Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1600
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-16T18:13:46.877303Z
status: published
lens: g_space
sector: health
headline: Development and Validation of a Machine Learning Model for Predicting 28-Day Mortality in Patients with Sepsis-Associated Encephalopathy and Heart Failure
dek: IntroductionPatients with sepsis-associated encephalopathy (SAE) and heart failure (HF) represent a clinically relevant but understudied subgroup in the intensive care unit, and early risk stratification may be important for guiding management. This study aimed to develop and temporally validate machine-learning models for predicting 28-day mortality in patients with SAE and HF.MethodsIn this retrospective cohort study, patients with SAE and HF were identified from the Medical Information Mart for Intensive Care…
gain_title: Models built on six routine variables predicted 28-day mortality in this ICU subgroup more accurately than SOFA, SAPS II and LODS, with logistic regression offering the most balanced performance.
problem_title: (none)
trace_subject: (none)
gain_reading: Models built on six routine variables predicted 28-day mortality in this ICU subgroup more accurately than SOFA, SAPS II and LODS, with logistic regression offering the most balanced performance.
gain_evidence: Machine-learning models and logistic regression outperformed conventional severity scores (SOFA, SAPS II, and LODS)
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers used records from two intensive care databases, MIMIC-IV and MIMIC-III, to identify 1249 patients who had both sepsis-associated encephalopathy and heart failure. They built five prediction models from six variables and tested them on held-out and temporally distinct patient groups.

The work suggests that early, individualized risk stratification for a poorly studied ICU subgroup may be feasible with simple models. The evidence comes from retrospective data, so it does not show that using the models improves patient care or survival.
limitation: The study is retrospective and based on MIMIC database records, and the source reports no testing of whether using the models changes clinical management or outcomes.
tag: Evidence-backed gain
key_points: The cohort had 1249 patients: 833 for training, 205 for internal testing and 211 for temporal testing. | Predictors were age, weight, respiratory rate, Logistic Organ Dysfunction System score, first 24-hour urine output and Charlson Comorbidity Index. | In the temporal test set, XGBoost reached an AUROC of 0.824 and logistic regression 0.816. | Logistic regression showed more stable calibration and the best overall balance across performance measures.
rundown: The study was a retrospective cohort analysis. Variables with more than 30% missing data were excluded and the remaining gaps were filled by multiple imputation. Candidate predictors were screened by univariable analysis, then by multivariable logistic regression.

Five model types were built: logistic regression, XGBoost, LightGBM, support vector machine and AdaBoost. They were evaluated on AUROC, classification metrics, calibration and decision-curve analysis.

The MIMIC-III cohort served as a temporally distinct test set. The authors conclude logistic regression may be the most clinically applicable model for individualized risk stratification.
sources:
- peer_reviewed | Journal of Intensive Care Medicine | https://doi.org/10.1177/08850666261495897 | 2026-10-09
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

e11fc49eaac15d55e0e6f7f2986c9bfccc726b12b2967b6d05fa6bf38cf829be

previous

0000000000000000000000000000000000000000000000000000000000000000