Recurrent early hypotension after continuous renal replacement therapy initiation and 28-day mortality: a MIMIC-IV landmark cohort study

Abstract Background Early hypotension after continuous renal replacement therapy (CRRT) initiation may reflect hemodynamic vulnerability, but prior studies mainly assessed first-hour events or single mean arterial pressure (MAP) values. We examined recurrent MAP < 65 mmHg during the first 6 h and its incremental information beyond pre-initiation clinical factors and simpler post-initiation MAP summaries. Methods We conducted a retrospective 6-hour landmark study using MIMIC-IV v3.1. Eligible patients were alive and remained in the index intensive care unit (ICU) at T0 + 6 h and had at least one valid MAP measurement during [T0, T0 + 6 h). Six T0-aligned hourly bins were classified as low when any valid MAP was < 65 mmHg. Exposure groups were no MAP < 65 mmHg (zero low bins), transient MAP < 65 mmHg (one low bin), and recurrent MAP < 65 mmHg (at least two low bins). The outcome was 28-day mortality counted from CRRT initiation (T0). Primary estimates were adjusted marginal risks, risk differences, and robust modified Poisson risk ratios. Exploratory paired 10-fold cross-validation assessed incremental information. Results Among 2,034 landmark-eligible patients (1,126 deaths), 924 had no, 401 transient, and 709 recurrent MAP < 65 mmHg. Adjusted mortality risks were 48.8% (95% confidence interval [CI], 45.7–52.0), 56.1% (51.6–60.5), and 64.3% (60.4–67.7), respectively. Risk differences versus no MAP < 65 mmHg were + 7.3% points (95% CI, 1.7–12.8) for transient and + 15.5 (10.2–20.5) for recurrent MAP < 65 mmHg. Modified Poisson risk ratios were 1.18 (95% CI, 1.06–1.32; p = 0.003) and 1.32 (1.20–1.46; p < 0.001). Adding the three-level pattern to the pre-initiation model yielded a small internally cross-validated discrimination increment (+ 0.010, 95% CI, 0.001–0.019), without clear gain beyond MAP nadir or mean-MAP summaries; the prespecified cumulative-duration comparison was not stably estimable. Conclusions Among patients who were alive and remained in the index ICU at T0 + 6 h and had valid exposure-window MAP data, transient and recurrent early hypotension were associated with higher adjusted 28-day mortality, with the highest risk in the recurrent group. These single-center findings lack independent external validation and do not establish causality, predictive superiority, clinical utility, a treatment target, or a basis for management changes.