RobotReviewer: evaluation of a system for automatically assessing bias in clinical trials
Open accessThe DOI matches its Crossref record; 22 of the 22 references it cites were found in a registry or an index.
Abstract
Objective To develop and evaluate RobotReviewer, a machine learning (ML) system that automatically assesses bias in clinical trials. From a (PDF-formatted) trial report, the system should determine risks of bias for the domains defined by the Cochrane Risk of Bias (RoB) tool, and extract supporting text for these judgments. Methods We algorithmically annotated 12,808 trial PDFs using data from the Cochrane Database of Systematic Reviews (CDSR). Trials were labeled as being at low or high/unclear risk of bias for each domain, and sentences were labeled as being informative or not. This dataset was used to train a multi-task ML model. We estimated the accuracy of ML judgments versus humans by comparing trials with two or more independent RoB assessments in the CDSR. Twenty blinded experienced reviewers rated the relevance of supporting text, comparing ML output with equivalent (human-extracted) text from the CDSR. Results By retrieving the top 3 candidate sentences per document (top3 recall), the best ML text was rated more relevant than text from the CDSR, but not significantly (60.4% ML text rated ‘highly relevant' v 56.5% of text from reviews; difference +3.9%, [−3.2% to +10.9%]). Model RoB judgments were less accurate than those from published reviews, though the difference was <10% (overall accuracy 71.0% with ML v 78.3% with CDSR). Conclusion Risk of bias assessment may be automated with reasonable accuracy. Automatically identified text supporting bias assessment is of equal quality to the manually identified text in the CDSR. This technology could substantially reduce reviewer workload and expedite evidence syntheses.
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
Figures





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.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Recent advances in natural language processing, text mining and machine learning have produced new algorithms that can accurately mimic human endeavour in systematic review activity, faster and more cheaply [5–8].”
From Making progress with the automation of systematic reviews: principles of the International Collaboration for the Automation of Systematic Reviews (ICASR) (2018) · cites this paper as [7]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Deep learning and machine learning have been used for assisting with the more complicated and time-intensive tasks of extracting and normalizing data that usually need to be hand curated [12].”
From A narrative review of recent tools and innovations toward automating living systematic reviews and evidence syntheses (2023) · cites this paper as [12]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“While other tools included in this review do support the manual extraction and appraisal of evidence, RobotReviewer, and Trialstreamer have gone furthest with the integration of automation and in making their tool and code accessible by using a transparent, well-reported, and open-source approach for validation [12,13].”
From A narrative review of recent tools and innovations toward automating living systematic reviews and evidence syntheses (2023) · cites this paper as [12]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“Validation studies of RobotReviewer have found that the article bias classifications (i.e. ‘low’ versus ‘high/unclear’ risk of bias) are reasonable but less accurate than those in published Cochrane reviews [12, 15].”
From Toward systematic review automation: a practical guide to using machine learning tools in research synthesis (2019) · cites this paper as [12]
The cited source says what the claim attributes to it.
In the citing paper
“For systematic reviews of RCTs, there exist only a few prototype platforms that make such technologies available (ExaCT [33] and RobotReviewer [12, 34, 35] being among these).”
From Toward systematic review automation: a practical guide to using machine learning tools in research synthesis (2019) · cites this paper as [12]
10 more claims about this paper
Create a free account to see every claim other papers make about this one.
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
The Cochrane Collaboration's tool for assessing risk of bias in randomised trials
Higgins, J. P. T., Altman, D. G., Gotzsche, P. C., Juni, P., Moher, D., Oxman, A. D., Savovic, J., Schulz, K. F., Weeks, L., Sterne, J. A. C., et al. (2011) · BMJ · doi:10.1136/bmj.d5928 · Cited by 33,817
Dermyshi, E., Wang, Y., Yan, C., Hong, W., Qiu, G., Gong, X., & Zhang, T. (2017) · Neonatology · doi:10.1159/000454668 · Cited by 24,451
A training algorithm for optimal margin classifiers
Boser, B. E., Guyon, I. M., & Vapnik, V. N. (1992) · Proceedings of the fifth annual workshop on Computational learning theory · doi:10.1145/130385.130401 · Cited by 11,992 · cited 2×
“We combine this with a squared L2 penalty term on the model parameters to form our objective, which describes a linear-kernel soft-margin SVM.14 This objective is shown in Equation 1.”
(2008) · Cited by 3,790
Distant supervision for relation extraction without labeled data
Mintz, M., Bills, S., Snow, R., & Jurafsky, D. (2009) · Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP: Volume 2 - ACL-IJCNLP '09 · doi:10.3115/1690219.1690287 · Cited by 2,752 · cited 1×
“Here we take a different approach: to obtain a large corpus of labeled data we use distant supervision, a machine learning methodology that exploits structured data in existing databases in place of direct human supervision.(10,11) Distant supervision involves automatically deriving labels for unlabeled data from exis…”
Frustratingly Easy Domain Adaptation
Daumé, H. (2009) · arXiv (Cornell University) · doi:10.48550/arxiv.0907.1815 · Cited by 1,383 · cited 1×
“Note that this can be viewed as an instantiation of Daumé’s frustratingly easy domain adaptation approach.(17) Our sentence model is thus trained jointly across all risk of bias domains; the shared component enables information sharing between them.”
Caruana, R. (1998) · Cited by 1,110
Feature hashing for large scale multitask learning
Weinberger, K., Dasgupta, A., Langford, J., Smola, A., & Attenberg, J. (2009) · Proceedings of the 26th Annual International Conference on Machine Learning · doi:10.1145/1553374.1553516 · Cited by 906 · cited 1×
“This has been shown to work well for large-scale multi-task text classification.(18) To summarize, we use a novel model for risk of bias prediction that 1) jointly makes article- and sentence-level predictions, and 2) borrows strength across the related risk of bias assessment tasks via multi-task learning, for both a…”
Evaluation Measures for Multi-class Subgroup Discovery
Abudawood, T., & Flach, P. (2009) · Lecture Notes in Computer Science · doi:10.1007/978-3-642-04180-8_20 · Cited by 778
Scaling to very very large corpora for natural language disambiguation
Banko, M., & Brill, E. (2001) · Proceedings of the 39th Annual Meeting on Association for Computational Linguistics - ACL '01 · doi:10.3115/1073012.1073017 · Cited by 675
Cited by
Items 301–321 of 321
Ferreira, G. F., Lewis, T. L., Filho, M. V. P., Machado, F., Arliani, G. G., Foni, N. O., Zambeli, R., Nery, C. A. D. S., & Pagnano, R. G. (2026) · Qeios · doi:10.32388/usaktb.3
Zhou, F., Lokker, C., Parrish, R., Haynes, R. B., Iorio, A., Saha, A., & Afzal, M. (2026) · JMIR AI · doi:10.2196/77311
Assisted authoring for avoiding inadequate claims in scientific reporting
Koroleva, A. (2020) · theses.fr (ABES) · doi:10.70675/f035bb8azddf1z4a7dzbf39zdf86c6947bb9 · Cited by 2
“Marshall et al. (2015) addressed the task of automating assessment of risk of bias in clinical trials.”
BackgroundMethodAhn, S., Kim, S., & Valdes, J. (2026) · Springer international handbooks of education · doi:10.1007/978-3-031-88898-4_24-1
Farquhar, H. (2026) · Research Square · doi:10.21203/rs.3.rs-9308492/v1
Thaker, N. G., Loaiza‐Bonilla, A., & Nabhan, C. (2026) · AI in Precision Oncology · doi:10.1177/2993091x261451036
“13,14 Extending these methods, an AI engine can parse the pre-speci fi ed primary and secondary endpoints, planned sample size, eligibility criteria, and analysis plans from protocols and registries and compare them directly with what is reported in submitted or published articles.”
Wang, D., Datta, S., Glasgow, J., Lee, K., Paek, H., Zhang, J., Zheng, Y., Huang, Y.-L., He, L., Rastegar-Mojarad, M., et al. (2026) · JMIR AI · doi:10.2196/81049 · Cited by 1
“Several existing SLR automation tools have been developed [38], including Trialstreamer [39], SWIFT-Review [40], DistillerSR [41], NestedKnowledge [42], SWIFT-ActiveScreener [43], Abstrackr [44], EPPI-Reviewer [45], and RobotReviewer [46].”
Automated tools for evidence quality assessment: a scoping review
Huang, J., Deng, X., Zhou, L., Tao, J., Liang, C., Yang, K., & Li, X. (2026) · BMC Medical Research Methodology · doi:10.1186/s12874-026-02868-3
“…were identified; 65% were reported as publicly available, while 35% did not specify accessibility: (1) ROB assessment: RobotReviewer ( n = 9) [17, 29–36], Evidence Mapping Tool ( n = 1) [37], ExaCT ( n = 1) [38], and STEED ( n = 1) [16] primarily targeted ROB assessment; ROB 2 Excel Macro…”
Gonzalez, F., Olmedo, F. A. M., Chavez, S., & López, V. (2026) · doi:10.2196/preprints.108848
Kurian, N., Cherian, J. M., Reji, R. M., & Varghese, K. G. (2026) · Evidence-Based Dentistry · doi:10.1038/s41432-026-01238-8 · Cited by 1
Comparative performance of large language models for appraising bias in real-world evidence studies
Okeke, C. M., Nechi, R. N., Khalid, J., Karaye, I. M., Thornton, J. D., & Yunusa, I. U. (2026) · Journal of Managed Care & Specialty Pharmacy · doi:10.18553/jmcp.2026.32.9.1062
Sánchez‐Piedra, C., Heyerdahl-Viau, I., Garcia-Carpintero, E.-E., Martínez-Núñez, J. M., & Prado-Galbarro, F. J. (2026) · Medical Sciences · doi:10.3390/medsci14050533
“Tools such as RobotReviewer have demonstrated the feasibility of automating structured risk of bias assessment in randomised trials, highlighting the potential of AI to enhance methodological consistency alongside reviewer judgement [10].”
Tan, H., Bin, M., Fang, J., Fang, C., & Shi, W. F. (2026) · Environmental Science & Technology · doi:10.1021/acs.est.6c09347
Quality Assessment Framework for Heterogeneous Sensor Data Standardization in Multimodal Research
Wu, Z. (2026) · doi:10.1145/3838457.3838557
“Performance is bench-marked against expert annotations obtained through systematic manual review, following established evaluation methodologies for automated assessment systems [14].”
Shi, B., Zhang, H., & Zhang, L. (2026) · Artificial Intelligence in Applied Sciences · doi:10.69882/adba.ai.2026023
“A tool like RobotRe-viewer excels in medical RCTs because it was likely trained and optimized on that specific literature (Marshall et al. 2016).”
Choi, J., Swan, H., Kim, M. J., & Kim, H. J. (2026) · Journal of Evidence-Based Practice · doi:10.63528/jebp.2026.00011
Klanjšček, T., Hackenberger, D. K., Đerđ, T., Geček, S., Klanjšček, J., Haberle, I., Terzić, I., Pečar-Ilić, J., Bavčević, L., & Hackenberger, B. K. (2026) · Ecological Modelling · doi:10.1016/j.ecolmodel.2026.111864
Adam, G. P., Davies, M. C., Caputo, E. L., Mai, H. J., Coppola, E. L., Holmer, H. K., Kuhn, E., Balk, E. M., Trikalinos, T. A., Ivlev, I., et al. (2026) · Research Synthesis Methods · doi:10.1017/rsm.2026.10118
X-SLR: A Hexagonal Architecture Framework for Reporting Systematic Literature Reviews
Aguirre, M., Ayala-Tipan, C., Andrade-Cabrera, A., Martínez-Mosquera, D., Recalde, L., & Suntaxi, G. (2026) · Information · doi:10.3390/info17101043
Nagendrababu, V., Dufey-Portilla, N., Abraham, S. B., Vinothkumar, T. S., Tzanetakis, G. N., Jaćimović, J., Jakovljević, A., Peña‐Bengoa, F., García‐Font, M., Sans, F. A., et al. (2026) · International Endodontic Journal · doi:10.1111/iej.70297
Metadata
- Authors
- Marshall, I. J., Kuiper, J., & Wallace, B. C.
- Affiliations
- King's College London (GB)
- University Medical Center Groningen (NL)
- The University of Texas at Austin (US)
- Corresponding author
- Marshall, Iain J
- Venue
- Journal of the American Medical Informatics Association
- Publisher
- Oxford University Press (OUP)
- Published
- 2015-06-22
- Volume
- 23
- Issue
- 1
- Pages
- 193-201
- ISSN
- 1067-5027 (ISSN-L)
- 1527-974X
- PMC
- PMC4713900
- OpenAlex
- W2184378182
- Cited by
- 317
- References
- 22
- Citations by year
- 2026: 34 · 2025: 39 · 2024: 39 · 2023: 32 · 2022: 23 · 2021: 41 · 2020: 49 · 2019: 13 · 2018: 23 · 2017: 16 · 2016: 5 · 2015: 2
- FWCI
- 13.97
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- Journal of the American Medical Informatics Association (publishedVersion, cc-by-nc)
- Europe PMC (PubMed Central) (publishedVersion, cc-by-nc)
- PubMed Central (publishedVersion, cc-by-nc)
- PubMed (publishedVersion)
- University of Groningen research database (University of Groningen / Centre for Information Technology) (publishedVersion, cc-by-nc)
- Research Portal (King's College London) (submittedVersion, cc-by-nc)
- HighWire Press Open Archive (submittedVersion)
- University of Groningen research database (University of Groningen / Centre for Information Technology) (submittedVersion, cc-by-nc)
- Funders
- Brown University
- King's College Hospital NHS Foundation Trust
- National Institute for Health and Care Research
- King's College London
- Monash University
- NIHR Collaboration for Leadership in Applied Health Research and Care South London
- Medical Research Council (MR/N015185/1)
- Institute of Education, University of London
- Topics
- Social Sciences › Decision Sciences › Statistics, Probability and Uncertainty › Meta-analysis and systematic reviews
- Social Sciences › General Social Sciences › Computational and Text Analysis Methods
- Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
- Keywords
- risk of bias assessment, clinical trials, Cochrane risk of bias tool, multi-task learning, Cochrane Database of Systematic Reviews, evidence synthesis
- MeSH terms
- Machine Learning, Algorithms, Clinical Trials as Topic, Natural Language Processing, Review Literature as Topic, Bias, Peer Review, Research, Databases as Topic, Data Mining
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.