AI Science Tracker
EN

University of Wisconsin-Milwaukee

1 discovery from 2022

Discoveries

Learned preoperative record features improve retrospective orthopedic VTE classification

Source check: Source checked. Not independent scientific confirmation.

Finding: Genetic-algorithm feature selection and a deep neural classifier distinguished recorded postoperative venous-thromboembolism events in a large orthopedic-procedure dataset. The final table reports better discrimination than using established risk factors alone. Patient overlap, procedure coding and model-selection uncertainty limit this retrospective result; no causal medication risk or improved prophylaxis outcome was demonstrated.

Technology: Genetic-algorithm feature selection with gradient boosting; fully connected deep neural network classification

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
Learned feature selection and nonlinear classification linked preoperative record patterns to measured/coded postoperative outcomes, providing a specific retrospective clinical evaluation beyond a generic architecture benchmark. No discovered causal risk factor or clinical intervention effect is adopted.
Test and result
The source derived 541,354 procedure records from 392,661 patients, using prior diagnosis/medication windows and coded VTE within thirty days. A 70:30 random procedure split evaluated a six-hidden-layer DNN. Final Table 6 reports AUC 0.901 after genetic feature selection versus 0.780 using established risk factors. The abstract instead cites 0.873, which main text assigns to a preliminary feature-selection stage.
Limitation / still unknown
Patients could contribute multiple procedures, and patient-level partition isolation was not established. Broad procedure-code ranges may include operations beyond hip/knee replacement. Genetic selection/tuning isolation from final evaluation remains unclear; excluding postoperative prophylaxis also limits preoperative applicability. The Table6/abstract difference is not silently resolved. Drug-feature importance is associational and does not establish medication harms or benefit from changing treatment.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9892391/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 21 Nov 2026. Review scope: Read actual full primary PMC clinical cohort/procedure construction, endpoint/windows, holdout, genetic/GBT feature-selection and DNN Methods; directly read primary Tables4/6 and model Results. Abstract-versus-main AUC association and procedure/patient separation caveats checked. Individual records and detailed patient-level test isolation not reconstructed. Exact original article front DOI/title and fresh PubMed primary relationship metadata checked..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open the interactive profile for the timeline, or browse all organizations and model families.