Documentation Gaps are the Primary Barrier to Reproducing Clinical ML Literature: Automated Replication with VERA. Chu, A. & Pettine, W. W. In Andreev, P., Van Woensel, W., Holmes, J., & Sauré, A., editors, Artificial Intelligence in Medicine, pages 412–422, Cham, 2027. Springer Nature Switzerland. doi abstract bibtex Machine learning models are increasingly embedded in clinical decision support systems (CDSS) for triage, risk stratification, and precision medicine, where patients’ lives depend on model reliability. Yet the scientific literature underpinning these tools is rarely verified independently, and recent evidence reveals that even FDA-authorized AI devices may lack adequate clinical validation. We present VERA (Verification Engine for Reproducible Analysis), an automated system that takes a research paper PDF and attempts end-to-end replication, enabling systematic trust verification of the clinical ML literature. VERA integrates hybrid PDF extraction, dual-path dataset resolution (UCI Repository and dynamic MIMIC-III cohort construction), deterministic execution, and quantitative scoring.
@inproceedings{chu_documentation_2027,
address = {Cham},
title = {Documentation {Gaps} are the {Primary} {Barrier} to {Reproducing} {Clinical} {ML} {Literature}: {Automated} {Replication} with {VERA}},
isbn = {978-3-032-30710-1},
shorttitle = {Documentation {Gaps} are the {Primary} {Barrier} to {Reproducing} {Clinical} {ML} {Literature}},
doi = {10.1007/978-3-032-30710-1_49},
abstract = {Machine learning models are increasingly embedded in clinical decision support systems (CDSS) for triage, risk stratification, and precision medicine, where patients’ lives depend on model reliability. Yet the scientific literature underpinning these tools is rarely verified independently, and recent evidence reveals that even FDA-authorized AI devices may lack adequate clinical validation. We present VERA (Verification Engine for Reproducible Analysis), an automated system that takes a research paper PDF and attempts end-to-end replication, enabling systematic trust verification of the clinical ML literature. VERA integrates hybrid PDF extraction, dual-path dataset resolution (UCI Repository and dynamic MIMIC-III cohort construction), deterministic execution, and quantitative scoring.},
language = {en},
booktitle = {Artificial {Intelligence} in {Medicine}},
publisher = {Springer Nature Switzerland},
author = {Chu, Andre and Pettine, Warren Woodrich},
editor = {Andreev, Pavel and Van Woensel, William and Holmes, John and Sauré, Antoine},
year = {2027},
keywords = {Automatic replication, Clinical decision support systems, Machine Learning, Practice guidelines, Reproducibility, Triage},
pages = {412--422},
}
Downloads: 0
{"_id":"JH7pWJTKEp73EGCmy","bibbaseid":"chu-pettine-documentationgapsaretheprimarybarriertoreproducingclinicalmlliteratureautomatedreplicationwithvera-2027","author_short":["Chu, A.","Pettine, W. W."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","address":"Cham","title":"Documentation Gaps are the Primary Barrier to Reproducing Clinical ML Literature: Automated Replication with VERA","isbn":"978-3-032-30710-1","shorttitle":"Documentation Gaps are the Primary Barrier to Reproducing Clinical ML Literature","doi":"10.1007/978-3-032-30710-1_49","abstract":"Machine learning models are increasingly embedded in clinical decision support systems (CDSS) for triage, risk stratification, and precision medicine, where patients’ lives depend on model reliability. Yet the scientific literature underpinning these tools is rarely verified independently, and recent evidence reveals that even FDA-authorized AI devices may lack adequate clinical validation. We present VERA (Verification Engine for Reproducible Analysis), an automated system that takes a research paper PDF and attempts end-to-end replication, enabling systematic trust verification of the clinical ML literature. VERA integrates hybrid PDF extraction, dual-path dataset resolution (UCI Repository and dynamic MIMIC-III cohort construction), deterministic execution, and quantitative scoring.","language":"en","booktitle":"Artificial Intelligence in Medicine","publisher":"Springer Nature Switzerland","author":[{"propositions":[],"lastnames":["Chu"],"firstnames":["Andre"],"suffixes":[]},{"propositions":[],"lastnames":["Pettine"],"firstnames":["Warren","Woodrich"],"suffixes":[]}],"editor":[{"propositions":[],"lastnames":["Andreev"],"firstnames":["Pavel"],"suffixes":[]},{"propositions":[],"lastnames":["Van","Woensel"],"firstnames":["William"],"suffixes":[]},{"propositions":[],"lastnames":["Holmes"],"firstnames":["John"],"suffixes":[]},{"propositions":[],"lastnames":["Sauré"],"firstnames":["Antoine"],"suffixes":[]}],"year":"2027","keywords":"Automatic replication, Clinical decision support systems, Machine Learning, Practice guidelines, Reproducibility, Triage","pages":"412–422","bibtex":"@inproceedings{chu_documentation_2027,\n\taddress = {Cham},\n\ttitle = {Documentation {Gaps} are the {Primary} {Barrier} to {Reproducing} {Clinical} {ML} {Literature}: {Automated} {Replication} with {VERA}},\n\tisbn = {978-3-032-30710-1},\n\tshorttitle = {Documentation {Gaps} are the {Primary} {Barrier} to {Reproducing} {Clinical} {ML} {Literature}},\n\tdoi = {10.1007/978-3-032-30710-1_49},\n\tabstract = {Machine learning models are increasingly embedded in clinical decision support systems (CDSS) for triage, risk stratification, and precision medicine, where patients’ lives depend on model reliability. Yet the scientific literature underpinning these tools is rarely verified independently, and recent evidence reveals that even FDA-authorized AI devices may lack adequate clinical validation. We present VERA (Verification Engine for Reproducible Analysis), an automated system that takes a research paper PDF and attempts end-to-end replication, enabling systematic trust verification of the clinical ML literature. VERA integrates hybrid PDF extraction, dual-path dataset resolution (UCI Repository and dynamic MIMIC-III cohort construction), deterministic execution, and quantitative scoring.},\n\tlanguage = {en},\n\tbooktitle = {Artificial {Intelligence} in {Medicine}},\n\tpublisher = {Springer Nature Switzerland},\n\tauthor = {Chu, Andre and Pettine, Warren Woodrich},\n\teditor = {Andreev, Pavel and Van Woensel, William and Holmes, John and Sauré, Antoine},\n\tyear = {2027},\n\tkeywords = {Automatic replication, Clinical decision support systems, Machine Learning, Practice guidelines, Reproducibility, Triage},\n\tpages = {412--422},\n}\n\n","author_short":["Chu, A.","Pettine, W. W."],"editor_short":["Andreev, P.","Van Woensel, W.","Holmes, J.","Sauré, A."],"key":"chu_documentation_2027","id":"chu_documentation_2027","bibbaseid":"chu-pettine-documentationgapsaretheprimarybarriertoreproducingclinicalmlliteratureautomatedreplicationwithvera-2027","role":"author","urls":{},"keyword":["Automatic replication","Clinical decision support systems","Machine Learning","Practice guidelines","Reproducibility","Triage"],"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"https://api.zotero.org/users/1528358/collections/DH4R6ZWA/items?key=L8lztpUgybScRoWSW76OhIE0&format=bibtex&limit=100","dataSources":["hoqThwEepw4WmWudJ"],"keywords":["automatic replication","clinical decision support systems","machine learning","practice guidelines","reproducibility","triage"],"search_terms":["documentation","gaps","primary","barrier","reproducing","clinical","literature","automated","replication","vera","chu","pettine"],"title":"Documentation Gaps are the Primary Barrier to Reproducing Clinical ML Literature: Automated Replication with VERA","year":2027}