The DOI matches its Crossref record; 21 of the 21 references it cites were found in a registry or an index; 11 of the 17 sources its claims cite were checked for support.
Abstract screening using the automated tool Rayyan: results of effectiveness in three diagnostic test accuracy systematic reviews
Open accessAbstract
OBJECTIVE: To evaluate the performance of the automated abstract screening tool Rayyan. METHODS: The records obtained from the search for three systematic reviews were manually screened in four stages. At the end of each stage, Rayyan was used to predict the eligibility score for the remaining records. At two different thresholds (≤2.5 and < 2.5 for exclusion of a record) Rayyan-generated ratings were compared with the decisions made by human reviewers in the manual screening process and the tool's accuracy metrics were calculated. RESULTS: Two thousand fifty-four records were screened manually, of which 379 were judged to be eligible for full-text assessment, and 112 were eventually included in the final review. For finding records eligible for full-text assessment, at the threshold of < 2.5 for exclusion, Rayyan managed to achieve sensitivity values of 97-99% with specificity values of 19-58%, while at the threshold of ≤2.5 for exclusion it had a specificity of 100% with sensitivity values of 1-29%. For the task of finding eligible reports for inclusion in the final review, almost similar results were obtained. DISCUSSION: At the threshold of < 2.5 for exclusion, Rayyan managed to be a reliable tool for excluding ineligible records, but it was not much reliable for finding eligible records. We emphasize that this study was conducted on diagnostic test accuracy reviews, which are more difficult to screen due to inconsistent terminology.
Overview
CoChat Summary
TL;DR
This study evaluated Rayyan’s automated abstract-screening performance in three diagnostic test accuracy systematic reviews, comparing its ratings against manual reviewer decisions. Using a threshold of <2.5 stars for exclusion, Rayyan achieved very high sensitivity but weak specificity; using ≤2.5 stars reversed that tradeoff.
Key findings
- A total of 2,054 records were screened manually; 379 were judged eligible for full-text assessment, and 112 reports were ultimately included in the final reviews.
- For identifying eligible records, at the < 2.5 exclusion threshold Rayyan achieved sensitivity of 97–99% and specificity of 19–58%; at ≤ 2.5, it had specificity of 100% but sensitivity of 1–29%.
- For identifying eligible reports, at < 2.5 Rayyan had almost perfect sensitivity but very poor specificity (3–5%); at ≤ 2.5, it had specificity of 78–99% with sensitivity up to 44%.
- The authors conclude that < 2.5 is the better threshold when maximizing sensitivity is the priority, while ≤ 2.5 may be used when specificity is more important.
Methods
- The study evaluated Rayyan on three ongoing diagnostic test accuracy systematic reviews of autism spectrum disorder, differing by index test: sMRI, rs-fMRI, and EEG machine-learning studies.
- Search results were drawn from Embase, MEDLINE, APA PsycINFO, IEEE Xplore, Scopus, Web of Science, plus grey literature sources (OpenGrey, CRL, OATD).
- Manual screening was done in four stages; Rayyan ratings were compared with human decisions using sensitivity, specificity, PPV, NPV, and F1 score at two exclusion thresholds: < 2.5 and ≤ 2.5.
Limitations
- The paper states that the reviews were all DTA reviews, which are harder to screen because of inconsistent terminology, limiting generalizability to other review types.
- Reviewers could see Rayyan’s computed ratings during screening, and although instructed to ignore them, the authors note some risk of bias may remain.
- The authors state they assessed only diagnostic accuracy measures and not other outcomes such as workload savings or user satisfaction.
- Evident from the text: the three reviews were on similar topics and machine-learning-based index tests, and data for the 3rd stage of the EEG review were missing because the results were accidentally lost.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures






Verification
Claims the paper makes
6 Valid · 5 Not valid · 6 Not checked · 38 Own results and interpretations
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The noticeable result was that NPV values remain consistent across all stages.”
§ Abstract screening using the automated tool Rayyan: results of effectiveness in three diagnostic test accuracy systematic reviews > Results > Test results
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“it seems that a threshold of < 2.5 (0.5 and 1.5) for exclusion is the optimal choice for record screening, as it achieved a good F1 score (0.354) with just 20% of the records manually screened.”
§ Abstract screening using the automated tool Rayyan: results of effectiveness in three diagnostic test accuracy systematic reviews > Discussion > Interpretation of the results
hedgedstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“In an analysis of 195 SRs in 2017 [3], the mean project length was 67.3 weeks with a range of 6–186 weeks.”
§ Abstract screening using the automated tool Rayyan: results of effectiveness in three diagnostic test accuracy systematic reviews > Background > Rationale A systematic review (SR) is a scientific investigation that focuses on a specific question and uses explicit, prespecified scientific methods to identify, select, assess, and summarize the findings of similar but separate studies [1]. SRs are considered to have the strongest level of evidence (level 1) in modern evidence-based medicine [2]. As the body of scientific literature is rapidly growing, SRs are more appreciated by healthcare decision-makers, due to providing brief robust reports of new interventions and phenomena. Unfortunately, current methods for conducting SRs are very time-consuming, resulting in the slow production of these important scientific reports. In an analysis of 195 SRs in 2017 [3], the mean project length was 67.3 weeks with a range of 6–186 weeks. In the same analysis, the number of studies found in the literature searches ranged from 27 to 92,020 with a mean of 1781. As a rough conservative estimate, it is believed that titles and abstracts of search results could be screened at a rate of 60–120 per hour [3]. With some basic calculations applied to the mean value of 1781, it results in about 14.8 to 29.6 hours of exhaustive work for reviewers with a maximum range of 766.8 to 1533.6 hours. Taking into consideration that most organizations prefer this task to be done in duplicate by at least two masked reviewers to minimize the risk of bias in study selection, the above numbers could be doubled. These numbers indicate the significant amount of time and energy a team of authors has to spend just selecting the potentially eligible studies. In recent years, a variety of automated tools have been introduced to facilitate the process of conduction of different parts of SRs, with different results. One of the main branches of these tools has been the study selection tools. Different automated tools have been developed for such tasks as Rayyan [4], Covidence [5], Abstrackr [6], Colandr [7], and EPPI-Reviewer [8]. These tools use text mining techniques to identify relevant information from text using statistical pattern learning that recognizes patterns in data. To achieve this, supervised learning algorithms are incorporated in their core, which tries to find patterns in the studies classified by the reviewers to predict the classification of unclassified records. These tools vary significantly in their core learning algorithm, offered features, and availability. In a scoping review in 2020 [9] Rayyan managed to get the highest score in weighted feature analysis and second place in the overall experience score (as rated by users in a survey) among these tools. Rayyan, a web-based automated screening tool, developed by Qatar Computing Research Institute (QCRI) was initially launched in 2014 and is currently accessible at www.rayyan.ai. It uses text mining methods to facilitate semi-automatic screening of records for SRs. As a reviewer screens some records and labels them for either inclusion, exclusion, or “maybe” relevant to the subject of the review, the tool thrives for finding patterns and similarities to give a similarity score to each of the remaining records as a five-star rating. Higher ratings reflect the computed underlying probabilities of the record being included are higher, and vice versa. The simplicity of using Rayyan, combined with its completely free access, has made it quite popular among users. It also provides some interesting features such as allowing independent masked screening of the records by more than one user, creating custom labels for records, highlighting words for inclusion and exclusion (which significantly assists manual screening), and choosing the reason(s) for excluding a record. Rayyan’s code is written in the open-source framework Ruby on Rails [10] and runs on Heroku [11] which is a Platform as a Service based on the cloud-hosting Amazon Web Services. The Rayyan classification system is described in a paper [4] by the developers as follows: Rayyan extracts all the words and pairs of words (bigrams) and MeSH terms following the removal of stop words and the stem of the remaining words from the titles/abstracts. These words are then used as features by the machine learning algorithm (support vector machine (SVM) classifier). As users label records as either excluded or included, the app uses the classifier to learn the features and build a model. The algorithm then runs on the records without a decision and gives a score to each of them, revealing how close it is to the include or exclude classes. That score is presented to the user as a five-star rating.
What the sources say
- Not checked — no full text of the sourcecites [3] Lefebvre C, et al. Searching for and selecting studies. Cochrane Handbook Syst Rev Intervent. 2019;67–107. source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“Such a significant imbalance could have strongly affected the training process of the learning algorithm.”
§ Abstract screening using the automated tool Rayyan: results of effectiveness in three diagnostic test accuracy systematic reviews > Discussion > Interpretation of the results
What the sources say
- Not checked — no full text of the sourcecites [22] Shultz, T. R. et al. Class imbalance problem. In encyclopedia of machine learning (eds. Sammut, C. & Webb, G. I.) 171–171 (Springer US, 2011). :https://doi.org/10.1007/978-0-387-30164-8_110. source
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Based on these results, Rayyan has the potential to reach acceptable PPV and NPV after manually screening 40% of records, considering a threshold of ≤2.5 for exclusion.”
§ Abstract screening using the automated tool Rayyan: results of effectiveness in three diagnostic test accuracy systematic reviews > Results > Test results
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
4 more claims
Create a free account to see all 55 claims and every source verdict.
- Probably not a substantial claim. 25
- It may state two separate claims as one. 30
- The quote only partly states this claim. 1
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 cited source says what the claim attributes to it.
In the citing paper
“Following deduplication of all retrieved records, the screening process was conducted using the Rayyan Professional systematic review platform (Valizadeh et al., 2022) to facilitate blinded and independent study selection.”
From Between cognitive offloading and critical autonomy: A systematic review of the epistemic implications of generative AI in higher education. (2026) · cites this paper as Valizadeh et al., 2022
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
PRISMA-P Group, Moher, D., Shamseer, L., Clarke, M., Ghersi, D., Liberati, A., Petticrew, M., Shekelle, P., & Stewart, L. A. (2015) · Systematic Reviews · doi:10.1186/2046-4053-4-1 · Cited by 26,223
Rayyan-a web and mobile app for systematic reviews.
Ouzzani, M., Hammady, H., Fedorowicz, Z., & Elmagarmid, A. (2016) · Systematic Reviews · doi:10.1186/s13643-016-0384-4 · Cited by 26,092 · cited 3×
“Different automated tools have been developed for such tasks as Rayyan [4], Covidence [5], Abstrackr [6], Colandr [7], and EPPI-Reviewer [8].”
BackgroundMethodThe PRISMA 2020 statement: an updated guideline for reporting systematic reviews
Page, M. J., McKenzie, J., Bossuyt, P., Boutron, I., Hoffmann, T., mulrow, C., Shamseer, L., Tetzlaff, J., Akl, E., Brennan, S. E., et al. (2020) · doi:10.31222/osf.io/v7gm2 · Cited by 5,177
Encyclopedia of Machine Learning
(2010) · Springer eBooks · doi:10.1007/978-0-387-30164-8 · Cited by 3,465
STARD 2015: An Updated List of Essential Items for Reporting Diagnostic Accuracy Studies
Bossuyt, P. M., Reitsma, J. B., Bruns, D. E., Gatsonis, C. A., Glasziou, P. P., Irwig, L., Lijmer, J. G., Moher, D., Rennie, D., de Vet, H. C. W., et al. (2015) · Clinical Chemistry · doi:10.1373/clinchem.2015.246280 · Cited by 984 · cited 1×
“This study’s design and methods are reported in line with the Standards for Reporting Diagnostic accuracy studies (STARD) checklist [13].”
MethodFinding What Works in Health Care: Standards for Systematic Reviews
Eden, J., Levit, L. A., Berg, A. O., & Morton, S. C. (2013) · Cited by 943
Searching for and selecting studies
Lefebvre, C., Glanville, J., Briscoe, S., Littlewood, A., Marshall, C., Metzendorf, M., Noel‐Storr, A., Rader, T., Shokraneh, F., Thomas, J., et al. (2019) · Cochrane Handbook for Systematic Reviews of Interventions · doi:10.1002/9781119536604.ch4 · Cited by 647 · cited 3×
“As a rough conservative estimate, it is believed that titles and abstracts of search results could be screened at a rate of 60–120 per hour [3].”
BackgroundMethodInfluentialHarrison, H., Griffin, S. J., Kuhn, I., & Usher-Smith, J. A. (2020) · BMC Medical Research Methodology · doi:10.1186/s12874-020-0897-3 · Cited by 449 · cited 1×
“In a scoping review in 2020 [9] Rayyan managed to get the highest score in weighted feature analysis and second place in the overall experience score (as rated by users in a survey) among these tools.”
BackgroundCost-Sensitive Learning and the Class Imbalance Problem
Ling, C. X., & Sheng, V. S. (2008) · Cited by 260
Research works printed but not linked: 3
- Covidence. Covidence systematic review software, veritas health innovation. Melbourne; 2021.
- Thomas J, Brunton J. EPPI-reviewer: software for research synthesis; 2007.
- Hupe M. EndNote X9. J ElectroResources Med Lib. 2019;16:117–9.
Cited by
Adaptive Learning Using Artificial Intelligence in e-Learning: A Literature Review
Gligorea, I., Cioca, M., Oancea, R., Gorski, A.-T., Gorski, H., & Tudorache, P. (2023) · Education Sciences · doi:10.3390/educsci13121216 · Cited by 980
“In its function, the tool excels by assigning a similarity score to each record during the review process, as records are categorized for inclusion, exclusion, or potential relevance [22].”
MethodEzzdine, L. B., Dhahbi, W., Dergaa, I., Ceylan, H. İ., Guelmami, N., Saad, H. B., Chamari, K., Ștefănică, V., & Omri, A. E. (2025) · Frontiers in Neuroscience · doi:10.3389/fnins.2025.1502417 · Cited by 119
Methodological insights into ChatGPT’s screening performance in systematic reviews
Issaiy, M., Ghanaati, H., Kolahi, S., Shakiba, M., Jalali, A. H., Zarei, D., Kazemian, S., Avanaki, M. A., & Firouznia, K. (2024) · BMC Medical Research Methodology · doi:10.1186/s12874-024-02203-8 · Cited by 100
“Historically, several studies have sought to apply machine learning or deep learning methods to assist in this process [11–14].”
BackgroundSulandari, S., Coats, R. O., Miller, A., Hodkinson, A., & Johnson, J. (2024) · The Gerontologist · doi:10.1093/geront/gnae128 · Cited by 81
High-performance automated abstract screening with large language model ensembles
Sanghera, R., Thirunavukarasu, A. J., Khoury, M. E., O’Logbon, J., Chen, Y., Watt, A., Mahmood, M., Butt, H., Nishimura, G., & Soltan, A. A. S. (2025) · Journal of the American Medical Informatics Association · doi:10.1093/jamia/ocaf050 · Cited by 86
Metadata
- Authors
- Valizadeh, A., Moassefi, M., Nakhostin-Ansari, A., Hosseini Asl, S. H., Saghab Torbati, M., Aghajani, R., Maleki Ghorbani, Z., & Faghani, S.
- Affiliations
- Tehran University of Medical Sciences (IR)
- Exceptional Talents Development Center (IR)
- Corresponding author
- Valizadeh, Amir
- Venue
- BMC Medical Research Methodology
- Published
- 2022-06-02
- Volume
- 22
- Issue
- 1
- Pages
- 160-160
- ISSN
- 1471-2288 (ISSN-L)
- PMC
- PMC9161508
- OpenAlex
- W4281646566
- Semantic Scholar
- 54f23028ce10416b843ca7d660f10ada0623fff1
- Cited by
- 159
- References
- 21
- Citations by year
- 2026: 50 · 2025: 59 · 2024: 35 · 2023: 14 · 2022: 1
- Type
- conference abstract
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- BMC Medical Research Methodology (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Topics
- Social Sciences › Decision Sciences › Statistics, Probability and Uncertainty › Meta-analysis and systematic reviews
- Physical Sciences › Computer Science › Information Systems › Information Retrieval and Search Behavior
- Social Sciences › Library and Information Sciences › scientometrics and bibliometrics research
- Keywords
- Rayyan, systematic review screening, eligibility determination, sensitivity and specificity
- MeSH terms
- Delivery of Health Care, Diagnostic Tests, Routine, Humans, Research
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Abstract screening using the automated tool Rayyan: results of effectiveness in three diagnostic test accuracy systematic reviews. https://app.cochat.ai/papers/10.1186/s12874-022-01631-8
- The paper
- Valizadeh, A., Moassefi, M., Nakhostin-Ansari, A., Hosseini Asl, S. H., Saghab Torbati, M., Aghajani, R., Maleki Ghorbani, Z., & Faghani, S. (2022). Abstract screening using the automated tool Rayyan: results of effectiveness in three diagnostic test accuracy systematic reviews. BMC Medical Research Methodology. https://doi.org/10.1186/s12874-022-01631-8
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.