Early Prediction of ICU Mortality in Patients with Acute Hypoxemic Respiratory Failure Using Machine Learning: The MEMORIAL Study
Bibliographic record
Abstract
Background: Early prediction of ICU death in acute hypoxemic respiratory failure (AHRF) could inform clinicians for targeting therapies to reduce harm and increase survival. We sought to determine clinical modifiable and non-modifiable features during the first 24 h of AHRF associated with ICU death. Methods: This is a development, testing, and validation study using data from a prospective, multicenter, nation-based, observational cohort of 1241 patients with AHRF (defined as PaO2/FiO2 ≤ 300 mmHg on mechanical ventilation [MV] with positive end-expiratory pressure [PEEP] ≥ 5 cmH2O and FiO2 ≥ 0.3) from any etiology. Using relevant features captured at AHRF diagnosis and within 24 h, we developed a logistic regression model following variable selection by genetic algorithm and machine learning (ML) approaches. Results: We analyzed 1193 patients, after excluding 48 patients with no data at 24 h after AHRF diagnosis. Using repeated random sampling, we selected 75% (n = 900) for model development and testing, and 25% (n = 293) for final validation. Risk modeling identified six major predictors of ICU death, including patient’s age, and values at 24 h of PEEP, FiO2, plateau pressure, tidal volume, and number of extrapulmonary organ failures. Performance with ML methods was similar to logistic regression and achieved a high area under the receiver operating characteristic curve (AUROC) of 0.88, 95%CI 0.86–0.90. Validation confirmed adequate model performance (AUROC 0.83, 95%CI 0.78–0.88). Conclusions: ML and traditional methods led to an encouraging model to predict ICU death in ventilated AHRF as early as 24 h after diagnosis. More research is needed to identify modifiable factors to prevent ICU deaths.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.007 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".