Researchers developed a simple, interpretable nomogram to predict sepsis-associated encephalopathy (SAE) on day 1 in septic ICU patients who also have acute kidney injury (AKI), using data from the large MIMIC-IV database (2008–2022).

This retrospective study focused on adult patients meeting KDIGO criteria for AKI, with SAE defined as any occurrence during the ICU stay. After processing with multiple imputation and applying LASSO regression with cross-validation on 44 baseline variables from the first 24 hours post-admission, the model was narrowed to just four easily obtainable predictors: age, SAPS II score, serum sodium, and mean arterial pressure (MAP).

In a cohort of 6,780 ICU stays (split into training and validation sets), SAE occurred in nearly 70% of cases. The nomogram demonstrated solid performance, with AUC values around 0.73–0.74 for discrimination, excellent calibration, and superior clinical net benefit compared to a more complex XGBoost model (particularly in calibration and decision-curve analysis across relevant risk thresholds).

SHAP analysis provided insights into the model’s interpretability, revealing key patterns such as increasing risk with mild hypernatremia (linear rise in sodium 138–144 mmol/L), elevated risk in older age (especially above ~70 years), and a U-shaped relationship with MAP (optimal protection around 55–75 mm Hg).

These findings underscore potential modifiable factors—like correcting sodium levels and tailoring blood pressure targets—to mitigate SAE risk, while highlighting interconnected kidney-brain-circulatory dynamics in sepsis pathophysiology.

The authors emphasize the tool’s practicality for real-time bedside application and call for prospective multicenter validation plus integration into clinical decision-support systems to promote standardized, precise brain protection strategies in septic care.

Read More