Personalized Mortality Prediction Driven by Electronic Medical Data and a Patient Similarity Metric
Bibliographic record
Abstract
BACKGROUND: Clinical outcome prediction normally employs static, one-size-fits-all models that perform well for the average patient but are sub-optimal for individual patients with unique characteristics. In the era of digital healthcare, it is feasible to dynamically personalize decision support by identifying and analyzing similar past patients, in a way that is analogous to personalized product recommendation in e-commerce. Our objectives were: 1) to prove that analyzing only similar patients leads to better outcome prediction performance than analyzing all available patients, and 2) to characterize the trade-off between training data size and the degree of similarity between the training data and the index patient for whom prediction is to be made. METHODS AND FINDINGS: We deployed a cosine-similarity-based patient similarity metric (PSM) to an intensive care unit (ICU) database to identify patients that are most similar to each patient and subsequently to custom-build 30-day mortality prediction models. Rich clinical and administrative data from the first day in the ICU from 17,152 adult ICU admissions were analyzed. The results confirmed that using data from only a small subset of most similar patients for training improves predictive performance in comparison with using data from all available patients. The results also showed that when too few similar patients are used for training, predictive performance degrades due to the effects of small sample sizes. Our PSM-based approach outperformed well-known ICU severity of illness scores. Although the improved prediction performance is achieved at the cost of increased computational burden, Big Data technologies can help realize personalized data-driven decision support at the point of care. CONCLUSIONS: The present study provides crucial empirical evidence for the promising potential of personalized data-driven decision support systems. With the increasing adoption of electronic medical record (EMR) systems, our novel medical data analytics contributes to meaningful use of EMR data.
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 | 0.002 | 0.006 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| 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".