Bibliographic Database

Bibliographic Database in Health Technology Assessment

A bibliographic database holds one record per publication, giving the citation, usually an abstract and a set of subject terms, and reviewers search it to find reports of studies rather than the study data themselves. MEDLINE, Embase and the Cochrane Central Register of Controlled Trials (CENTRAL) are the main general sources for trials, and health economists add economic sources such as the NHS Economic Evaluation Database (NHS EED) archive and the Tufts CEA Registry. In health technology assessment, many inputs, from treatment effects to health state utility values and published cost-effectiveness studies, are found this way. Which databases are searched, and how well the strategy fits each one's indexing, decides which studies a committee ever sees.

How a database record is built and indexed

The Cochrane Handbook notes that MEDLINE and Embase include abstracts for most recent records and can be searched both by words in the title or abstract and by the controlled vocabulary assigned to each record. Indexers or automated methods choose those terms from a thesaurus.

MEDLINE is the US National Library of Medicine's bibliographic database, and its records are indexed with Medical Subject Headings (MeSH), a controlled and hierarchically organised vocabulary. Embase uses its own thesaurus, Emtree. Because one study can produce several reports, the Handbook treats studies, not reports, as the unit of interest in a systematic review, and linking a conference abstract, main paper and follow-up of the same trial belongs to data extraction and the evidence table.

Databases searched for clinical and economic evidence

Database choice follows the review question, and an HTA usually needs several. General biomedical databases carry most reports of trials, while specialist sources hold economic evaluations, utility weights and earlier assessments.

PubMed also carries citations not yet indexed for MEDLINE, so MeSH alone misses the newest records, and PRISMA-S notes that the platform, such as Ovid or EBSCOhost, can change what is retrieved.

The economic sources have limits of their own. NHS EED was funded from 1994 to March 2015, with its searches run until the end of 2014 and its last records published on 31 March 2015. The HTA database is now produced by INAHTA, and the CRD copy has taken no new records since 31 March 2018. The CEA Registry admits only original analyses that measure benefit in QALYs and excludes HTA reports, cost-benefit analyses and systematic reviews, so it covers cost-utility analyses, not all economic evaluation.

Controlled vocabulary and free text in the search strategy

A search strategy combines two retrieval approaches. The Cochrane Handbook makes it mandatory for Cochrane reviews to identify controlled vocabulary, including exploded terms, together with free-text terms that allow for spelling variants, synonyms, acronyms, truncation and proximity. Exploding a heading retrieves records indexed with that heading and with the narrower headings below it in the thesaurus.

The thesauri differ, so a strategy cannot be moved between databases unchanged. CRD's guidance gives the example of pressure sores, indexed under the MeSH term pressure ulcer in MEDLINE and the Emtree term decubitus in Embase. Free text stays necessary because new records may not yet carry subject terms and some study types are indexed unreliably. The NICE manual (PMG36) says filters should not be used to narrow searches for diagnostic test accuracy studies, because such studies are often poorly indexed.

Why reviews search more than one database

The Cochrane Handbook states that searching two or more databases lowers the risk of missing eligible studies, and that studies comparing MEDLINE and Embase have generally concluded that a comprehensive search needs both. For Cochrane intervention reviews, searching CENTRAL, MEDLINE and Embase (where Embase is available) is mandatory, and searching subject-specific and regional databases such as CINAHL is highly desirable.

In a study of 58 published systematic reviews, Bramer and colleagues found that 16% of included references were found in only one database, with Embase contributing the most unique references. The Handbook also reports work in which 75% to 92% of included studies could be found through PubMed, with little effect on meta-analyses from adding the rest. It concludes that abbreviated searches may suit rapid evidence syntheses, while decisions demanding high certainty require comprehensive searches.

What NICE and CRD guidance expect in technology appraisal

NICE's manual (PMG36) asks for evidence on outcomes to come from a systematic review and for every record retrieved by the search to be assessed by more than one reviewer. For published economic evaluations the stopping rule differs, as searches should be transparent and reproducible and run until sufficient relevant evidence is found, without needing to be exhaustive if further studies would only confirm it. When EQ-5D data are taken from the literature, the methods for identifying them should be systematic and transparent, and the methods for finding cost sources should be defined, preferably through systematic review.

The company evidence submission template (PMG24) turns these into reporting duties. For clinical evidence it asks for a full list of information sources and the full electronic search strategies for all databases, including limits, in enough detail to reproduce the results. Separate appendices ask how cost-effectiveness studies, health-related quality of life data and cost and resource use data were searched for, in enough detail to be reproduced, with the search strategy supplied for quality of life and cost data. PRISMA-S adds that each database should be named with its platform, each strategy copied exactly as run and the deduplication process described.

Search filters for economic evaluations and utility studies

A search filter is a tested set of terms designed to retrieve one type of record, such as randomised trials or economic evaluations, and is combined with the topic terms. CRD's guidance notes that a specialist economic database needs only subject terms, while a MEDLINE or Embase search must add economic terms. For a brief economic commentary, Cochrane's economic evidence chapter recommends searching NHS EED and, for full economic evaluations, applying specialist filters in general databases from 1 January 2014, since NHS EED stopped being updated at the end of 2014.

The InterTASC Information Specialists' Sub-Group (ISSG) Search Filter Resource lists published filters for economic evaluations and utility values, without endorsing them, and warns that older filters should be checked against current indexing. Filter studies show the trade-off between sensitivity and precision. Glanville and colleagues tested filters against 2,070 full economic evaluations from NHS EED and found several with sensitivity above 0.99 in MEDLINE and Embase, but with low precision, perhaps 5 relevant records per 100 scanned. A NICE team built paired MEDLINE and Embase filters for cost-utility studies, whose sensitive pair reached 100% recall in validation and was estimated to cut screening by a factor of 5 against the broad economic search NICE had used for its guidelines. Arber and colleagues validated MEDLINE filters for utility studies with sensitivities of 88% to 95%.

Worked example adding Embase to a MEDLINE search for utility studies

All numbers in this example are illustrative. A modelling team needs utility values and judges its searches against a reference set of 40 eligible studies, using the Cochrane Handbook's definitions of sensitivity, also called recall, and precision.

$$S = \frac{a}{a+b}, \qquad P = \frac{a}{a+c}, \qquad NNR = \frac{1}{P}$$

where $a$ is the number of relevant records retrieved, $b$ the number of relevant records missed, $c$ the number of irrelevant records retrieved, $S$ the sensitivity, $P$ the precision and $NNR$ the number of records screened per relevant record found.

The MEDLINE strategy retrieves 1,200 records containing 30 of the 40 studies, and the Embase strategy 1,800 records containing 34. After deduplication, 900 records appear in both, including 26 of the relevant studies, so the combined set holds 1,200 plus 1,800 minus 900, or 2,100 records, and 30 plus 34 minus 26, or 38 relevant studies.

Adding Embase costs 900 more records for 8 more studies, or 112.5 records per extra study. The Handbook puts screening at a conservative 60 to 120 abstracts an hour, so the extra records take about 7.5 to 15 hours. If one of the 8 studies is the only source of utility values for a health state in the model, a MEDLINE search alone would have left that input to assumption or to a less relevant source.

Boundaries with nearby concepts

A literature search is the process, a search strategy is the set of terms, limits and filters, and the bibliographic database is the source the strategy is run against. Administrative and claims databases hold records of patients' care and are analysed as data, whereas a bibliographic database describes publications and the data stay in the reports. Cochrane treats trials registers, such as ClinicalTrials.gov, as a separate mandatory source, and lists grey literature and the reference lists of included studies as further sources.

Limitations and common misreadings

A database search finds what has been published and indexed. NICE's manual asks for attempts to find evidence not in the public domain, such as unpublished trial data, and the Cochrane Handbook notes that removing language limits in English-language databases is no substitute for searching non-English journals and databases.

A total of records retrieved is not a number of studies, since the same report appears in several databases and the same study in several reports, and low precision is expected in a sensitive search. A filter's published sensitivity applies to the database, platform and indexing it was tested on, and archive sources such as NHS EED describe the literature only up to their closing date.

Sources

• Lefebvre C, Glanville J, Briscoe S, et al. Chapter 4: Searching for and selecting studies (last updated March 2025). In: Higgins JPT, Thomas J, et al, editors. Cochrane Handbook for Systematic Reviews of Interventions, version 6.5.1. Cochrane; 2025. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-04
• Aluko P, Graybill E, Craig D, et al. Chapter 20: Economic evidence (last updated August 2023). In: Higgins JPT, Thomas J, et al, editors. Cochrane Handbook for Systematic Reviews of Interventions, version 6.5. Cochrane; 2024. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-20
• Centre for Reviews and Dissemination. Systematic Reviews: CRD's guidance for undertaking reviews in health care. York: University of York; 2009. https://www.york.ac.uk/media/crd/newpagesspring2025/Systematic_Reviews_CRD_guidance.pdf
• Centre for Reviews and Dissemination. CRD Database, About the databases and News (2015, 2018). https://www.crd.york.ac.uk/CRDWeb/AboutPage.asp
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). London: NICE; 2022, last updated 13 October 2026. https://www.nice.org.uk/process/pmg36
• National Institute for Health and Care Excellence. Single technology appraisal and highly specialised technologies evaluation: user guide for company evidence submission template (PMG24), appendices B, E, F and G. London: NICE; 2015, last updated 31 March 2026. https://www.nice.org.uk/process/pmg24
• US National Library of Medicine. About MEDLINE (last reviewed 2025); Medical Subject Headings. https://www.nlm.nih.gov/medline/medline_overview.html
• Center for the Evaluation of Value and Risk in Health. Cost-Effectiveness Analysis (CEA) Registry. Tufts Medical Center; accessed 2026. https://cevr.tuftsmedicalcenter.org/databases/cea-registry
• InterTASC Information Specialists' Sub-Group. ISSG Search Filters Resource. Accessed 2026. https://sites.google.com/a/york.ac.uk/issg-search-filters-resource/home
• Rethlefsen ML, Kirtley S, Waffenschmidt S, et al. PRISMA-S: an extension to the PRISMA Statement for Reporting Literature Searches in Systematic Reviews. Syst Rev. 2021;10:39. https://doi.org/10.1186/s13643-020-01542-z
• Bramer WM, Rethlefsen ML, Kleijnen J, Franco OH. Optimal database combinations for literature searches in systematic reviews: a prospective exploratory study. Syst Rev. 2017;6:245. https://doi.org/10.1186/s13643-017-0644-y
• Glanville J, Kaunelis D, Mensinkai S. How well do search filters perform in identifying economic evaluations in MEDLINE and EMBASE. Int J Technol Assess Health Care. 2009;25(4):522-529. https://doi.org/10.1017/S0266462309990523
• Hubbard W, Walsh N, Hudson T, et al. Development and validation of paired MEDLINE and Embase search filters for cost-utility studies. BMC Med Res Methodol. 2022;22:310. https://doi.org/10.1186/s12874-022-01796-2
• Arber M, Garcia S, Veale T, et al. Performance of Ovid MEDLINE search filters to identify health state utility studies. Int J Technol Assess Health Care. 2017;33(4):472-480. https://doi.org/10.1017/S0266462317000897