Measurement of intraoperative hypotension exposure and ICU resource-use associations: a retrospective methodological cohort study using VitalDB

This is the authors' abstract. We don't add key points for this paper.

Intraoperative hypotension (IOH) exposure depends on monitoring source, sampling, duration, signal quality and observation window. We examined these measurement choices and the distinction between exposure–outcome association and incremental prediction of intensive care unit (ICU) resource use. Primary analyses used operative-window VitalDB exposures and patient-grouped prediction models. Exposures used invasive arterial pressure (ART) or non-invasive blood pressure (NIBP). Exploratory analyses evaluated sparse ART sampling, continuous low-pressure episodes, effective coverage, directional ART/NIBP discordance and anesthesia-window exposure. A waveform-available ART subset underwent a fixed technical-quality screen, not a clinically validated artifact classifier. Three separate ART severity, duration and deficit-burden association models used covariate adjustment and patient-cluster confidence intervals (CIs). Prediction was assessed using area under the receiver operating characteristic curve (AUROC). The cohort contained 5825 cases from 5587 patients; 3510 used ART and 2315 NIBP. ART sampling every 180 s yielded MAP < 55 mmHg prevalence of 20.1–21.3%, compared with 60.7% in full operative records. Among 3485 ART cases with valid anesthesia bounds, extending the window increased any MAP < 55 prevalence from 60.7% to 97.6%; continuous episodes ≥ 1 min occurred in 31.9%, and median supported coverage was 86.6%. In 3407 cases with positive support before and after waveform screening, supported MAP < 55 prevalence was 60.8% before and 58.8% after screening. In 3230 complete ART cases, doubling (1 + cumulative minutes < 55) had adjusted ICU-use odds ratio 1.31 (95% CI 1.20–1.44). The source-adjusted pooled prediction increment was 0.00020 AUROC (conditional 95% CI − 0.00038 to 0.00079). Only 44 NIBP cases had ICU use. IOH severity, duration and burden require explicit measurement rules. Broader observation windows and technical screening alter classification but do not ensure complete induction capture or artifact-free signals. Burden associations with ICU resource use do not establish treatment effects or clinically useful prediction. Sparse NIBP events and single-centre selection limit interpretation.