Previous-day physical restraint duration as a bedside signal for following-day delirium: a retrospective patient-day analysis in a mixed ICU.

Abstract

Background: Delirium is common in the intensive care unit and associated with adverse outcomes. However, bedside indicators that flag patients at high risk for delirium the following day are lacking. Although physical restraint has been linked to delirium, many prior studies relied on cross-sectional, same-day assessments that cannot exclude reverse causation. This study aimed to evaluate whether cumulative previous-day physical restraint duration was a bedside signal for following-day delirium.

Methods: We conducted a single-center retrospective observational cohort study using patient-days (00:00-24:00 h) as the unit of analysis, with the exposure and covariates derived from more granular, time-stamped nursing records aggregated into calendar-day units. The exposure was cumulative physical restraint duration on the previous day, reported per 8-h increase. The outcome was incident delirium on the following day, defined as an Intensive Care Delirium Screening Checklist (ICDSC) score ≥ 4. The primary analysis was restricted to patient-days without previous-day delirium (ICDSC 0-3). We fitted a generalized linear mixed model with a random intercept for each patient. The restraint duration was modeled using a natural cubic spline (degrees of freedom = 3).

Results: Of 1,482 patient-days from 281 patients, 263 patient-days were excluded due to coma. The primary analysis included 729 patient-days from 247 patients without previous-day delirium (115 incident delirium events, 15.8%). The association between restraint duration and following-day incident delirium was non-linear (likelihood ratio test, P < 0.001). Using no restraint as the reference, the odds ratios were 2.43 (95% confidence interval [CI], 1.64-3.61) at 8 h, 5.16 (95% CI 3.39-7.85) at 16 h, and 10.01 (95% CI 6.31-15.88) at 24 h. Sensitivity analyses using complete-case analysis, last observation carried forward, best-case/worst-case extreme scenarios, and adjustment for previous-day mental status yielded consistent results.

Conclusions: In this single-center cohort, longer restraint duration on the previous day was associated with a higher risk of incident delirium the following day and may serve as a bedside-observable signal rather than evidence of a causal effect. Residual confounding may remain; prospective observational studies are needed to confirm this signal before its clinical utility is evaluated.