Filters (ie. hedges)

From HLWIKI Canada

Jump to: navigation, search
Typical Medline search using a combination of MeSH & search filters (ie. hedges)
Are you interested in contributing to HLWIKI Canada - hlwiki.ca? contact: dean.giustini@ubc.ca

To browse other articles on a range of HSL topics, see the wiki index.

Contents

Introduction

See also Canadian search filter for MEDLINE and Systematic review searching

Filters (ie. hedges) or search filters are defined as: "...a [single] search term or terms (such as 'randomi?ation' or 'random allocation' for RCTs) that select studies that are at the most advanced stages of testing for clinical application" (Wilczynski et al 1995, p436). In other words, "...search filters or hedges consist of a combination of terms, generally drawing on free-text, controlled vocabulary and metadata."
Search filters are often described as predefined strategies to improve recall and retrieve maximum recall of primary research of gold standard studies i.e. randomized trials (RCTs), systematic reviews, meta-analysis etc. and clinical queries such as diagnosis, prognosis, etiology and therapy. Typically, search filters are used to refine search results, particularly for complex topics with multiple synonyms, problems with polysemy and so on. In terms of their development, filters are created by combining Medical Subject Headings (MeSH) with a series of keywords and wildcard variants to optimize retrieval. (See McMaster University's HIRU Project). Filters are also referred to as hedges, PubMed Clinical Queries, or optimal search strategies. Filters are sometimes referred to as methodological search strategies as they are composed of words that relate to study methodology. For example, a randomized control (RCT) filter may contain randomi*ed and clinical trial as phrases or separate keyword occurrences. A diagnosis filter might use sensitivity, specificity and so on. (For more information about search filters, see InterTASC Information Specialists' Sub-Group. Search filter resource.)

See McKibbon KA et al. Retrieving randomized controlled trials from Medline: a comparison of 38 published search filters. Health Information & Libraries Journal 2009;26(3):187-202.

The aim of using a search filter

Search filters were first devised by Haynes et al and K. Ann McKibbon (a Canadian health librarian) at McMaster University in the early 1990s. The investigators entered search terms that related to diagnosis, prognosis, therapy and etiology and put them into a computer program to identify optimal search strategies. They were tested and validated in MEDLINE and compared against a 'gold standard'. Further studies have tested and validated search filters in several areas relating to research methodology and evidence-based practice. It must be said that search filters are not a guarantee of retrieving quality research or complete recall in a given area of biomedical research. While they retrieve studies related to research methodologies, the onus continues to remain on information retrievalists and the principle investigators to ensure that all evidence has been cumulated and that it be put through an evaluation process to assess its quality, relevance and applicability to the clinical question at hand.

Filter types

See Canadian search filter for MEDLINE

Many filters are interface-specific (OVID, Dialog, PubMed), database-specific (MEDLINE, CINAHL, EMBASE, PsychInfo) or focussed on publication types (RCT, systematic reviews, diagnosis, etc.). Filters are designed to be highly-sensitive, precise or as one-line strategies.

RCT filter for OVID (Dickersin K et al, 1996) see also http://www.cancerview.ca/portal/server.pt/community/Search%20Filters/528/Search%20Filters/5661

  1. RANDOMIZED CONTROLLED TRIAL.pt.
  2. CONTROLLED CLINICAL TRIAL.pt.
  3. RANDOMIZED CONTROLLED TRIALS.sh.
  4. RANDOM ALLOCATION.sh.
  5. DOUBLE BLIND METHOD.sh.
  6. SINGLE-BLIND METHOD.sh.
  7. or/1-6
  8. ANIMALS.sh. not HUMAN.sh.
  9. 7 not 8
  10. CLINICAL TRIAL.pt.
  11. exp CLINICAL TRIALS
  12. (clin$ adj25 trial$).ti,ab.
  13. ((singl$ or doubl$ or trebl$ or tripl$) adj25 (blind$ or mask$)).ti,ab.
  14. PLACEBOS.sh.
  15. placebo$.ti,ab.
  16. random$.ti,ab.
  17. RESEARCH DESIGN.sh.
  18. or/10-17
  19. 18 not 8
  20. 19 not 9
  21. COMPARATIVE STUDY.sh.
  22. exp EVALUATION STUDIES
  23. FOLLOW UP STUDIES.sh.
  24. PROSPECTIVE STUDIES.sh.
  25. (control$ or prospectiv$ or volunteer$).ti,ab.
  26. or/21-25
  27. 26 not 8
  28. 27 not (9 or 20)
  29. 9 or 20 or 28

Handy tools

OVID Commands

To help you understand the filter above, you will need to be aware of various commands in the OvidSP interface. For example, the OVID command for subject headings is the forward slash (placed after the subject heading) - "common cold/". In this case, the MeSH is chosen to retrieve all items indexed with the appropriate subject heading assigned to it by indexers:

  • .exp. Explode retrieves all items with the subject heading and all associated narrower MeSH terms assigned
  • .ti. Title retrieves items with words contained in the title
  • .ab. Abstract retrieves items with words contained in the abstract
  • .mp. Keyword retrieves items with words contained in the title, abstract or subject heading
  • .pt. Publication type retrieves items of a specified publication type (e.g. Multicenter study)
  • $ Truncation replaces any number of characters including zero (e.g. child$ will retrieve items containing child, children, childhood, etc.)
  • adj Adjacency operator retrieves items with query terms on either side in the specified order
    • N.B. Other systems may use different commands, refer to the 'help' page in the particular system/database in which the search is conducted for further explanations.

Why use a search filter?

Search filters are used to limit topics with a voluminous literature or when searches have high recall beyond what would be reasonable to process by hand. Certain segments of the medical bibliography such as oncology, genetics and cardiology are areas with a voluminous literature. In addition, search filters apply a level of consistency to searching that make it much more systematic and, if applied well, will save researchers a considerable amount of time in cumulating and evaluating the relevant literature for the clinical question at hand.

When filters or hedges are required

  • To limit large search results or when looking for a particular publication type (RCTs, etc) or clinical query (diagnosis, etc).
  • To retrieve quality research for advanced or expert searching
  • To conduct highly sensitive searches combining MeSH and free text.
  • To check that filters are relevant to database, study design, etc.

To avoid typographical errors highlight each line in turn, copy and paste each line of the search into the database and perform search. Alternatively type each line in manually. (Take care to note any changes in the written numbers of search statements when combining with 'OR'/'AND' e.g. 'or/10-17', which may occur as a result of adding on the search filter to your subject search. To avoid confusion it may be easier to type in the filter first and save it, then start a new subject search and re-run the saved filter at the end).

Use 'AND' to combine the results of the subject search with the filter. Although filters may be downloaded to disk, at this time they cannot be uploaded from disk into OVID and therefore they have to be typed in manually. For frequent users, type in the filter and save it as a permanent search. The filter can then be re-run as required. If in doubt seek help from a librarian/information professional.

Search filters are not a guarantee of retrieving the best evidence. A recent article by Glanville et al (2008) looked at the challenges posed by so many extant search filters and how health librarians can manage them using checklists and appraisal tools. Obviously, search results will need to be appraised for quality and relevance against the research or clinical question. Filters developed in MEDLINE, for example, will not have the same sensitivity/specificity in a database reflecting a different focus. One example is EMBASE because its indexing practices are different from what we see in CINAHL and MEDLINE. Due to the annual updates, MeSH terms must be checked to see if any new terms have been created. Specific types of filters, such as RCT or systematic review filters, do not necessarily retrieve all relevant clinical trials or systematic reviews. Due to the limitations of human indexing, some major or even seminal articles may be missed. Therefore, reference harvesting and hand-searching are also recommended.

See also

References
Retrieved from "http://hlwiki.slais.ubc.ca/index.php/Filters_(ie._hedges)"
Personal tools