First aggression as a planning signal in post-acute stroke rehabilitation: Service benchmarks, goal setting, and a 60-day precaution window from a comprehensive observational cohort.

Abstract

Background: Aggressive behavior after stroke is common and linked to poorer outcomes, yet prior studies have not translated this knowledge into operational benchmarks that guide planning (time-limited risk windows, realistic goals, discharge timelines) during postacute rehabilitation.

Objective: To quantify service benchmarks after the first electronic health record (EHR)-documented aggression and map them to a standardized, time-boxed response: event rates and short-horizon risks (30/60 days), functional/length-of-stay (LOS) gaps in clinical units, and practical targets for planning.

Design: Retrospective observational cohort with prespecified 1:1 propensity-matched sensitivity analysis.

Setting: Single tertiary neurorehabilitation hospital (postacute, inpatient).

Patients: Adults with first-ever ischemic or hemorrhagic stroke admitted ≤3 months post stroke between 2005 and 2023 (N = 1683). Aggression was flagged at the first qualifying EHR annotation; 317 (18.8%) had more than one episode.

Outcome measure(s): Incidence densities (per 1000 patient-days) and derived 30-/60-day risks for first aggression, urinary tract infection, and falls; between-group differences in LOS and motor Functional Independence Measure (mFIM) outcomes (gain, efficiency, effectiveness) summarized as median differences (Hodges-Lehmann).

Results: Across 111,945 patient-days, first-aggression incidence was 2.83/1000 (95% CI: 2.53-3.16), implying 8.1% 30-day and 15.6% 60-day risks. Compared with nonaggressive patients, the aggressive group had higher urinary tract infection rates (rate ratio 1.92, 95% confidence interval [CI]: 1.43-2.57) and falls (rate ratio 1.28, 95% CI: 1.00-1.63). Functional gaps (median differences, Hodges-Lehmann) favored the nonaggressive group: LOS was longer (+17.9 days, 95% CI +13.8 to +22.0), mFIM efficiency and effectiveness were lower in the full cohort; in the matched subset (317 vs. 317), Hodges-Lehmann differences persisted (eg, LOS +9.8 days, 95% CI +4.1 to +15.2).

Conclusions: A single EHR-documented aggression provides a planning signal: event rates and 30-/60-day risks justify a 60-day precaution window, and scaled gaps support realistic goal setting (mFIM) and discharge date updates. These benchmarks offer a framework for bedside counseling, service targets, and audit; multicenter validation is warranted.