Hostname: page-component-77f85d65b8-6c7dr Total loading time: 0 Render date: 2026-03-27T14:21:22.480Z Has data issue: false hasContentIssue false
  • English
  • Français

Combining search filters for randomized controlled trials with the Cochrane RCT Classifier in Covidence: a methodological validation study

Published online by Cambridge University Press:  28 August 2025

Klas Moberg Open the ORCID record for Klas Moberg [Opens in a new window]  and
Carl Gornitzki Open the ORCID record for Carl Gornitzki [Opens in a new window]
Klas Moberg*
Affiliation:
Swedish Agency for Health Technology Assessment and Assessment of Social Services (SBU) , Stockholm, Sweden
Carl Gornitzki
Affiliation:
Swedish Agency for Health Technology Assessment and Assessment of Social Services (SBU) , Stockholm, Sweden
*
Corresponding author: Klas Moberg; Email: klas.moberg@sbu.se
Rights & Permissions [Opens in a new window]

Abstract

Our objective was to evaluate the recall and number needed to read (NNR) for the Cochrane RCT Classifier compared to and in combination with established search filters developed for Ovid MEDLINE and Embase.com. A gold standard set of 1,103 randomized controlled trials (RCTs) was created to calculate recall for the Cochrane RCT Classifier in Covidence, the Cochrane sensitivity-maximizing RCT filter in Ovid MEDLINE and the Cochrane Embase RCT filter for Embase.com. In addition, the classifier and the filters were validated in three case studies using reports from the Swedish Agency for Health Technology Assessment and Assessment of Social Services to assess impact on search results and NNR. The Cochrane RCT Classifier had the highest recall with 99.64% followed by the Cochrane sensitivity-maximizing RCT filter in Ovid MEDLINE with 98.73% and the Cochrane Embase RCT filter with 98.46%. However, the Cochrane RCT Classifier had a higher NNR than the RCT filters in all case studies. Combining the RCT filters with the Cochrane RCT Classifier reduced NNR compared to using the RCT filters alone while achieving a recall of 98.46% for the Ovid MEDLINE/RCT Classifier combination and 98.28% for the Embase/RCT Classifier combination. In conclusion, we found that the Cochrane RCT Classifier in Covidence has a higher recall than established search filters but also a higher NNR. Thus, using the Cochrane RCT Classifier instead of current state-of-the-art RCT filters would lead to an increased workload in the screening process. A viable option with a lower NNR than RCT filters, at the cost of a slight decrease in recall, is to combine the Cochrane RCT Classifier with RCT filters in database searches.

Keywords

literature searchingmachine learningrandomized controlled trialssearch filtersstudy classifierssystematic review software

Information

Type
Research Article
Information
Research Synthesis Methods , Volume 16 , Issue 6 , November 2025 , pp. 953 - 960
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (https://creativecommons.org/licenses/by/4.0), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2025. Published by Cambridge University Press on behalf of The Society for Research Synthesis Methodology

Highlights

What is already known?

  • • The Cochrane RCT Classifier is a useful tool for automatically identifying possible RCTs and non-RCTs with a 99.5% recall.

What is new?

  • • The Cochrane RCT Classifier in Covidence does not reduce the number of hits retrieved in database searches to the same extent as existing state-of-the-art RCT filters in Ovid MEDLINE and Embase.com.

  • • Using RCT filters in Ovid MEDLINE and Embase.com prior to the application of the Cochrane RCT Classifier in Covidence reduces the number of hits to screen while maintaining a high recall.

Potential impact for RSM readers

  • • A reduced workload with a slight decrease in recall can be achieved in the title and abstract screening process by combining the use of RCT filters in Ovid MEDLINE and Embase.com with the Cochrane RCT Classifier in Covidence.

1 Introduction

Randomized controlled trials (RCTs) are the standard study design for inclusion in systematic reviews on effects of health care interventions. For this purpose, RCT search filters have been developed and validatedReference Cooper, Varley-Campbell and Carter 1 – Reference McKibbon, Wilczynski and Haynes 5 for efficient information retrieval.

In recent years, machine classifiers have been developed to increase efficiency in the abstract screening process. These classifiers are trained on data sets of title and abstract records meeting specific criteria using supervised machine learning. One such tool is the Cochrane RCT Classifier, originally built as a part of the Cochrane evidence pipeline with a purpose to populate the Cochrane CENTRAL database.Reference Thomas, McDonald and Noel-Storr 6 It has also been used as one of three components in the Cochrane Screen4Me workflow to aid authors of Cochrane reviews by distinguishing RCTs from non-RCTs.Reference Thomas, McDonald and Noel-Storr 6 – Reference Noel-Storr, Dooley and Wisniewski 9 In addition, the Cochrane RCT Classifier is integrated in EPPI-ReviewerReference Thomas, Graziosi and Brunton 10 and Covidence. 11

In this study, we evaluate the performance of the Cochrane RCT Classifier in Covidence. This tool enables the automatic tagging of references as either being a “Possible RCT” or “Not RCT” and facilitates the removal of references reporting on non-RCTs before screening. References marked as non-RCTs can be screened separately allowing the user to move individual references back to the screening workflow. This feature can be disabled at any point moving references marked as non-RCTs back to the title and abstract screening step for manual screening. 12

To the best of our knowledge, no validation study evaluating the combination of search filters and the Cochrane RCT Classifier has been published. Therefore, the aim of this study is to evaluate the recall (also known as “sensitivity”), number needed to read (NNR) and number of hits to screen for the Cochrane RCT Classifier compared to and in combination with established search filters developed for the bibliographic databases Ovid MEDLINE and Embase.com.

For this purpose, we identified RCTs from a repository of included studies at the Swedish Agency for Health Technology Assessment and Assessment of Social Services (SBU) to generate a gold standard set (GS). The repository is populated with bibliographic references from tables of included studies for all SBU reports from year 2019 onwards with SBU specific metadata such as information on study type and risk of bias. As of April 2025, the repository contains more than 6,000 references, including 1,249 manually classified RCT studies, and is continuously updated.

2 Methods

2.1 Generating the gold standard set

The relative recall method was used to identify RCTs to include in the GS. This method involves the collection of references included in evidence syntheses on a specific topic to calculate the recall of search filters.Reference Sampson, Zhang and Morrison 13 The GS was populated with RCTs (n = 1,249) from all 47 SBU reports published between 2019 and 2024 that has included RCTs (see Appendix 1 of the Supplementary Material for the SBU reports). The number of RCTs included in each report ranges from 1 to 139. One hundred forty-six references were then excluded to achieve the exact same GS, including the same set of references, for both Ovid MEDLINE and Embase.com. The final GS contains 1,103 RCTs published between 1970 and 2024 (see Appendix 2 of the Supplementary Material for the GS references). Approximately, 20% of the GS was identified in reports searching without a filter for study design, 30% with an RCT filter and 50% with a broader search filter for clinical trials.

2.2 Testing RCT Classifier and search filter recall

We validated the Cochrane RCT Classifier in Covidence, the Cochrane sensitivity-maximizing RCT filter in Ovid MEDLINEReference Lefebvre, Glanville, Briscoe, Higgins, Thomas and Chandler 14 and the Cochrane Embase RCT filter for Embase.comReference Glanville, Foxlee, Wisniewski, Noel-Storr, Edwards and Dooley 3 using recall, that is, the proportion of studies correctly identified as relevant relative to the total number of relevant studies.Reference Jenkins 15 The search filters were tested in Ovid MEDLINE and Embase.com by first running the search filter and then adding a search line for the GS with the PubMed identifiers (PMID) and Embase identification numbers for these records. The searches were then compared using the NOT Boolean operator to detect references in the GS not identified by the search filters. The Cochrane RCT Classifier was tested by exporting the GS to Covidence from the EndNote software, using the RefMan (RIS) Export format, to identify the number of references labeled as either “Possible RCT” or “Not RCT.” Recall was then calculated as: (number of gold standard records identified by a search filter, the RCT Classifier or a combination of both/total number of gold standard records) × 100 to express as a percentage.Reference Jenkins 15

2.3 Case study

To establish the efficiency of the Cochrane RCT Classifier in Covidence compared to and in combination with the Cochrane sensitivity-maximizing RCT filter in Ovid MEDLINE and the Cochrane Embase RCT filter for Embase.com, we conducted a case study. We reran the original search strategies from three SBU reports in Ovid MEDLINE (see Appendix 3 of the Supplementary Material for the Ovid MEDLINE search strategies) and Embase.com (see Appendix 4 of the Supplementary Material for the Embase.com search strategies) in January 2025, using date limits matching the search period from the original searches. Search results with and without RCT filters were documented and exported separately from Ovid MEDLINE, using the EndNote export format, and from Embase.com, using the RIS format. Number of included references retrieved by each individual search was recorded in order to be able to calculate recall. Search results were then imported into empty Covidence libraries to run the Cochrane RCT Classifier. The selection of the SBU reports was based on the topical coverage of the organization and for the purposes of illustrating the impact on searches of varied size. The topics of the three reports are listed below:

  • - SBU Assessment 337: Internet-delivered psychological treatment versus other available treatment options for common mental disorders .Reference Bachmann, Coray, Estermann and Ter Riet 16

  • - SBU Assessment 372: Treatment and social support for adults with co-occurring addictive and psychiatric disorders. 17

  • - SBU Policy support 379: Treatment and rehabilitation of post-COVID-19 and other post-infectious conditions. 18

NNR, the number of studies a researcher must read to identify a relevant study, 19 was used to assess the impact on screening workload. NNR was calculated as: 1/Precision, with precision equal to the proportion of references retrieved by a filter or classifier that are relevant: (Number of gold standard records identified by a search filter, the RCT Classifier or a combination of both/Number of records retrieved in a retrospective search). 19

3 Results

3.1 RCT Classifier and search filter recall

The Cochrane RCT Classifier in Covidence had the highest recall with 99.64% (1,099 out of 1,103 references). Three out of four missed references were indexed as RCTs in Ovid MEDLINE and therefore identified by the RCT filter. Two out of four missed references were indexed as RCTs in Embase.com and identified by the Cochrane Embase RCT filter for Embase.com. All four missed references lacked terms related to the RCT study design in the titles and abstracts. The Cochrane sensitivity-maximizing RCT filter in Ovid MEDLINE reached a recall of 98.73% and the Cochrane Embase RCT filter 98.46% missing 14 and 17 references, respectively. When combining the Cochrane RCT Classifier with RCT filters in Ovid MEDLINE and Embase.com, there was a slight decrease in recall in comparison to using the RCT filters alone (Ovid MEDLINE: 98.46%, Embase.com: 98.28%) missing three and two additional references, respectively (Table 1).

Table 1 Validation of recall with the entire gold standard (1,103 articles)

3.2 Case study

In the case study, the Cochrane RCT Classifier in Covidence and the RCT filters all achieved 100% recall for the included studies found by the original search strategies (Table 2). The Cochrane RCT Classifier in Covidence had a higher NNR than the RCT filters in all case studies, producing an average of 34% more references to screen. The lowest NNR was achieved when combining the Cochrane RCT Classifier with RCT filters in Ovid MEDLINE and Embase.com. The RCT classifier and RCT filter combination reduced NNR compared to when using RCT filters alone: For SBU Assessment 337, NNR was reduced from 263 to 219 in Ovid MEDLINE and from 270 to 236 in Embase.com. For SBU Assessment 372, NNR was reduced from 284 to 132 in Ovid MEDLINE and from 172 to 78 in Embase.com. For SBU Policy support 379, NNR was reduced from 77 to 34 in Ovid MEDLINE and from 81 to 34 in Embase.com. The RCT classifier and RCT filter combination reduced the number of references to screen with more than 50% compared to using RCT filters alone in two out of three case studies while the decreased screening burden remained at 13–17% in one case study (Table 2).

Table 2 Results of the case studies

4 Discussion

This study shows that the Cochrane RCT Classifier has a higher recall (99.64%) than state-of the-art RCT filters in Ovid MEDLINE (98.46%) and Embase.com (98.28%). However, at the cost of an average of 34% more references to screen than the validated search filters. An alternative is to combine searching with RCT filters and using the Cochrane RCT Classifier. In our study, this combination reduced NNR while achieving a recall comparable to using search filters in Ovid MEDLINE and Embase.com missing only three and two extra references, respectively.

The lack in overlap in missed RCTs between the Cochrane RCT Classifier and the search filters is due to the fact that none of the four studies not identified by the Cochrane RCT Classifier contained terms related to the RCT study design in titles or abstracts. However, three out of four missed references were indexed as RCTs in either MEDLINE or Embase. In other words, three out of four missed references would have been identified by the Cochrane RCT Classifier if it incorporated metadata from databases describing study design (e.g., Publication Type in MEDLINE). In a previous work on an RCT classifier, the highest recall was achieved when metadata for study design was incorporated in a machine classifier.Reference Marshall, Noel-Storr, Kuiper, Thomas and Wallace 7 However, for the purposes of creating the Cochrane RCT Classifier with a main focus on new records, usually lacking metadata about study design, a model was preferred that uses titles and abstract text without additional metadata.Reference Thomas, McDonald and Noel-Storr 6

There is no universal definition of an acceptable level of recall for a search filter or a machine classifier. Considering recall, the best option is not to use a search filter or machine classifier at all since every limiting concept added to a search increases the risk of missing relevant studies. Nevertheless, 90%Reference Beynon, Leeflang and McDonald 20 and 95%,Reference Glanville, Fleetwood, Yellowlees, Kaunelis and Mensinkai 21 respectively, have been suggested as thresholds in previous filter validation studies, numbers that can be compared to 99% which was the target recall used when creating the Cochrane RCT Classifier.Reference Thomas, McDonald and Noel-Storr 6 The results from our study can thus, based on their idea of an acceptable level of recall, assist review authors decisions on choosing RCT-filters, the Cochrane RCT Classifier or a combination of the two.

An important aspect to consider when using the Cochrane RCT Classifier is quality of metadata. For instance, we discovered that during export of references from Ovid MEDLINE to Covidence there were minor differences in performance of the Cochrane RCT Classifier depending on the export format used. Using the RIS export format, special characters appeared in some abstracts causing a slight decrease in recall as opposed to when using the EndNote format. It should be stressed that the RIS export format in Ovid MEDLINE is not identical to the RefMan (RIS) Export format in EndNote, using the latter does not cause any changes in the performance of the Cochrane RCT Classifier. Another metadata issue concerns short titles and presence or absence of abstracts. In the training, calibration, and validation of the Cochrane RCT Classifier references with less than 15 characters in the title or less than 400 characters in the abstract where excluded and in a secondary analysis of the validation recall was reduced to 94% when records with limited information in their titles and/or abstracts were included.Reference Thomas, McDonald and Noel-Storr 6 This issue is resolved in the Cochrane RCT Classifier in Covidence by only processing references that meet the minimum text requirements mentioned above leaving references that do not meet these thresholds to manual screening. 12

Systematic review handbooks have started to mention the use of machine classifiers in general as well as the Cochrane RCT Classifier specifically. The IQWiG General Methods handbook declares that validated search filters or machine classifiers are used if available. 22 NICE states in their handbook for developing guidelines that they support the use of machine classifiers if they improve efficiency in the search and screening process and their performance characteristics are known. 23 Neither IQWiG nor NICE comment on the possible combination of search filters and machine classifiers. The Cochrane Handbook acknowledges the use of study design classifiers in general but underlines that they should probably not be used in combination with study design filters.Reference Lefebvre, Glanville, Briscoe, Higgins, Thomas and Chandler 24

This study provides validation data demonstrating a slight decrease in recall when the Cochrane RCT Classifier in Covidence is combined with an RCT filter as opposed to using an RCT filter alone, while reducing the burden of screening references by more than 50% in two out of three case studies.

4.1 Study limitations

The Cochrane RCT Classifier is currently available in Cochrane Screen4Me, EPPI-Reviewer, and Covidence. Due to lack of access, we have only evaluated it in Covidence. A head-to-head comparison would provide valuable information about the performance of the Cochrane RCT Classifier in different systematic review software tools.

The GS was developed using the relative recall method. A potential limitation of this method is that it is dependent on the quality of the individual searches used to create the GS.Reference Sampson, Zhang and Morrison 13 Ideally, the GS would consist of references identified in searches without an RCT filter, otherwise it could be argued that the tested filters perform better because they were, in part, used to create the GS. The GS in this study does not exclusively contain references identified in searches without an RCT search filter. Approximately, 20% of the GS was identified in reports searching without a filter for study design, 30% with an RCT filter and 50% with a broader search filter for clinical trials. The share of references in the GS identified in searches without an RCT search filter is comparable to a much larger RCT filter validation study published in 2020.Reference Glanville, Kotas, Featherstone and Dooley 4

Finally, the majority of SBU reports used to populate the GS address issues within the health care domain making the results most applicable to this context. Furthermore, the reduction in NNR when combining the Cochrane RCT Classifier with RCT filters was more significant in two out of the three case studies. A more comprehensive evaluation including a larger set of case studies on different topics would provide more conclusive data on possible screening workload reductions.

5 Conclusions

The Cochrane RCT Classifier in Covidence has a higher recall than established search filters but also a higher NNR. Thus, using the Cochrane RCT Classifier instead of current state-of-the-art RCT filters would lead to an increased workload in the screening process. A viable option with a lower NNR than the RCT filters, at the cost of a slight decrease in recall, is to combine the Cochrane RCT Classifier with RCT filters in database searches. Larger evaluations of the combination of RCT filters and the Cochrane RCT Classifier are needed to further investigate time and resource savings and impact on recall.

Author contributions

Conceptualization: K.M., C.G.; Formal analysis: K.M.; Investigation: K.M.; Methodology: K.M., C.G.; Writing—original draft: K.M.; Writing—review and editing: K.M., C.G.

Competing interest statement

The two authors are employed by the Swedish Agency for Health Technology Assessment and Assessment of Social Services (SBU) and declares no competing interests.

Data availability statement

Data available within the article or its Supplementary Material (Appendices 1–4).

Funding statement

The project was conducted within assignment of the Swedish Agency for Health Technology Assessment and Assessment of Social Services (SBU) and external funding was not sought or used.

Supplementary material

To view supplementary material for this article, please visit http://doi.org/10.1017/rsm.2025.10023.

References

Cooper, C, Varley-Campbell, J, Carter, P. Established search filters may miss studies when identifying randomized controlled trials. J Clin Epidemiol. 2019;112:12–19. https://doi.org/10.1016/j.jclinepi.2019.04.002.CrossRefGoogle ScholarPubMed
Glanville, J, Dooley, G, Wisniewski, S, Foxlee, R, Noel-Storr, A. Development of a search filter to identify reports of controlled clinical trials within CINAHL Plus. Health Info Libr J. 2019;36(1):73–90. https://doi.org/10.1111/hir.12251.CrossRefGoogle ScholarPubMed
Glanville, J, Foxlee, R, Wisniewski, S, Noel-Storr, A, Edwards, M, Dooley, G. Translating the Cochrane EMBASE RCT filter from the Ovid interface to Embase.com: a case study. Health Info Libr J. 2019;36(3):264–277. https://doi.org/10.1111/hir.12269.CrossRefGoogle ScholarPubMed
Glanville, J, Kotas, E, Featherstone, R, Dooley, G. Which are the most sensitive search filters to identify randomized controlled trials in MEDLINE? J Med Libr Assoc. 2020;108(4):556–563. https://doi.org/10.5195/jmla.2020.912.CrossRefGoogle ScholarPubMed
McKibbon, KA, Wilczynski, NL, Haynes, RB. Retrieving randomized controlled trials from medline: a comparison of 38 published search filters. Health Info Libr J. 2009;26(3):187–202. https://doi.org/10.1111/j.1471-1842.2008.00827.x.CrossRefGoogle Scholar
Thomas, J, McDonald, S, Noel-Storr, A, et al. Machine learning reduced workload with minimal risk of missing studies: development and evaluation of a randomized controlled trial classifier for Cochrane Reviews. J Clin Epidemiol. 2021;133:140–151. https://doi.org/10.1016/j.jclinepi.2020.11.003.CrossRefGoogle ScholarPubMed
Marshall, IJ, Noel-Storr, A, Kuiper, J, Thomas, J, Wallace, BC. Machine learning for identifying randomized controlled trials: an evaluation and practitioner’s guide. Res Synth Methods. 2018;9(4):602–614. https://doi.org/10.1002/jrsm.1287.CrossRefGoogle ScholarPubMed
Noel-Storr, A, Dooley, G, Elliott, J, et al. An evaluation of Cochrane Crowd found that crowdsourcing produced accurate results in identifying randomized trials. J Clin Epidemiol. 2021;133:130–139. https://doi.org/10.1016/j.jclinepi.2021.01.006.CrossRefGoogle ScholarPubMed
Noel-Storr, AH, Dooley, G, Wisniewski, S, et al. Cochrane Centralised Search Service showed high sensitivity identifying randomized controlled trials: a retrospective analysis. J Clin Epidemiol. 2020;127:142–150. https://doi.org/10.1016/j.jclinepi.2020.08.008.CrossRefGoogle ScholarPubMed
Thomas, J, Graziosi, S, Brunton, J, et al. EPPI-Reviewer: advanced software for systematic reviews, maps and evidence synthesis. London: EPPI Centre, UCL Social Research Institute, University College London; 2023. Accessed February 14, 2025. https://eppi.ioe.ac.uk/cms/Default.aspx?tabid=2914.Google Scholar
Covidence Systematic Review Software. Melbourne: Veritas Health Innovation. Accessed February 14, 2025. www.covidence.org.Google Scholar
Automation using the Cochrane RCT Classifier. Covidence. Accessed April 7, 2025. https://support.covidence.org/help/automatically-tag-studies-not-reporting-on-rcts.Google Scholar
Sampson, M, Zhang, L, Morrison, A, et al. An alternative to the hand searching gold standard: validating methodological search filters using relative recall. BMC Med Res Methodol. 2006;6:33. https://doi.org/10.1186/1471-2288-6-33.CrossRefGoogle Scholar
Lefebvre, C, Glanville, J, Briscoe, S, et al. Chapter 4: Searching for and selecting studies. Cochrane Highly Sensitive Search Strategy for identifying randomized trials in MEDLINE: sensitivity-maximizing version (2023 revision); Ovid format. In: Higgins, JPT, Thomas, J, Chandler, J, et al., eds. Cochrane Handbook for Systematic Reviews of Interventions version 65. Cochrane; 2024. Accessed February 14, 2025. https://training.cochrane.org/handbook/current/chapter-04#section-4-4-7.Google Scholar
Jenkins, M. Evaluation of methodological search filters--a review. Health Info Libr J. 2004;21(3):148–163. https://doi.org/10.1111/j.1471-1842.2004.00511.x.CrossRefGoogle ScholarPubMed
[Internet-Delivered Psychological Treatment versus Other Available Treatment Options for Common Mental Disorders]. Stockholm: Swedish Agency for Health Technology Assessment and Assessment of Social Services (SBU); 2021. SBU Assessment 337. Published November 7, 2021. Accessed February 14, 2025. https://www.sbu.se/337.Google Scholar
[Treatment and Social Support for Adults with Co-occurring Addictive and Psychiatric Disorders - Part I: Pharmacological Interventions, A Preliminary Report]. Stockholm: Swedish Agency for Health Technology Assessment and Assessment of Social Services (SBU); 2024. SBU Assessment 372. Published February 6, 2024. Accessed February 14, 2025. https://www.sbu.se/372e.Google Scholar
[Treatment and Rehabilitation of Post-Covid and Other Post-Infectious Conditions]. Stockholm: Swedish Agency for Health Technology Assessment and Assessment of Social Services (SBU); 2024. SBU Policy support 379. Published August 13, 2024. Accessed February 14, 2025. https://www.sbu.se/379.Google Scholar
Bachmann, LM, Coray, R, Estermann, P, Ter Riet, G. Identifying diagnostic studies in MEDLINE: reducing the number needed to read. J Am Med Inform Assoc. 2002;9(6):653–658. https://doi.org/10.1197/jamia.m1124.CrossRefGoogle Scholar
Beynon, R, Leeflang, MM, McDonald, S, et al. Search strategies to identify diagnostic accuracy studies in MEDLINE and EMBASE. Cochrane Database Syst Rev. 2013;2013(9):Mr000022. https://doi.org/10.1002/14651858.MR000022.pub3.Google ScholarPubMed
Glanville, J, Fleetwood, K, Yellowlees, A, Kaunelis, D, Mensinkai, S. Development and Testing of Search Filters to Identify Economic Evaluations in MEDLINE and EMBASE. Ottawa; 2009. Published October 2009. Accessed May 3, 2025. https://www.cadth.ca/media/pdf/H0490_Search_Filters_for_Economic_Evaluations_mg_e.pdf.Google Scholar
IQWiG General Methods: Version 7.0. Cologne: Institute for Quality and Efficiency in Health Care; 2020. Accessed February 14, 2025. https://www.iqwig.de/methoden/general-methods_version-7-0.pdf.Google Scholar
Identifying the evidence: literature searching and evidence submission. Developing NICE guidelines: the manual. NICE. Published October 31, 2014. Accessed May 29, 2024. https://www.nice.org.uk/process/pmg20/chapter/identifying-the-evidence-literature-searching-and-evidence-submission.Google Scholar
Lefebvre, C, Glanville, J, Briscoe, S, et al. Chapter 4: Searching for and selecting studies [last updated September 2024]. In: Higgins, JPT, Thomas, J, Chandler, J, et al., eds. Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane; 2024. Accessed February 14 2025. https://training.cochrane.org/handbook/current/chapter-04.Google Scholar
Figure 0

Table 1 Validation of recall with the entire gold standard (1,103 articles)

Figure 1

Table 2 Results of the case studies

Supplementary material: File

Moberg and Gornitzki supplementary material

Moberg and Gornitzki supplementary material
Download Moberg and Gornitzki supplementary material(File)
File 464.1 KB