Although intraoperative hypotension exposure is associated with increased intensive care unit resource use, it yields virtually no incremental value for predicting intensive care unit admission, according to a retrospective methodological cohort study analyzing surgical cases from the VitalDB database.
The investigation evaluated 5,825 surgical cases across 5,587 patients to examine how measurement choices—including monitoring modality, sampling intervals, exposure duration, signal quality, and observation windows—affect hypotension classification and outcomes. Among the analyzed procedures, 3,510 cases utilized invasive arterial pressure monitoring, while 2,315 cases relied on non-invasive blood pressure monitoring. Primary analyses utilized operative-window exposures alongside patient-grouped prediction models.
The authors observed that sampling frequency dramatically influenced recorded hypotension rates. When invasive arterial pressure was sampled sparsely at 180-second intervals, the prevalence of mean arterial pressure below 55 mmHg ranged from 20.1% to 21.3%. By contrast, continuous measurement across full operative records demonstrated a mean arterial pressure below 55 mmHg prevalence of 60.7%.
Shifting the observation window and applying signal processing filters also altered exposure estimates. Among 3,485 invasive arterial pressure cases with valid anesthesia bounds, expanding the observation window from the operative period to the broader anesthesia window increased the prevalence of any mean arterial pressure below 55 mmHg from 60.7% to 97.6%. Within this extended window, continuous episodes lasting at least one minute occurred in 31.9% of cases, with a median supported coverage of 86.6%. Furthermore, applying a fixed technical-quality screen to 3,407 cases with positive support adjusted the supported hypotension prevalence from 60.8% before screening to 58.8% after screening.
In association analyses of 3,230 complete invasive arterial pressure cases using covariate adjustment and patient-cluster confidence intervals, doubling cumulative minutes with mean arterial pressure below 55 mmHg was associated with an adjusted odds ratio of 1.31 for intensive care unit resource use, with a 95% confidence interval of 1.20 to 1.44. However, assessing discrimination using the area under the receiver operating characteristic curve revealed that the source-adjusted pooled prediction increment for intensive care unit use was only 0.00020, with a conditional 95% confidence interval ranging from −0.00038 to 0.00079. In the non-invasive monitoring cohort, only 44 cases involved intensive care unit resource use.
The authors concluded that quantifying intraoperative hypotension severity, duration, and deficit burden requires explicit measurement rules. While broader observation windows and technical quality screens alter case classification, they do not guarantee complete induction capture or artifact-free signals. Additionally, statistical associations between hypotension burden and intensive care unit resource use do not demonstrate treatment effects or provide clinically useful prediction, and findings remain constrained by single-centre selection and sparse events among non-invasive monitoring cases.
The study design also differentiated specific dimensions of low blood pressure by evaluating three separate arterial association models rather than relying on a single exposure definition.
Subject of Research: Measurement and predictive value of intraoperative hypotension for ICU resource use
Article Title: Measurement of intraoperative hypotension exposure and ICU resource-use associations: a retrospective methodological cohort study using VitalDB
Article References: Wang, S., Zhang, X., & Yin, L. (2026). Measurement of intraoperative hypotension exposure and ICU resource-use associations: a retrospective methodological cohort study using VitalDB. BMC Anesthesiology. https://doi.org/10.1186/s12871-026-04430-6
Image Credits: AI Generated
DOI: 10.1186/s12871-026-04430-6
Keywords: intraoperative hypotension, VitalDB, invasive arterial pressure, non-invasive blood pressure, intensive care unit resource use, mean arterial pressure, area under the receiver operating characteristic curve, retrospective cohort study, hemodynamic monitoring, exposure measurement, waveform screening, sampling frequency, observation windows, deficit burden, patient-grouped prediction models, covariate adjustment
