Utilizing Machine Learning for Proactive Post-Operative Patient Management (Preprint)
Bibliographic record
Abstract
Background: Postoperative complications contribute significantly to patient morbidity and mortality. Early prediction of such complications could enable the care team to intervene promptly and improve patient outcomes. Existing surgical risk scores are easy to use but lack accuracy and do not provide individualized risk or guidance for clinical decision-making. Objective: This study aimed to develop and evaluate machine learning models using the INSPIRE perioperative dataset to predict whether an individual patient will require critical postoperative interventions-ventilatory support, extracorporeal membrane oxygenation (ECMO), an intra-aortic balloon pump (IABP), or continuous renal replacement therapy (CRRT). Methods: Four artificial neural network models and 4 random forest classifiers were trained and tested using the publicly available INSPIRE dataset (131,000 surgical cases). Preoperative laboratory data, medication use, surgery type, and intraoperative vital signs were used as inputs. The dataset was randomly divided into a 70% training set (91,000 cases) and a 30% test set (38,998 cases), with a temporal split to prevent leakage. Model performance was assessed using accuracy, sensitivity, specificity, positive and negative predictive values, and area under the curve (AUC). Results: The artificial neural network models achieved high predictive performance for ECMO (accuracy 98.9%; AUC 0.992), ventilatory support (accuracy 97.7%; AUC 0.995), IABP (accuracy 98.0%; AUC 0.992), and CRRT (accuracy 94.9%; AUC 0.983). In the test set of 38,998 patients, the neural network models correctly classified 38,569 cases for ECMO, 38,097 for ventilatory support, 38,230 for IABP, and 37,021 for CRRT, corresponding to accuracies of 98.9%, 97.7%, 98.0%, and 94.9%, respectively. The corresponding random forest classifiers achieved accuracies of 98.0%, 99.0%, 98.2%, and 96.5%, respectively, with similar AUC values. These results indicate that the models can reliably identify patients who will or will not require lifesaving postoperative interventions despite the marked class imbalance. Our work began in April 2023, with the research concept taking shape over the following 1 to 2 months. Model development spanned approximately 3 to 4 months and was followed by 1 month of testing and validation conducted by other computer scientists. Conclusions: Machine learning models trained and tested on the INSPIRE dataset can provide individualized risk estimates for critical postoperative interventions. By supplying accurate predictions before or during surgery, these models can support proactive perioperative planning (eg, intensive care unit bed allocation and device readiness).
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
Direct model labels (unvalidated)
Per-model category and study-design labels from the labeling rounds. They are machine output, unvalidated, and the disagreement between models ships as data. No study design here is MEDLINE-validated yet.
| Model arm | Categories | Study design | Confidence |
|---|---|---|---|
| gemma | no category Domain: not available · Genre: Empirical About the Canadian research system: no · About a Canadian topic: no | Simulation or modeling | low |
| gpt | no category Domain: not available · Genre: Empirical About the Canadian research system: no · About a Canadian topic: no | Other design | low |
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 | 0.001 | 0.007 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.006 | 0.001 |
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, unvalidatedLabeled directly by 2 models reading the full record.
The models disagree on parts of this classification; every voice is preserved in the section at the end of the page.
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".