This retrospective cohort study quantified interruptions in scheduled antiseizure medication (ASM) dosing that occur during the ICU-to-floor transfer. The authors used the MIMIC‑IV (version 3.1) clinical database to identify admissions with epilepsy or status epilepticus that included an ICU stay followed by transfer to a hospital floor, and where a scheduled ASM order remained active at the time of ICU departure. The study sought to measure how often ASM administration gaps occur across the transfer and whether gap frequency is higher than during matched non‑transfer intervals in the same patient and for the same drug.
Data were drawn from MIMIC‑IV (v3.1). Eligible observations required an ICU stay followed by a floor transfer, a scheduled ASM order active at ICU departure, and sufficient medication administration records to define dosing intervals. The analysis included 2,469 transition-by-drug observations originating from 1,583 admissions and 1,335 unique patients.
An administration gap was defined as either: (1) an interval exceeding 1.5 times the expected dosing interval between the last ICU dose and the first floor dose, or (2) no additional dose recorded before discharge. Each transfer interval was compared with a matched non‑transfer control interval drawn from the same patient and the same ASM drug, enabling a within‑patient paired comparison that controls for individual dosing schedules and patient factors.
The primary paired comparisons used the McNemar test to assess differences in gap frequency between transfer and matched non‑transfer intervals in the same patient and drug. Confidence intervals for proportions and paired differences were reported. A prespecified multivariable model evaluated six clinical predictors for association with gap occurrence; sociodemographic variables were summarized descriptively. The authors applied correction for multiple comparisons when assessing predictors.
Across the 2,469 ASM transition observations, an administration gap occurred in 251 cases, representing 10.2% of observations (95% CI, 8.7%–11.7%). When compared with matched non‑transfer control intervals in the same patient and drug, gaps were significantly more frequent across the transfer. Specifically, the paired rate difference was 5.8 percentage points before transfer comparisons (7.5% vs 1.7%; P = 7.3 × 10^-22) and 6.4 percentage points in after‑transfer comparisons (8.9% vs 2.5%; P = 1.9 × 10^-23). These paired differences indicate a consistent excess in gap risk associated with the ICU‑to‑floor transition.
The authors compared gap rates between ASMs that have an intravenous formulation available and those that are oral‑only. Gap rates were 9.9% for IV‑available drugs and 11.4% for oral‑only drugs. The observed difference (1.5 percentage points) did not reach statistical significance (95% CI, −1.6 to 4.5; P = .34), indicating similar interruption risk across formulation types in this dataset.
Six prespecified clinical predictors were evaluated in a multivariable model to identify factors associated with administration gaps. After correction for multiple comparisons, none of these predictors reached statistical significance. The authors therefore did not attribute the excess gap frequency to any single medication characteristic or measured patient factor in their model.
The study found that an ASM administration gap occurred in roughly one in ten drug‑transition observations at the ICU‑to‑floor transfer, and that this rate exceeded matched non‑transfer gap rates by approximately 5.8 to 6.4 percentage points. Because the excess risk appears linked to the transfer process itself rather than a single drug or patient characteristic, the authors advocate for a structured medication‑continuity check at the time of ICU discharge to the floor. Such a check is proposed as a systems‑level intervention to reduce preventable interruptions in ASM therapy during care transitions.
All analyzed data derive from the publicly available MIMIC‑IV (v3.1) resource, accessible to credentialed users through PhysioNet under the PhysioNet data use agreement after completion of required human‑subjects research training. Analysis code for this study is available at the authors' GitHub repository: https://github.com/Alon-Gorenshtein/ASM_Continuity. The Institutional Review Board at Beth Israel Deaconess Medical Center waived ethical approval for this secondary analysis of deidentified data; the MIMIC‑IV resource itself was created under IRB review with a waiver of informed consent.