Bibliographic record
Abstract
I t is well established that secondary mitral regur- gitation (sMR) is associated with poor outcomes regardless of left ventricular (LV) function and that mitral valve interventions could improve outcomes in this population.1 However, identification of those who would benefit most from intervention is unclear.Recent percutaneous mitral edge-to-edge repair trials have indicated that echocardiographic parameters related to LV dimensions and function could offer guidance in decision-making.[2][3][4][5][6] But, by focusing only on these echocardiographic parameters, we ignore the wealth of clinical and laboratory data available that could improve patient selection.Historically, performance of studies that could include these data have not been executed because of limitations in traditional statistical methods.Developments in machine learning have now provided the tools that could be applied to this problem.7 In this issue of JACC: Advances, Heitzinger et al 8 apply machine learning methods to develop a structured decision tree-like approach to risk stratify patients with severe sMR and a wide range of LV function.The authors used contemporaneously collected clinical, echocardiographic, and laboratory variables from 1,317 severe sMR patients.Their primary outcome was all-cause mortality.First, using 70% of the entire cohort, the authors performed univariate Cox proportional hazards ratio analysis followed by bootstrap resampling on only the clinical, the laboratory, or the echocardiographic variables to identify parameters associated with mortality for inclusion in a type of supervised learning called survival tree modeling.Then, they performed survival tree-based modeling with analysis stopped when there were 95 patients left in the terminal leaves.Finally, parameter cutoffs determined from this analysis were applied to the remaining 30% of their sMR cohort for validation using Kaplan-Meier and univariate Cox proportional hazards regression.The authors then repeated each step to identify parameters and their cutoff for the 3 heart failure (HF) subtypes within their sMR population.These subtypes included HF with preserved ejection fraction (left ventricular ejection fraction (left ventricular ejection fraction [LVEF]) >50%), HF with mildly reduced ejection fraction (LVEF 40%-50%), and HF with reduced ejection fraction (LVEF <40%).Because of the small numbers within each HF subgroup, there was no division into derivation and validation subgroups for this analysis.From the entire sMR cohort, after the first step, age and peripheral artery disease were most associated with all-cause mortality in the clinical variables analysis, LV end-diastolic diameter and the presence of mild, moderate, or severe LV dysfunction in the echocardiographic analysis, and blood urea nitrogen (BUN), creatinine, and albumin in the laboratory analysis.These parameters were then used in the survival tree modeling, and 8 subgroups were identified that differ in their long-term survival.A near 20-fold risk difference in mortality was seen between the highest and lowest risk subgroups.The lowest risk subgroup had survival of 97% and 85% at 1 and 6 years, respectively, and the highest risk subgroup had survival of 48% and 11%.Patients with low risk of death were younger patients (aged <66 years) with normal hemoglobin (>12.7 g/dL) and albumin
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 | 0.011 | 0.045 |
| Meta-epidemiology (narrow) | 0.005 | 0.002 |
| Meta-epidemiology (broad) | 0.008 | 0.004 |
| Bibliometrics | 0.005 | 0.002 |
| Science and technology studies | 0.002 | 0.002 |
| Scholarly communication | 0.007 | 0.003 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.020 | 0.026 |
| Insufficient payload (model declined to judge) | 0.014 | 0.009 |
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".