1 No NEWS is good NEWS – a machine learning approach to improve physiological early warning scoring
Bibliographic record
Abstract
Background Current Early Warning Scoring (EWS) systems in clinical practice are threshold rules-based systems using physiological data to highlight patients at risk of impending in-hospital death. Examples include the National Early Warning Score (NEWS2) and the Electronic Cardiac Arrest Triage (eCART) score, the standards of care within the UK and the USA, respectively. The current EWS modelling framework has two limitations. Firstly, they consider risk at a single time point and, therefore, cannot consider trajectories. Secondly, they negate relational information between covariates by decomposing physiological signals into a single value. We propose using Long Short-Term Memory (LSTM) units, a Machine Learning (ML) technique that uses time-series modelling and neural networks to address these limitations and better utilise the available information. Methods We extracted the sequences of vital signs, NEWS2, and eCART values within a 72-hour observation window from MIMIC-IV, a dataset containing anonymised electronic healthcare records. We masked the last 24 hours for each sequence and trained multiple ML models to identify patients at risk of death. We compare the models’ discriminative ability using the metrics, recall, F1-score, AUROC and AUPRC and conducted 10-fold stratified cross-validation. Results The LSTM model has statistically significant performance advantages in F1-score ( 0.3391 ±0.0093), AUROC ( 0.7399 ±0.0119), and AUPRC ( 0.337 ±0.0212). The clinical benefit of such a model allows clinicians to correctly identify more at-risk patients without increasing the false alarm rate. We hypothesise that this ML pipeline can be used to predict other clinical outcomes of interest, such as a requirement for escalation of care/additional organ support, which is the focus of future work. Conclusions Our results demonstrate that the ML time-series framework can utilise trajectory information to give further context and significantly improve the prediction of impending death. We propose any future EWS systems should incorporate physiological time-series.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 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 teacher head, 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".