<sup>−5</sup>.</p> Results <p>In the invasive A-line dataset, 94.6% of the HPI alerts at threshold of 85 are followed in 15&#xa0;min by haemodynamic instability, and 91.4% of HPI alerts at threshold 50 are followed by haemodynamic instability. In the non-invasive finger cuff dataset, they are 96.2% and 94.1% for HPI alerts at threshold 85 and 50, respectively.</p> Conclusion <p>The Hypotension Prediction Index predicts the upcoming hypotensive haemodynamic instability events with high accuracy. Further research is needed to investigate whether HPI alerts guided treatment of haemodynamic instability improves patient outcomes.</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

Performance of the hypotension prediction index as an indicator of hypotensive haemodynamic instability in surgical patients

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月25日

查看全文
  • Simon Davies,
  • Sai Buddi,
  • Zhongping Jian,
  • Neal W. Fleming,
  • Maxime Cannesson,
  • Michael Sander,
  • Denise P. Veelo,
  • Alexander P. J. Vlaar,
  • Thomas W. L. Scheeren,
  • Monty Mythen,
  • Feras Hatib

摘要

Background

The Hypotension Prediction Index (HPI) is a machine learning algorithm that was designed to detect alterations in the cardiovascular physiological control mechanisms that may lead to hypotension through multidimensional analyses of the physiological interactions of core haemodynamic parameters. Hypotension is a physiological condition where the central perfusion pressure in the cardiovascular system is too low to ensure adequate flow regulation of local organs and its threshold is the pressure at the lower limit of autoregulation. As such, hypotension can occur in a wide pressure range that can exceed considerably the conventional MAP threshold of 65 mmHg and can vary between individuals as well as within the same individual. We hypothesize that HPI is an indicator of haemodynamic instability. We used direct physiological signs of haemodynamic instability and hypoperfusion from the cardiovascular haemodynamic parameters available on the HPI technology. We conducted a retrospective analysis to investigate the performance of HPI as an indicator of haemodynamic instability related to hypotension and hypoperfusion.

Methods

A dataset of 1,683 cardiac and non-cardiac surgical patients were analyzed, including 871 patients monitored with invasive arterial line and 812 monitored with non-invasive finger cuff. There were 19,685 and 21,097 HPI alerts at the threshold of 85 and 50, respectively. Haemodynamic parameters used to define instability were MAP < 65 mmHg, SVV ≥ 13%, CI ≤ 2 L·min−1·m−2, and SVR ≤ 800 dyn·s·cm−5.

Results

In the invasive A-line dataset, 94.6% of the HPI alerts at threshold of 85 are followed in 15 min by haemodynamic instability, and 91.4% of HPI alerts at threshold 50 are followed by haemodynamic instability. In the non-invasive finger cuff dataset, they are 96.2% and 94.1% for HPI alerts at threshold 85 and 50, respectively.

Conclusion

The Hypotension Prediction Index predicts the upcoming hypotensive haemodynamic instability events with high accuracy. Further research is needed to investigate whether HPI alerts guided treatment of haemodynamic instability improves patient outcomes.