High-performance automated abstract screening with large language model ensembles
Open accessThe DOI matches its Crossref record; 13 of the 13 references it cites were found in a registry or an index.
Abstract
Objective: Abstract screening is a labor-intensive component of systematic review involving repetitive application of inclusion and exclusion criteria on a large volume of studies. We aimed to validate large language models (LLMs) used to automate abstract screening. Materials and Methods: LLMs (GPT-3.5 Turbo, GPT-4 Turbo, GPT-4o, Llama 3 70B, Gemini 1.5 Pro, and Claude Sonnet 3.5) were trialed across 23 Cochrane Library systematic reviews to evaluate their accuracy in zero-shot binary classification for abstract screening. Initial evaluation on a balanced development dataset (n = 800) identified optimal prompting strategies, and the best performing LLM-prompt combinations were then validated on a comprehensive dataset of replicated search results (n = 119 695). Results: On the development dataset, LLMs exhibited superior performance to human researchers in terms of sensitivity (LLMmax = 1.000, humanmax = 0.775), precision (LLMmax = 0.927, humanmax = 0.911), and balanced accuracy (LLMmax = 0.904, humanmax = 0.865). When evaluated on the comprehensive dataset, the best performing LLM-prompt combinations exhibited consistent sensitivity (range 0.756-1.000) but diminished precision (range 0.004-0.096) due to class imbalance. In addition, 66 LLM-human and LLM-LLM ensembles exhibited perfect sensitivity with a maximal precision of 0.458 with the development dataset, decreasing to 0.1450 over the comprehensive dataset; but conferring workload reductions ranging between 37.55% and 99.11%. Discussion: Automated abstract screening can reduce the screening workload in systematic review while maintaining quality. Performance variation between reviews highlights the importance of domain-specific validation before autonomous deployment. LLM-human ensembles can achieve similar benefits while maintaining human oversight over all records. Conclusion: LLMs may reduce the human labor cost of systematic review with maintained or improved accuracy, thereby increasing the efficiency and quality of evidence synthesis.
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
The PRISMA 2020 statement: An updated guideline for reporting systematic reviews
Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., Shamseer, L., Tetzlaff, J. M., Akl, E. A., Brennan, S. E., et al. (2021) · BMJ · doi:10.1136/bmj.n71 · Cited by 108,283
Saito, T., & Rehmsmeier, M. (2015) · PLOS ONE · doi:10.1371/journal.pone.0118432 · Cited by 5,309
Large language models in medicine
Thirunavukarasu, A. J., Ting, D. S. J., Elangovan, K., Gutierrez, L., Tan, T. F., & Ting, D. S. W. (2023) · Nature Medicine · doi:10.1038/s41591-023-02448-8 · Cited by 3,944
Self-citationSiddaway, A. P., Wood, A. M., & Hedges, L. V. (2019) · Annual Review of Psychology · doi:10.1146/annurev-psych-010418-102803 · Cited by 2,550
Five steps to conducting a systematic review
Khan, K. S., Kunz, R., Kleijnen, J., & Antes, G. (2003) · Journal of the Royal Society of Medicine · doi:10.1177/014107680309600304 · Cited by 1,873
Meta-analysis and the science of research synthesis
Gurevitch, J., Koricheva, J., Nakagawa, S., & Stewart, G. (2018) · Nature · doi:10.1038/nature25753 · Cited by 1,709
The Systematic Review: An Overview
Aromataris, E., & Pearson, A. (2014) · AJN, American Journal of Nursing · doi:10.1097/01.naj.0000444496.24228.2c · Cited by 598
Systematic review adherence to methodological or reporting quality
Pussegoda, K., Turner, L., Garritty, C., Mayhew, A., Skidmore, B., Stevens, A., Boutron, I., Sarkis-Onofre, R., Bjerre, L. M., Hróbjartsson, A., et al. (2017) · Systematic Reviews · doi:10.1186/s13643-017-0527-2 · Cited by 322
Large Language Models for Chatbot Health Advice Studies: A Systematic Review
Huo, B., Boyle, A., Marfo, N., Tangamornsuksan, W., Steen, J. P., McKechnie, T., Lee, Y., Mayol, J., Antoniou, S. A., Thirunavukarasu, A. J., et al. (2025) · JAMA Network Open · doi:10.1001/jamanetworkopen.2024.57879 · Cited by 283
Self-citationGartlehner, G., Affengruber, L., Titscher, V., Noel-Storr, A., Dooley, G., Ballarini, N., & König, F. (2020) · Journal of Clinical Epidemiology · doi:10.1016/j.jclinepi.2020.01.005 · Cited by 206 · cited 2×
“43 Further work is required to integrate automated abstract screening into the conventional workflow of systematic review: our approach requires accessing an API with a spreadsheet containing details from every study identified at the search stage.”
Cited by
Generative artificial intelligence in medicine
Teo, Z. L., Thirunavukarasu, A. J., Elangovan, K., Cheng, H., Moova, P., Soetikno, B., Nielsen, C., Pollreisz, A., Ting, D. S. J., Morris, R. J. T., et al. (2025) · Nature Medicine · doi:10.1038/s41591-025-03983-2 · Cited by 119
Self-citationAccelerating clinical evidence synthesis with large language models
Wang, Z., Cao, L., Danek, B., Jin, Q., Lu, Z., & Sun, J. (2025) · npj Digital Medicine · doi:10.1038/s41746-025-01840-7 · Cited by 69
Artificial intelligence across the cancer care continuum
Riaz, I. B., Khan, M. A., & Osterman, T. J. (2025) · Cancer · doi:10.1002/cncr.70050 · Cited by 40
Metadata
- Authors
- Sanghera, R., Thirunavukarasu, A. J., El Khoury, M., O’Logbon, J., Chen, Y., Watt, A., Mahmood, M., Butt, H., Nishimura, G., & Soltan, A. A. S.
- Venue
- Journal of the American Medical Informatics Association
- Publisher
- Oxford University Press (OUP)
- Published
- 2025-03-22
- Volume
- 32
- Issue
- 5
- Pages
- 893-904
- ISSN
- 1067-5027 (ISSN-L)
- 1527-974X
- OpenAlex
- W4408763854
- Cited by
- 52
- References
- 13
- FWCI
- 32.36
- Type
- journal article
- Language
- en
- Open access
- hybrid
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.