Joint associations of timestamp-aligned early ICU shock index and glucose with 28-day mortality in critically Ill patients with atrial fibrillation: a MIMIC-IV cohort study

Abstract Background Early bedside risk characterization in critically ill patients with atrial fibrillation is challenging because hemodynamic instability and metabolic stress may coexist and evolve rapidly after ICU admission. Shock index and glucose are routinely available, but their joint interpretation requires careful attention to measurement timing. Methods We conducted a retrospective cohort study using the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 3.1) database. Adult ICU patients with atrial fibrillation were included. Timestamp-aligned exposure profiles were reconstructed from source records: shock index was calculated only from heart rate and systolic blood pressure recorded at the same chart time, and the nearest glucose measurement within ± 6 h was matched to the synchronized shock-index time point. The exposure index time was the later timestamp of the synchronized shock-index and matched glucose measurements. Cox models used delayed entry at the index time. The primary temporally ordered model adjusted for age and sex; expanded pre-index laboratory and SOFA- and APSIII-adjusted models were evaluated as sensitivity analyses. Results Among 1,635 eligible ICU admissions with atrial fibrillation, 925 patients had a complete timestamp-aligned profile using the primary ± 6-h window; 209 (22.6%) died within 28 days. The median index time was 195.0 min after ICU admission (IQR, 96.0–374.6), and no death occurred before profile ascertainment. In the primary delayed-entry model, synchronized shock index was associated with mortality (HR per SD increase, 1.27; 95% CI, 1.13–1.43; P < 0.001), whereas matched glucose showed a weaker association (HR per SD increase, 1.10; 95% CI, 0.98–1.23; P = 0.112). The shock-index estimate remained similar in an expanded pre-index laboratory complete-case model (HR, 1.26; 95% CI, 1.06–1.49; P = 0.009) but was attenuated in SOFA- and APSIII-adjusted sensitivity analyses. Joint categories differed descriptively, but the high-SI/high-glucose group did not exceed the high-SI/low-glucose group, and interaction analyses did not support positive synergy. Conclusions After aligning exposure ascertainment with the start of risk estimation, synchronized early ICU shock index was associated with 28-day mortality in the primary pre-index model, whereas matched glucose showed weaker adjusted associations. Joint shock-index and glucose categories were exploratory descriptive groupings and did not establish incremental risk stratification or positive biological or statistical interaction.