Machine learning to predict myocardial injury and death after non-cardiac surgery
Bibliographic record
Abstract
Abstract Background/Introduction Myocardial injury after non-cardiac surgery (MINS) is defined as prognostically relevant myocardial injury due to ischaemia that occurs within 30-days of non-cardiac surgery. Purpose The aim of this study was to test whether machine learning, using neural networks, can accurately predict this frequent and important complication. Methods Using data from 24,589 participants in the Vascular Events in Noncardiac Surgery Patients Cohort Evaluation (VISION) study, who had non-cardiac surgery and post-operative high-sensitivity troponin T (hs-TnT) levels measured, a deep neural network was trained to predict the primary outcome of MINS and the secondary outcome of death within 30-days. Validation was performed on a separate, randomly selected, subset of the study population with model discrimination and accuracy (number of correct predictions) determined. Results Using only data available pre-operatively, the deep neural network predicted MINS with an area under the receiver operating characteristic curve (AUROC) of 0.75 (95% confidence interval [95% CI] 0.74-0.76) and death at 30-days with an AUROC of 0.83 (95% CI 0.79-0.86). Addition of basic intra-operative and early post-operative data increased the AUROC for MINS to 0.77 (95% CI 0.76-0.78) and death to 0.87 (95% CI 0.85-0.90). The deep neural network trained on the full dataset (pre-operative, intra-operative and early post-operative) predicted MINS with an accuracy of 70% and death within 30-days with an accuracy of 89%. Conclusions Neural networks can be trained to predict MINS and death within 30-days of non-cardiac surgery and the inclusion of intra-operative and early post-operative data improves predictive accuracy. These techniques may be useful clinically to predict adverse outcomes after non-cardiac surgery.MINS outcomesDeath Outcomes
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 | 0.008 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 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".