About this guide. This guide aims to make Quality Improvement filters easier to identify, understand, access, and implement in user customized searches.
In order to use this guide effectively, you will need to understand 1) basic searching in PubMed, using both MeSH terminology and keywords, 2) the use of search field tags in PubMed, 3) the use of built-in filters on the left sidebar of the PubMed Search Results page, and 4) how PubMed's Advanced Search page functions. Information in this guide builds on your understanding of these fundamental PubMed search features.
What is Health Services Research (HSR)? The Agency for Healthcare Research and Quality (AHRQ) outlines the broad benefits of HSR for patients in 3 major categories: "Health services research examines how people get access to health care, how much care costs, and what happens to patients as a result of this care. The main goals of health services research are to identify the most effective ways to organize, manage, finance, and deliver high-quality care; reduce medical errors; and improve patient safety." (Source: What Is AHRQ? Pamphlet, 2002. Available at: http://archive.ahrq.gov/about/whatis.pdf)
Are there tools to help me with HSR research? The National Library of Medicine's (NLM) web page “Finding and Retrieving HSR Tools and Databases" introduces researchers to a variety of tools for finding Health Services Research, including search filters that enable the quick retrieval of literature citations from PubMed for quality improvement projects.
What is a filter? Perhaps you are familiar with the built-in PubMed filters on the left sidebar of the search results page. These filters work by limiting citation search results by language, year of publication, article type, etc. Other filters, like the ones in this guide, limit search results by subject concepts. Subject filters are pre-fabricated search strings designed to capture citations according to the subjects referenced in them. Subject filters may be simple and contain only a few search terms or they may be complex and run on for pages.
Below is an example of a filter created by NLM to capture "Comparative Effectiveness Research" (Source: NLM's Resources for Informing Comparative Effectiveness page at: http://www.nlm.nih.gov/nichsr/cer/cerqueries.html).
If you click the link below, you will get a sense of how the filters in this guide have been automated to perform a PubMed search and then allow you to use your browser back button to return to the guide.
(pragmatic trial [tw] OR pragmatic trials [tw] OR practical clinical trial [tw] OR “comparative effectiveness” OR (promising [tw] AND unproven [tw]) OR resource allocation [majr] OR therapies, investigational [majr] OR agency for healthcare research and quality [mh] OR centers for medicare and Medicaid services [majr])
How does the PubMed search engine determine if a subject is referenced in an article citation? PubMed is a citation database made up of millions citation records. These records are constructed of searchable fields. Search fields may be designated by search field tags when coding a search (e.g., Folkman, Judah[author] codes the search for literature authored by the well-known cancer researcher). The searchable fields include titles, authors, author affiliations, abstracts, publishers, grant funding sources, dates of publication, Medical Subject Headings [MeSH], and more. The Medical Subject Headings [MeSH] field is particularly important to subject searching and subject filters.
The MeSH field in the citation record is created by a medical librarian who reads the original article, determines what it is about, and then uses standardized Medical Subject Headings to "tag" the article with the subject terms. MeSH terms are uniquely able to communicate what an article is about because they are chosen by a human reader for precisely this purpose. Article titles are often undescriptive and many articles have no abstract. In these cases, the MeSH tagging is invaluable and often the only way to retrieve the article.
Why are Medical Subject Headings important tools for searching? Medical Subject Headings (MeSH fields) are important because there are usually many different words for the same concept and it is impractical to remember them all. If you are able to find and include the standardized MeSH terms for a subject search, your chances of finding the right citations increases dramatically. Again, consider that many articles have no abstract or are ill served by an undescriptive title, and you will quickly realize the crucial role of MeSH indexing in citation retrieval. Many citations are only found because of their MeSH indexing. Note, if you are conducting a search in a broad area like quality improvement or patient safety, you are likely to need multiple MeSH terms to cover even a narrow sub-field.
Why do some filters use only MeSH terms while others include both MeSH terms with free-text terms? Search strategies and filters vary in their comprehensiveness and rigor, depending on researchers needs and resources (i.e., time to review citations) as well as the likely population of citations in the domain they are searching. Some subjects have thousands of citations entering PubMed every year, while other subjects have less than a hundred over the lifetime of the database. The same filter for a QI search might retrieve 100,000 results when paired with a high publication volume subject, while it would retrieve only a hundred results for a low volume subject. Searchers must always strike a balance and aim to be transparent in reporting about the kind of trade-offs they have made.
Generally, more comprehensive and better searches are built using both MeSH and free-text keywords. Searches are more current in their retrievals when they use free-text keywords. Free-text keywords target cutting edge jargon, emerging concepts that have not been assimilated into MeSH yet, and citations that have not yet been indexed. It is worth noting that not all articles are indexed and not all citation records contain MeSH fields. The delay in the indexing may be few months (common case) to several years (uncommon case), and free-text keywords bridge the gap between what has been just published and what is indexed. Finally, be aware of the timeline for entry of a term in to the MeSH. For example, patient safety became a MeSH term in 2012. If you do a MeSH-only search for articles in patient safety, you will miss 40 years of research published before the MeSH was created. The MeSH entry definition for a term will give you the year in which the term entered the MeSH database.
A middle path for busy searchers and informal filter designers, in domains where there is abundant publication, is to utilize all the applicable MeSH vocabulary and selectively add in specialized disciplinary jargon as free-text. MeSH searching takes advantage of the human indexing process and free-text searching takes advantage of key language in the title, abstract, or other term fields. Another practice in areas where there are many publications and a dearth of distinctive jargon is to rely exclusively on the MeSH indexing and avoid irrelevant retrieval due to the generic nature of available keywords. Researchers with limited time may miss a few citations, but are more likely to review more of what they do retrieve. It is also sometimes helpful to preview a search topic by using an extremely comprehensive MeSH and keyword search, but limit the results to systematic reviews or other strong reviews (e.g., an integrative review is usually thorough). By using a very complete MeSH and free-text filter for a subject or topical search, researchers maximize the likelihood of finding strong reviews to help them get a comprehensive overview of a topic without undertaking the inspection of too many single study citations or irrelevant citations.
It is helpful to use all of the non-topical filter components (e.g., language, publication dates, and study design limits) in a test of any master filter before making final determinations on the master filter's composition. For example, a search string component aimed at quality improvement terms may retrieve a million citations, but when this component is paired with a health condition component string and a study design limit filter, the results may be less than a few hundred citations.
Published filters as scaffolding. Filter creation is labor intensive because it involves reiterative searching and article index and abstract scanning to identify the right terms to include in a search string. Any time you can build on the work of others to create a better filter, it is well to do so. For example, the search strings used to retrieve citations for well-done systematic reviews may do double-duty as the first drafts for your filters. That said, a "good enough" filter for a QI project does not neccessarily need to rise to the level of systematic review, which must aim at perfect comprehesiveness by definition.
Topical and concept filters have been developed by large organizations to assist communities of researchers with common topics (e.g., cancer topics, quality improvement processes, study methodologies, etc.). These pre-fabricated expert filters provide a kind of search scaffolding, which enables others to achieve higher levels of efficiency and effectiveness than they might have otherwise. Increasingly, filters support research in areas where the literature is evolving rapidly and knowledge is less well organized (i.e., the organization of literature in a topic depends on its being indexing).
Filters create convenience in the first place by suggesting possible search terminology for specific subject areas. Since the MeSH database contains over 27,000 subject heading terms, the availability of lists of expertly identified MeSH terms is indeed a convenience. Often, more than one MeSH term may cover a topic and this has to be discovered by looking at the indexing of numerous articles to see the variations. MeSH terms may be superceded by new ones when the evolution of a concept in the literature makes it sensible to change the terms applied to it - yet, there is no retrospective indexing and it may be important to add the historical terms for a concept as keywords to the search string. It is also helpful to be aware of the history of jargon in a subject area. The current jargon for a concept may be very different than the jargon of 15 years ago when key or foundational studies on the topic were being done. Again, the best searches take advantage of the strengths of MeSH indexing and historically sensitive free-text to capture significant specialty jargon.
Clinical Queries - The National Library of Medicine's (NLM) well-known filters. The NLM's most well-known filters appear on the PubMed Clinical Queries (CQ) page. CQ filters illustrate the benefits of expertly prepared search strings applied to the problem of finding the highest level evidence in major clinical study categories, including diagnosis, therapy, etiology, prognosis, clinical prediction guidance, or medical genetics. Broad or narrow versions of CQ filters allow users to maximize either the sensitivity or specificity of their search. The CQ filters, developed by RB Haynes et al, are invaluable aids to busy clinicians who wish to check for high level evidence that may not have been assimilated into point-of-care evidence summaries or guidelines, which necessarily lag behind the most current research.
Filters In This Guide. Filters in this guide are drawn from two sources:
1) National Library of Medicine (NLM) web pages with topical filters of interest to quality improvement researchers
2) A technical report published by the Agency for Healthcare Research and Quality (AHRQ) in its series Closing The Quality Gap. This AHRQ filter was develped by a Stanford team (Shojania et. al.) to retrieve citations for Diabetic Management in the QI context. It will be of interest to quality improvement researchers in the area of chronic diseases. Additional filters can be acquired by investigating AHRQ publications and peer-reviewed systematic reviews.
Since the main NLM web page offers only a search box to automate its filters, users are unable see the coded search strings underlying these filters. By running a dummy search, it is possible to extract the filter component and make it available for editing.
To enable easy examination and customization of the filters, this guide presents filters in the form of the underlying coded search strings. Additionally, some filter strings have also been ungrouped to permit the selection of individual terms. Search strings can be copied to a Word document for easy customization and then be made into "Saved Searches" using MyNCBI tools.
Two tabs in this guide derive from the NLM's Health Services Research (HSR) PubMed Queries Filters (HSR-PQF) page:
1) The NLM-HSR-PQF FILTERS & SPECIAL TOPICS TAB contains both the broad (sensitive) and narrow (specific) versions of the Appropriateness, Process Assessment, Outcomes Assessment, Costs, Economics, and Qualitative Research filters, as well as the original, unmodified broad and narrow Quality Improvement filters. Several filters have been ungrouped to aid user customization. For convenience, the NLM's Health Literacy Search Filter has also been added to this tab. Definitions of these HSR filters are available on the HSR Study Design Definitions page.
2) The NLM HSR-QI-BROAD FILTER TAB contains a version of the broad version of the "Quality Improvement" filter from the NLM-HSR-PQF page. However, it has been placed on its own tab so it can be used as an ungrouped set of terms, allowing the user to customize a reduced search. The grouped set of terms in the broad QI filter may retrieve a burdensome number of citations when paired with certain health conditions.
The original NLM HSR-QI-BROAD filter is constructed from a single, high-level MeSH (Medical Subject Heading) term and second string aimed at retrieving high quality studies. The adapted HSR-QI-BROAD filter in this guide presents this sensitive NLM filter in an "exploded" version that shows the full expansion of subordinate MeSH terms beneath the single high-level term. Exploding the high-level term and ungrouping its component subordinate strings enables researchers to select only the component strings that are relevant. Searchers also have the option of using the study design search set designed for the origina HSR-QI-BROAD filter or using another study design filter of their own choosing. Additional study design filters are provided on the STUDY DESIGN FILTERS tab.
In addition to the NLM filters, the AHRQ-SHOJANIA FILTER tab presents another QI filter published by Shojania et al. as part of the AHRQ's technical report series on quality improvement. This filter has been partially deconstructed to allow for user customization where appropriate. The AHRQ-SHOJANIA FILTER is particularly useful for chronic health condition searches. The original AHRQ filter includes a component for study design filtering and this is included on the AHRQ-SHOJANIA FILTER tab as well as the STUDY DESIGN FILTERS tab.
STUDY DESIGN FILTERS TAB
The STUDY DESIGN FILTER TAB provides a number of filters designed to retrieve citations associated with specific study designs. These filters can be matched with QI searches AND Disease searches in order to narrowly target studies of interest. The study design filters represent the middle term (and only dispensable term) in a master search formula: QI filter AND STUDY DESIGN filter AND HEALTH CONDITION filter.
Please see the box below for a description of the master search string formula. Search strings can be copied to a Word document for easy customization and then be made into "Saved Searches" using MyNCBI tools.
Generally, a master search string is constructed of two or three sub-strings linked by the Boolean operator AND:
(QI String) AND (Study Designs String) AND (Health Condition String)
Sub-strings or component strings of the master search string may be constructed in a more or less complex and comprehensive fashion.
The guide tabs labelled AHRQ-SHOJANIA-QI FILTER, NLM-HSR-PQF FILTER FOR SPECIAL TOPICS, and HSR-QI-BROAD FILTER wlll help you build the QI portion of the master search string. The guide tab labelled Study Designs String will help you build the second portion of the master search string. The AHRQ-SHOJANIA study design component is also on the AHRQ_SHOJANIA_QI tab in a separate box. Finally, a tab labelled DISEASE STRING will help you create your own unique third component of the master search string.
The copyright notice associated with original filters should be considered as having effect with respect to derived work and attribution in publications. These filters are used in this guide for non-comercial purposes and in keeping with fair use requirements.
Filters derived from NLM are in the public domain. The NLM Copyright page does request that the National Library of Medicine be properly acknowledged as the source for its filters.
The NLM-HSR-PQF FILTERS FOR SPECIAL TOPICS and the HSR-QI-BROAD FILTER TABS:
- The filters on these tabs derive from the filters on the NLM's Health Services Research (HSR) PubMed Queries page, except where noted. Filters from this NLM page include the Appropriateness category, Process Assessment category, Outcomes Assessment category, Costs category, Economics category, Qualitative Research Studies category, and Quality Improvement Studies category.
- Descriptions of these filters are quoted from the NLM's Health Services Research (HSR) PubMed Queries: Study Design Definitions page. These search filters are based on the work of Haynes RB et al. Readers are advised to see the Clinical Queries Filters page on the NLM website for further details. Additional information is linked to the filter table on NLM's HSR Filter page as well.
- The Health Literacy Filter derives from the NLM's MEDLINE/PubMed Search and Health Literacy Information Resources page. It was created by Reference and Web Services Section of the National Library of Medicine.
Two filters on this tab derive from NLM's Health Services Research (HSR) PubMed Queries page and include the study design or "SCOPE" component addressed by the "Broad, sensitive search" and the "Narrow, specific search".
The filters labelled as Randomized Controlled Trials, Observational Studies (cohort, administrative data, registries, and electronic health records), and Systematic Reviews, Simulations, Models, derive from the NLM's Resources for Informing Comparative Effectiveness page.
AHRQ-SHOJANIA FILTER TAB
- The AHRQ-Shojania filter may have some restrictions with respect to copyright and commercial use. Users are advised to consider the published copyright notice in the event of downstream publication.
- Source Citation:
Shojania KG, Ranji SR, Shaw LK, Charo LN, Lai JC, Rushakoff RJ, McDonald KM, Owens DK.Diabetes Mellitus Care. Vol. 2 of : Shojania KG, McDonald KM, Wachter RM, Owens DK. Closing The Quality Gap: A Critical Analysis of Quality Improvement Strategies. Technical Review 9 (Contract No. 290-02-0017 to the Stanford University–UCSF Evidence-based Practice Center). AHRQ Publication No. 04-0051-2. Rockville, MD: Agency for Healthcare Research and Quality. September 2004. Accessed February 16, 2015 at:http://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0067768/pdf/TOC.pdf
"Health services research examines how people get access to health care, how much care costs, and what happens to patients as a result of this care. The main goals of health services research are to identify the most effective ways to organize, manage, finance, and deliver high-quality care; reduce medical errors; and improve patient safety."
(Quoted from the What Is AHRQ? Online pamphlet, 2002. U.S. Department of Health and Human Services Agency for Healthcare Research and Quality. Available at: http://archive.ahrq.gov/about/whatis.pdf)