Development and validation of the Hypotensive Exposure Duration Index for mortality risk prediction in critically ill patients

Comprehensive Summary

Ding et al. developed and validated the Hypotensive Exposure Duration Index (HEDI) to quantify cumulative hypotensive burden and evaluate its association with ICU mortality. Using minute-by-minute invasive mean arterial pressure (MAP) data from 11,059 adult ICU admissions in the Salzburg Intensive Care Database, investigators constructed odds ratio deviation heatmaps across MAP thresholds (52-120 mmHg) and exposure durations (5 minutes to 5 hours). HEDI was calculated by integrating patient-specific hypotensive exposure frequencies weighted by mortality-associated odds ratios and normalized across all threshold-duration combinations. Non-survivors demonstrated significantly higher HEDI values than survivors (0.47 vs. -0.15, p<0.001). HEDI's discriminative ability for ICU mortality improved over time, with AUC values increasing from 0.624 at 24 hours to 0.700 at 72 hours after ICU admission. Machine learning models incorporating HEDI demonstrated robust predictive performance, with the best model achieving a test AUC of 0.843. External validation using the eICU database confirmed HEDI's prognostic performance across different patient populations, including those with and without vasopressor use.

Outcomes and Implications

Current approaches to hypotension monitoring in the ICU typically rely on single MAP thresholds (commonly 65 mmHg) or time-weighted averages, but these methods fail to capture the complex relationship between the depth, duration, and cumulative burden of hypotensive episodes. HEDI addresses this gap by integrating both the intensity and duration of hypotension into a single metric that reflects mortality risk. This is clinically important because not all hypotensive episodes carry equal risk; brief, mild hypotension likely differs from prolonged, severe hypotension in its impact on organ perfusion and outcomes. The finding that HEDI's predictive performance improves over the first 72 hours of ICU admission suggests it captures cumulative physiologic injury that unfolds over time rather than just initial severity. The robust performance in machine learning models (AUC 0.843) and successful external validation across different patient populations, including those with and without vasopressors, supports HEDI's generalizability. For critical care clinicians, HEDI could potentially serve as a real-time risk stratification tool to identify patients accumulating dangerous hypotensive burden who may benefit from more aggressive hemodynamic management. However, several important limitations and questions remain. The study's retrospective design cannot establish whether HEDI-guided interventions would improve outcomes, only that HEDI associates with mortality. The optimal HEDI threshold for triggering interventions is unclear. Implementation requires minute-by-minute invasive arterial monitoring and computational infrastructure to calculate HEDI in real-time, which may limit applicability in resource-constrained settings or in patients without arterial lines. The study does not address whether HEDI adds value beyond existing severity scores like APACHE or SOFA, or whether it helps guide specific interventions. Future prospective studies should examine whether HEDI-guided hemodynamic protocols reduce mortality compared to standard care, identify optimal intervention thresholds, and determine which patient subgroups benefit most from HEDI monitoring. Until then, HEDI represents a promising tool for quantifying hypotensive burden but requires validation as a clinical decision support tool before routine implementation.