Advancing the beneficial use of machine learning in health care and medicine: Toward a community understanding
Open accessThe DOI matches its Crossref record; 4 of the 4 references it cites were found in a registry or an index.
Abstract
Due to the abundance of health data and growing computational power, machine learning (ML) is engaging health researchers in a process of discovery around developing data-driven algorithms to make clinically reliable predictions.ML has the potential to provide effective tools to improve outcomes and reduce costs in health care, and the clinical community should partake in developing and evaluating these discoveries.However, if misguided applications are to be avoided, methodological savvy will be needed to develop, interpret and implement ML in medicine [1,2].In preparing PLOS Medicine's Special Issue (SI) on Machine Learning in Health and Biomedicine, Guest Editors Atul Butte, Suchi Saria, and Aziz Sheikh, and the PLOS Medicine Editors, have identified two principles in the design and reporting of ML studies that we believe should guide researchers in advancing the beneficial use of ML in healthcare and medicine.These principles, which also inform PLOS Medicine's editorial priorities for manuscript submissions in this field, require first, that models derived through ML are demonstrably fit for their stated clinical purpose, and second, that researchers undertake and report appropriate efforts to validate these models in external datasets. Fit-for-purpose performanceAn ML model need not be ready for off-the-shelf, practice-changing implementation to make a valuable contribution, but must achieve a clear purpose.The current SI includes reports on ML approaches that have undergone retrospective validation and are now ready for prospective testing, ML at early stages of validation, and head-to-head comparisons between standard epidemiological and ML approaches that suggest future directions without themselves establishing clinical utility.The articles vary widely in the intended application or "use cases" for their models.In a study from Soo-Jin Kang and colleagues, ML deploying imaging data from intravascular coronary angiography was used to diagnose coronary ischemia without a more invasive measurement-fractional flow reserve-that is currently the diagnostic standard [3].The authors were not trying to challenge the standard of care, but to improve diagnosis when resource constraints or clinical indications make fractional flow reserve impractical or unsuitable.For applications intended to provide pragmatic options for nonideal circumstances, model performance can be benchmarked against current practice rather than recommended practice, as long as the limitations of the advance are clear to readers.Early research planning should consider clinically acceptable performance characteristics for the targeted application, and a clear description of its intended use-and inappropriate
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Zech, J. R., Badgeley, M. A., Liu, M., Costa, A. B., Titano, J. J., & Oermann, E. K. (2018) · PLOS Medicine · doi:10.1371/journal.pmed.1002683 · Cited by 1,719
Prediction models need appropriate internal, internal-external, and external validation
Steyerberg, E. W., & Harrell, F. E. (2016) · Journal of Clinical Epidemiology · doi:10.1016/j.jclinepi.2015.04.005 · Cited by 1,304
Taylor, A. G., Mielke, C., & Mongan, J. (2018) · PLOS Medicine · doi:10.1371/journal.pmed.1002697 · Cited by 212
Corey, K. M., Kashyap, S., Lorenzi, E., Lagoo-Deenadayalan, S. A., Heller, K., Whalen, K., Balu, S., Heflin, M. T., McDonald, S. R., Swaminathan, M., et al. (2018) · PLOS Medicine · doi:10.1371/journal.pmed.1002701 · Cited by 204
Cited by
Artificial intelligence for mental health and mental illnesses: an overview
Graham, S., Depp, C., Lee, E. E., Nebeker, C., Tu, X., Kim, H.-C., & Jeste, D. V. (2019) · Current Psychiatry Reports · doi:10.1007/s11920-019-1094-0 · Cited by 970
Evaluation of clinical prediction models (part 1): from development to external validation
Collins, G. S., Dhiman, P., Ma, J., Schlussel, M. M., Archer, L., Van Calster, B., Harrell, F. E., Martin, G. P., Moons, K. G. M., van Smeden, M., et al. (2024) · BMJ · doi:10.1136/bmj-2023-074819 · Cited by 798
“50 Some journals refuse to publish model development studies without an external validation 51 ; this stance is outdated and misinformed, and could encourage researchers to perform a meaningless and misleading external validation (eg, non-representative convenience sample, too small, even data…”
BackgroundMachine intelligence in peptide therapeutics: A next‐generation tool for rapid disease screening
Basith, S., Manavalan, B., Hwan Shin, T., & Lee, G. (2020) · Medicinal Research Reviews · doi:10.1002/med.21658 · Cited by 312
Thomas, J., McDonald, S., Noel-Storr, A., Shemilt, I., Elliott, J., Mavergames, C., & Marshall, I. J. (2021) · Journal of Clinical Epidemiology · doi:10.1016/j.jclinepi.2020.11.003 · Cited by 253
Predictive analytics in health care: how can we know it works?
Van Calster, B., Wynants, L., Timmerman, D., Steyerberg, E. W., & Collins, G. S. (2019) · Journal of the American Medical Informatics Association · doi:10.1093/jamia/ocz130 · Cited by 223
Metadata
- Authors
- Nevin, L., & Editors, O. B. O. T. P. M.
- Venue
- PLOS Medicine
- Publisher
- Public Library of Science (PLoS)
- Published
- 2018-11-30
- Volume
- 15
- Issue
- 11
- Pages
- e1002708
- ISSN
- 1549-1277 (ISSN-L)
- 1549-1676
- OpenAlex
- W2902239438
- Cited by
- 58
- References
- 4
- FWCI
- 1.09
- Type
- journal article
- Language
- en
- Open access
- gold
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.