Prospective Real-World Performance Evaluation of a Machine Learning Algorithm to Predict 30-Day Readmissions in Patients with Heart Failure Using Electronic Medical Record Data
Bibliographic record
Abstract
Background: Heart failure (HF) patients have a high readmission rate with approximately 20% of patients being readmitted within 30-days after discharge. Hospital interventions to reduce HF readmissions are resource- and effort-intensive. Widespread availability of electronic medical record data has spurred interest in using machine learning-based techniques for risk stratification of heart failure patients. The predictive performance of machine learning-based predictive models is often evaluated solely using the Area Under the Receiver Operating Characteristic (AUROC) curve. However, the AUROC is independent of prevalence therefore predictive models with the same AUROC can have differential clinical utility. Furthermore, the AUROC does not provide any insight about the presence of overfitting or decay in predictive performance of a model over time, both of which can affect its real-world performance. Objective: Our primary objective is to assess real-world performance of a 30-day readmission risk prediction model for HF patients, which had an AUROC of 0.71 in the training dataset. Methods: Predictions for risk of 30-day readmissions in HF patients in the Partners Healthcare System were prospectively obtained from the model. We assessed the positive (PPV) and negative predictive value (NPV), in addition to sensitivity, specificity, accuracy, model calibration and Brier score. Results: Four hundred twenty index admissions that were not part of the training dataset were included in this prospective evaluation. Readmission rate was 24% (101 30-day readmissions). The AUROC of the predictive model was 0.57. At a discrimination threshold of 0.2 for flagging high-risk index admissions, the sensitivity and specificity of the model were 53.46% and 63.32%, respectively. The PPV and NPV were 31.57% and 81.12%, respectively. The Brier score was 0.19. Conclusions: Our analysis offers important insights about the real-world performance of this predictive model. The NPV suggests that the model’s prediction about patients at low risk for readmission are reliable. This insight can be useful in optimizing resource allocation for patients with heart failure.
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 | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| 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".