Hypotension prediction index: comparison between invasive and non-invasive pressure inputs

The Hypotension Prediction Index (HPI) derived using an Acumen™ arterial pressure transducer can decrease the incidence of intraoperative hypotension and possibly decrease perioperative complications. The HPI can also be obtained from the ClearSight™ continuous non-invasive blood pressure monitor. Concurrent comparison of HPI values obtained from these two pressure inputs is limited and additional comparisons could increase clinician confidence in non-invasive hemodynamic monitoring, expand its applications and improve patient outcomes.

Simultaneous hemodynamics were recorded using two HemoSphere monitors with the HPI software and either intra-arterial Acumen^® (hereinafter invasive) or ClearSight™ (hereinafter non-invasive) pressure inputs. Data collected from the non-invasive system was compared to corresponding invasive, intra-arterial data using Bland-Altman analysis, Spearman correlation, concordance analysis and relative performance with respect to prediction of hypotensive events using ROC analysis and HPI alerts agreement analysis.

6,862 paired data points were available from 36 patients. Bland-Altman comparisons demonstrated a bias of -8.4 (± 23) with limits of agreement from - 53 to 36. The correlation between HPI values was strong with an r value of 0.76 (95%CI:0.75-0.77). Concordance was also strong at 74% (10% exclusion zone). Using ROC analysis, the AUC for prediction of hypotension was similar and at 5 min was 0.883 [0.786,0.953] for the invasive pressure and 0.860 [0.770,0.939] for the non-invasive pressure inputs. At the same time points, the agreement between HPI alerts was high with an accuracy of 86.3%.

HPI values and predictive performance were comparable when derived from either invasive or non-invasive pressure inputs.

The study was approved by the UC Davis Institutional Review Board (IRB #1791102-1) and registered on ClinicalTrials.gov (NCT05025176) before the enrollment of the first patient.