D93-02 Impact of ICU-PAUSE on ICU Readmissions and Hospital Length of Stay: A Multi-site Interrupted Time Series Analysis

Abstract Rationale Communication failures during patient handoffs from the ICU to the ward are common and harmful. The ICU-PAUSE ICU-ward handoff bundle - a structured communication tool, educational materials, and implementation guide - has been implemented at 20+ United States hospitals with positive perceptual and implementation outcomes including reach and sustainability. It is not yet known, however, whether real-world implementation improves patient outcomes. We therefore evaluated the association between ICU-PAUSE implementation and ICU readmissions, hospital length of stay (LOS), and 30-day mortality. Methods Using deidentified patient-level data from the Vizient(R) Clinical Database, we conducted a quasi-experimental, multi-site interrupted time series (ITS) analysis at nine U.S. hospitals with ≥1 year of data before and after ICU-PAUSE implementation. We included adult medical ICU admissions that resulted in discharge to an acute care ward or intermediate care unit. The primary outcome was any ICU readmission before hospital discharge, indexed at the initial ICU-to-ward transfer. Secondary outcomes were 30-day in-hospital mortality and post-ICU LOS (days). To account for secular trends and detect associations between intervention and changes in outcomes over time, we fit linear segmented regressions on monthly aggregated outcome data, with fixed effects for time, post-implementation level change, post-implementation time trend change, and Newey-West standard errors. Results The 9-hospital cohort included 40,990 admissions with one or more medical ICU-to-ward transfer (19,985 pre-intervention, 21,005 post-intervention). Patients were demographically similar before and after the intervention. The ICU readmission rate was 15% pre- (n = 3,014) and post-implementation (n = 3,262, p = 0.19). Crude 30-day in-hospital mortality was 15% (n = 2,991) pre-implementation and 14% (n = 2,944) post-implementation (p = 0.006). In ITS models, ICU-PAUSE implementation was not associated with a change in ICU readmission rate level (p = 0.14) or trend (p = 0.30) but was associated with a significant reversal in the 30-day mortality trend (-0.28% per month change in trend, 95% CI -0.45 to -0.1, p = 0.001) and an immediate 1.05 day (95% CI -1.64 to -0.47, p 0.001) reduction in post-ICU LOS. Conclusions In this multi-site ITS analysis, preliminary results indicate that ICU-PAUSE implementation was not associated with unadjusted ICU readmission rates, but we observed associations with unadjusted 30-day in-hospital mortality and post-ICU LOS. ITS deals effectively with time trends, and while we plan adjusted analyses using hierarchical methods, it is unlikely that patient mix changed substantially over the study period, making our results plausible. This abstract is funded by: None