Outcome-blind acute metabolic displacement identifies QRS vulnerability and transient low-heart-rate susceptibility during hyperkalaemia

Abstract Background Electrocardiographic responses to hyperkalaemia vary substantially at similar potassium concentrations. We sought to determine whether acute displacement from a normal-potassium metabolic reference might identify the physiological context in which hyperkalaemia is associated with QRS slowing, and whether this construct could also mark acute heart-rate response or low-heart-rate burden. Methods We linked MIMIC-IV version 3.0 to MIMIC-IV-ECG and assembled normal-potassium–hyperkalaemia–normal-potassium ECG–laboratory trajectories. An acute metabolic displacement score (AMED) was calibrated without reference to electrophysiological outcomes in 14,400 complete normal-potassium reference trajectories, after excluding all hyperkalaemia-evaluation subjects. AMED was calculated as the shrinkage-covariance Mahalanobis distance of paired changes in creatinine, bicarbonate, and sodium, expressed per calibration standard deviation; potassium was excluded and adjusted for separately. The primary outcome in MIMIC-IV was immediate QRS change. In MIMIC-III, a separately locked post-result exploratory protocol analysed six-hour pre-event and post-event heart-rate windows. It distinguished absolute median-heart-rate response, time-weighted depth and duration below 50 beats/min, any post-event low-heart-rate exposure, and the original new sustained heart-rate-<50 endpoint. Episodes were retained, and inference was clustered by patient. Results The MIMIC-IV evaluation cohort comprised 1,342 complete episodes in 847 patients. Each standard deviation of AMED was associated with 0.85-ms greater immediate QRS prolongation (95% confidence interval [CI], 0.12–1.58; P = 0.022) and higher odds of dynamic QRS response (OR, 1.50; 95% CI, 1.25–1.81; false-discovery-rate P < 0.001). In MIMIC-III, 2,943 episodes from 2,409 patients had complete AMED and heart-rate windows. Each AMED standard deviation was associated with any post-event heart-rate-<50 exposure after adjustment for pre-event exposure (OR, 1.43; 95% CI, 1.14–1.80; P = 0.002; false-discovery-rate P = 0.014). The absolute median-heart-rate response was 0.42 beats/min greater per AMED standard deviation (95% CI, 0.04–0.80; nominal P = 0.029; Holm P = 0.057). AMED was not clearly associated with time-weighted low-heart-rate deficit change or the original new sustained heart-rate-<50 endpoint. Conclusions Outcome-blind non-potassium metabolic displacement was associated with greater QRS vulnerability and a higher probability of transient entry into a low-heart-rate state during hyperkalaemia. The pattern observed in MIMIC-III did not suggest greater cumulative or sustained bradycardic burden, which may help distinguish threshold-entry susceptibility from persistence. These findings suggest that AMED represents a research phenotype rather than a validated bedside predictor.