Changes Related to MEDLINE in the Period between 1995 and 2004 that Affect its Usefulness for Various User Groups

by Shawn Thomas (Dec. 2005)

Advertisement

Table of Contents:

Introduction

MEDLINE ( Medical Literature Analysis and Retrieval System Online) is the National Library of Medicine’s bibliographic database for life sciences literature. MEDLINE contains over 13.5 million references, primarily to biomedical literature from scholarly, peer-reviewed journals. The most common means to access MEDLINE is PubMed (http://www.ncbi.nih.gov/entrez/query.fcgi). In addition to MEDLINE citations, PubMed includes OLDMEDLINE for pre-1966 citations, citations for out-of-scope articles from MEDLINE- indexed journals, in-process citations, citations that precede the date that a journal was selected for indexing in MEDLINE, some life sciences journals that are included in PubMedCentral but not MEDLINE, and certain physics journals that were part of a prototype PubMed in the early to middle 1990s. The increasing number of ways to retrieve information from MEDLINE and PubMed are ever more likely to go unnoticed by user groups that could benefit from a greater understanding of the options available to them. The present article will describe changes related to MEDLINE and PubMed in the ten year period from 1995 to 2004 that may reflect its usefulness for various user groups. In addition, potential solutions to problems brought about by these changes will be suggested.

General Statistics

In general, the usefulness of PubMed for all users increases as more and more citations are added to the database. In the period between 1978 and 2001, 8.1 million journal articles were indexed in MEDLINE (Druss & Marcus, 2005). An average of 272,344 articles were added to MEDLINE between 1978 and 1985 (Druss & Marcus, 2005). This average increased 46% to 442,756 articles per year for the period from 1994 to 2001 (Druss & Marcus, 2005). In 1995, 432,751 citations were added to the PubMed database. By contrast, 621,054 citations were added to PubMed in 2004. This record number of citations represented a 6.72% increase in the number of citations added compared to the previous year, when 581,947 citations were added to PubMed. Over the period 1995 to 2004, the average percent increase in citations relative to the previous year was 3.96%. Additional data about changes in the number of citations added to PubMed in the period from 1995 to 2004 are presented in Table 1.

Table 1

Citations Added to PubMed in the Period from 1995 to 2004

Year

Citations Added

% Change Relative to Previous Year

2004

621,054

6.72

2003

581,947

5.57

2002

551,266

3.67

2001

531,750

2.23

2000

520,165

8.63

1999

478,863

4.11

1998

459,959

4.27

1997

441,125

- 0.45

1996

443,143

2.40

1995

432,751

2.48

The introduction of free access to MEDLINE in early 1997 also increased the usefulness of MEDLINE on a grand scale. In January 1997, 163,000 MEDLINE searches were conducted (National Library of Medicine, 2005). During December of the same year, the number of searches performed had risen to 4.654 million for the month (National Library of Medicine, 2005). By September 2005, MEDLINE was searched 70.143 million times in a single month (National Library of Medicine, 2005). These amazing figures reflect the growing usefulness of MEDLINE, and increasing the usability of this resource for a variety of users groups is therefore a concern that deserves special consideration from the biomedical informatics community.

Non-English Speakers

Citations included in PubMed and MEDLINE predominately reference English language material. Thus, individuals who are fluent in English are likely to find these databases more useful than are non-English speakers. To shed additional light on this problem, Loria and Arroyo (2005) sought to characterize MEDLINE- indexed articles by language (English/non-English) and country of publication (Anglos/non-Anglos). They examined articles from the years 1966 to 2000 at five-year intervals. The results showed that the number of English language papers increased linearly by 9,199 papers per year and that the number of non-English papers decreased at a rate of 1,056 papers per year over the time period considered (Loria & Arroyo, 2005). Record lows for new non-English articles were observed for the years 1995 and 2000 (Loria & Arroyo, 2005). These findings are similar to those of Sousa Escandon et al. (2000; as cited in Loria & Arroyo, 2005), who found a 40% drop in non-English papers in MEDLINE from 1966 to 1999.

To determine if a similar trend for increasing numbers of English language articles is evident in the PubMed database, the percentage of newly indexed PubMed articles published in English was found for each of the years in the period from 1995 to 2004. In 1995, 87.64% of the articles indexed in PubMed were written in English. By 2004, this figure had increased to 90.02%, a difference of 2.38 percentage points. Full data for the years from 1995 to 2004 are presented in Table 2.

Table 2

Preponderance of Newly Indexed English Language Articles in the Period from 1995 to 2004

Year

English Language Articles Added to PubMed

% of Total Articles Added to PubMed

2004

559,052

90.02

2003

522,172

89.73

2002

495,201

89.83

2001

476,883

89.68

2000

465,345

89.46

1999

425,425

88.84

1998

405,625

88.19

1997

385,396

87.37

1996

390,146

88.04

1995

379,281

87.64

Thus, several sources suggest that a growing number of papers indexed in both MEDLINE and PubMed are written in English. What might be causing this phenomenon? Loria and Arroyo (2005) suggest at least two explanations. First, editorial policy changes in MEDLINE may favor journals from Anglo countries (i.e., published in the United Kingdom, Australia, Canada, Ireland, New Zealand, or the United States) and journals that publish solely in English (Loria & Arroyo, 2005). In addition, the editors of some journals from non-Anglo countries may have contributed to the shift by accepting papers only in English (Loria & Arroyo, 2005). Second, certain factors that affect non-Anglo researchers in non-developed countries may reduce these researchers’ capacities to submit non-English papers to MEDLINE- indexed journals (Loria & Arroyo, 2005). These factors could include publication constraints, migration, and a lack of sufficient support (Loria & Arroyo, 2005). Because of these considerations, Loria and Arroyo (2005) also examined the number of MEDLINE- indexed papers published by non-Anglo journals. On one hand, the number of papers from Anglo journals was found to increase by a factor of 4.2 between 1966 and 2000 (Loria & Arroyo, 2005). On the other hand, the number of MEDLINE- indexed articles from non-Anglo journals increased by a factor of just 1.5 during the same time period (Loria & Arroyo, 2005). In 2000, 68% of all MEDLINE- indexed papers were published in an Anglo country (Loria & Arroyo, 2005). The greater proportions of English and Anglo journals indexed in MEDLINE and PubMed may lend support to arguments that suggest that English has become the lingua franca of scientific discourse. Garfield (1998) has suggested that globalization is responsible for an emerging role for English as a scientific lingua franca:

While science was fundamentally always international in character, globalization has increased the dominance of English as the language of science. But this does not mean there is no longer a need for journals published in vernacular languages. However, the role of these local journals is, I believe, fundamentally changed. (p. 497)

While the existence of an apparent language of science makes databases such as PubMed substantially more useful to those fluent in the language, users who cannot read the language are left at a large disadvantage. Perhaps the best solution would be an international alternative to MEDLINE specifically for non-English articles, but such a solution would require an extensive amount of time and money.

Users Fluent in English that Lack a Life Sciences Background

Unfortunately, many English speakers are unable to understand much of the biomedical jargon present in articles’ abstracts and assigned Medical Subjects Headings (MeSH). Each year, the National Library of Medicine (NLM) introduces dozens of new Medical Subject Headings. In 2004, 666 new headings were added to the MeSH database for subjects such as 1-alkyl-2-acetylglycerophosphocholine esterase, emphysematous cholecystitis, and wigglesworthia. How can English speakers without a life science background begin to comprehend terms like “emphysematous cholecystitis?” First, the MeSH Browser at http://www.nlm.nih.gov/mesh/MBrowser.html offers PubMed users a “scope note” for each MeSH term. The scope note for the MeSH term emphysematous cholecystitis describes it as “A variant of acute cholecystitis with inflammation of the gallbladder that is characterized by the pockets of gas in the gallbladder wall. It is due to secondary infection caused by gas-forming organisms, and has a high risk of perforation.” In cases where MeSH scope notes are insufficient, users can enhance their vocabularies with a medical dictionary. The On-Line Medical Dictionary (http://cancerweb.ncl.ac.uk/cgi-bin/omd) is an excellent free dictionary produced by the Department of Medical Oncology at the University of Newcastle upon Tyne. Another useful resource is Stedman’s Online Medical Dictionary (http://www.stedmans.com). Additional links to biomedical dictionaries are available from MedBioWorld (http://www.sciencekomm.at/advice/dict.html).

Acronyms and abbreviations are additional possible areas of confusion for PubMed users without a life sciences background. The appropriate means to solve this problem are biomedical acronym and abbreviation databases. The Medi-Lexicon database (http://www.medilexicon.com) is perhaps the most well known database of this type. Two other databases, Biomedical Acronym Resolver (http://invention.swmed.edu/argh) and the Biomedical Abbreviation Server (http://bionlp.stanford.edu/abbreviation/) are also potentially useful resources. The latter two databases utilize a set of algorithms to automatically extract acronyms and abbreviations along with definitions from MEDLINE records. Chang et al. (2002) have described the methods used to create Stanford’s Biomedical Abbreviation Server.

Users Who Do not Search with Qualified Search Terms

A feature that all PubMed users should be aware of is automatic term mapping, which can potentially improve the search results of users unfamiliar with the use of MeSH terms in search queries. This feature matches unqualified search terms to qualified search terms with translation tables in the following order: MeSH translation table, journals translation table, full author translation table, and author index. The results of PubMed’s automatic term mapping can be viewed by clicking on the details tab. For example, an unqualified (i.e., without tags such as [All Fields]) search for adrenaline will be mapped to ("epinephrine"[TIAB] NOT Medline[SB]) OR "epinephrine"[MeSH Terms] OR adrenaline[Text Word]. The [TIAB] qualifier is used in a translation that allows one to search the Title and Abstract fields of non-MEDLINE indexed citations with a MeSH term that was mapped from the entry term. This enhancement to PubMed’s automatic term mapping was implemented on November 22, 2004 and was discussed in the Nov-Dec 2004 edition of the NLM Technical Bulletin. An additional enhancement to the automatic term mapping feature was implemented on April 10, 2003. In cases where a user searches for a phrase enclosed in double quotes, a hyphenated term, or a truncated term, automatic term mapping is not applied. Thus, a search for “adrenaline” in quotation marks will result in the search statement “adrenaline”[All Fields] rather than the mapped statement associated with a search for adrenaline sans double quotes. A search on December 1, 2005 returned 102,545 results for a search for adrenaline and 14,607 results for a search for “adrenaline.” Clearly, all PubMed users should be aware of how automatic term mapping affects their search results. More detailed information about PubMed’s automatic term mapping can be found at http://www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=helppubmed.section.pubmedhelp.Appendices On a related note, users who do search PubMed with MeSH terms should be aware of the fact that PubMed automatically explodes MeSH terms. For example, a search for the MeSH term “neoplasms” will also search for more specific MeSH terms that relate to neoplasms, such as “cysts” and “precancerous conditions.” Automatic explosion can be turned off with the tag [mh:noexp].

Users Who Desire an Alternative to the NLM PubMed Interface

Over the last ten years, a number of alternative interfaces to the standard PubMed interface maintained by the National Library of Medicine have been introduced. Over the same period, a major alternative interface known as ELHILL was discontinued. Thus, Internet Grateful Med, which was based on the ELHILL system, is no longer available. Some databases that were previously accessible via Internet Grateful Med can now be accessed from the NLM Gateway (http://gateway.nlm.nih.gov). A variety of alternative interfaces may interest users that dislike the standard MEDLINE interface. HubMed (http://www.hubmed.org) is one such interface that offers links to potentially useful services such as SFX by Ex Libris Inc. SFX is sometimes a useful means to find out about databases where an article’s full text can be retrieved. HubMed was previously the only service to offer RSS feeds for users to keep track of recent articles that match some search statement. However, the National Library of Medicine gave PubMed the ability to output RSS 2.0 (Really Simple Syndication) feeds on June 2, 2005. At this time, the RSS feeds available from PubMed seem preferable to RSS feeds generated by outside sources. However, users that prefer the Atom Syndication Format should continue to use HubMed, because PubMed does not currently support this alternative format. An RSS feed reader is necessary to view RSS feeds; a number of free options are available for download at software depositories such as Download.com (http://www.download.com).

Perhaps the most feature rich means to access the MEDLINE database is Ovid MEDLINE. After an advanced keyword search is performed, users are provided with a list of relevant MeSH terms to use rather than an immediate list of results; this feature insures that most users will use MeSH terms more effectively. Users can also check or uncheck boxes to turn off features such as auto explode and focus (i.e., limiting the results of a search to documents where a MeSH term is considered a major subject of an article). In other MEDLINE interfaces, the [majr] tag can be used to access the focus feature. Ovid MEDLINE users who use the advanced keyword search option are also provided with a list of subheadings that can help to restrict the scope of a search. Ovid MEDLINE is a subscription service, but a web-based demo is available at http://demo.ovid.com/libpreview/index.html Additional means to access MEDLINE citations via a subscription service include Dialog (http://www.dialog.com/products/productline/dialog.shtml) and PaperChase (http://www.paperchase.com). The latter service offers a free trial that requires registration.

Another useful MEDLINE interface is available at Infotrieve Online (http://www4.infotrieve.com/newmedline/search.asp). This service is free, but a subscription is required to access Infotrieve’s ArticleFinder feature, which offers access to citations to MEDLINE, ERIC, and other major databases. The full text of articles found via Infotrieve MEDLINE can typically be purchased for $12.00 in addition to a varying copyright royalty fee. The subscription only MEDLINE interface produced by the Community of Science (http://medline.cos.com) also utilizes Infotrieve’s document retrieval and delivery service. Users who wish to purchase articles can also try Aries Systems Corporation's UltraMED (http://www1.kfinder.com). UltraMed is a subscription service, but users can sign up for a free trial.

Some users may wish to use MEDLINE interfaces that are enhanced by advanced technology. One such interface is Akwanmed, a free service that allows users to input queries in languages such as Portuguese, Spanish, and English. Akwanmed’s most sophisticated feature is a document ranking system that permits users to order search results by either relevance or date. Users who wish to use document clustering to enhance their PubMed searches can try Clusty the Clustering Engine (http://clusty.com/search?v%3Asources=pubmed-limited), a free product of Vivísimo Inc. Clusty allows users to cluster results by a number of search- related topics, which are listed on the left side of a search results page. Clusty’s algorithms cluster documents based on textual and linguistic similarity and on a set of heuristics coded by Vivísimo’s programmers. Another MEDLINE interface that utilizes clustering technology is XplorMed (http://www.ogic.ca/projects/xplormed), which groups search results together according to their subjects. Users can then select groups of results on which to perform additional analyses. One such option for additional analysis is the “eXploring word context” feature, which allows users to click on a word ordered by an association score before being provided with a list of sentences from abstracts that contain the word. XplorMed users can search with a search statement, with a properly formatted file containing an array of abstracts, or with an identifier obtained from a database that offers links to PubMed citations.

Scirus (http://www.scirus.com) is a free scientific information service that offers a means to search multiple science databases at once. With Scirus, users can search journal sources such as MEDLINE and BioMed Central in addition to preferred web sources such as Cogprints and the arXiv.org e-Print archive. Another free service that may interest some users is Ijs MEDLINE search (http://www.ijs.co.nz/med/medline.htm). This service is essentially a javascript front-end to MEDLINE and other National Center for Biotechnology (NCBI) databases. Ijs MEDLINE search provides a variety of filtering options that may be of interest to users who are unfamiliar with MEDLINE’s manifold qualifier tags. Individuals that prefer to enter their own tags can access a table of tags and their descriptions at PubMed Help (http://www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=helppubmed.table.pubmedhelp.T37), an online book created by the NCBI.

Users Who Need to be Informed about Articles Recently Added to PubMed

The previously mentioned RSS format is a suitable solution for most users who need to keep up-to-date about recent additions PubMed. An RSS feed can be set up from PubMed by performing a search and then selecting RSS Feed from the “Send to” menu. A link is then provided that can be accessed by clicking on the XML icon. HubMed users can access RSS or Atom feeds by conducting a search and then clicking on the orange “feeds” button.

Users who do not wish to use RSS can opt to receive notices about newly added PubMed citations via e-mail. BioMail (http://www.biomail.org) is free service from the Medical Informatics Department at the State University of New York at Stony Brook. A grant from the National Library of Medicine supports new developments to BioMail. BioMail is released under a General Public License (GPL), and its source code is freely available from SourceForge (http://sourceforge.net/project/showfiles.php?group_id=3196). The primary alternative to BioMail is PubCrawler (http://pubcrawler.gen.tcd.ie). PubCrawler searches PubMed and Entrez Gene each day with a user’s search statements. The results are presented on an HTML page. When the page is updated with new articles, users receive a notification via e-mail.

Additional PubMed- oriented e-mail services are directed at specific user groups. Experts at these services review articles before they are suggested to users, which can save time for busy users who just want to know about the major developments in their fields. DailyUpdates from LeadDiscovery (http://www.leaddiscovery.co.uk/PubMed-dailyupdates.html) is an example of this type of service. DailyUpdates offers both a basic service for individuals involved in drug development and a premium package that focuses on specific therapeutic areas such as oncology; cancer immunotherapy; psychiatric, addictive, and sleep disorders; neurodegenerative and neuroelectrophysiological disorders; pain; cardiovascular disease; metabolic disorders; immunology; infectious diseases; genitourinary tract disorders; and technology. Another service of this type is Amedeo (http://amedeo.com). Amedeo sends users a weekly e-mail that contains a list of recently published articles that were screened from major journals. Amedeo’s users can elect to receive articles related to topics such as infectious diseases, cardiovascular disorders, respiratory system disorders, renal disorders, disorders of the gastrointestinal system, immunology, hematology, oncology, endocrinology and metabolism, neurologic disorders, psychiatric disorders, women’s health, neonatology, men’s health, dermatology, ophthalmology, otorhinolaryngology, surgery, dentistry, emergency medicine, intensive care, rehabilitation medicine, diagnostic procedures, nutrition, and substance abuse.

Users who need to keep track of additions to PubMed can also visit certain web pages to learn about new articles. Shawn Thomas has developed a set of automatically updated pages that display the recent contents of top tier journals (http://www.neurotransmitter.net/recentneuroscience.html). The pages are organized by topics that include neuroscience, psychiatry, anesthesiology, anatomy & morphology, development, immunology, general chemistry, organic chemistry, multidisciplinary chemistry, anthropology, surgery, nephrology & urology, dentistry, and allergy. Unlike most web pages that list recent additions to PubMed, Mr. Thomas’s offerings include abstracts from articles in addition to PubMed links.

Users Who Wish to Obtain Full Text Articles

Most users with institutional proxy access to PubMed should use that service. Links to full text articles are often available for institutional users when they display an article’s abstract. Some institutions offer links that allow users to order interlibrary loans as a means to obtain the full text of articles that are not available electronically. Users with institutional access may also wish to use the LinkOut option available from PubMed’s display menu. The LinkOut feature allows users to find different sources that an institution may offer for full text articles. Users whose institutions also offer Ovid MEDLINE may prefer it as a means to obtain full text articles.

Users without institutional access to PubMed may wish to use a document ordering service such as Infotrieve. When funds are unavailable, a variety of means to obtain free full text articles can be entertained. Users can add the search term free full text[sb] to a search statement in order to find free full text articles that are available to the public. Users with institutional access to PubMed should note that this search term does not return full text articles that are institutionally accessible but not publicly accessible. Many of the free full text articles indexed in PubMed are part of PubMed Central (http://www.pubmedcentral.nih.gov), the National Institute of Health’s free digital archive of biomedical and life science articles. PubMed search results can be restricted to articles in PubMed Central with the pubmed pmc local[sb] search term. The free full text[sb] search term returns both free full text articles from PubMed Central and free full text articles that are maintained on publishers’ web sites.

When the free full text of an article is not linked to from PubMed, users can try entering an article’s title into a search engine like Google. This method will occasionally turn up papers that authors have posted to their personal web spaces. To facilitate the use of Google to find full text articles, Shawn Thomas developed PubMed Gold, a free full text search engine (http://www.neurotransmitter.net/ftsearch.html). PubMed Gold finds PDFs for PubMed citations by automatically searching Google. PubMed Gold also displays links to free full text articles indexed by PubMed with some citations. Mr. Thomas has also created a series of web pages (http://www.neurotransmitter.net/fulltextinformatics.html) that are automatically updated when new records for articles from free full text journals are added to PubMed. There, users can view recent abstracts from free full text journals that relate to topics such as biology, molecular & cellular biology, neuroscience, genetics & proteomics, psychiatry, pharmacology, endocrinology, microbiology, cancer, biotechnology, and medical informatics & bioinformatics. When all these methods fail, users can e-mail an article’s author or publisher to find a means to acquire it.

Users Who Need Information Related to Cancer

MEDLINE is the premier source of information about articles related to cancer. Cancer researchers have a number of options available for entering MEDLINE search queries. Options include cancer[sb], cancer, "neoplasms"[MeSH Terms], and “cancer.” To assess how these different search statements affect the number of search results retrieved from PubMed, data was collected for each year for the period from 1995 to 1994. The search term cancer[sb] consistently returned the greatest number of results, with an average of 135,367.10 citations per year. An average percent change relative to the previous year of 4.32% was found for this search term. Additional data for this search term are available from Table 3.

Table 3

Citations Added to PubMed in the Period from 1995 to 2004 that Match the Search Statement Cancer[sb]

Year

Citations Added

% Change Relative to Previous Year

2004

165,830

5.13

2003

157,738

6.32

2002

148,363

4.43

2001

142,072

3.12

2000

137,774

6.62

1999

129,220

3.53

1998

124,811

3.97

1997

120,050

3.25

1996

116,274

4.25

1995

111,539

2.55

The search term cancer (sans double quotes) returned an average of 64,678.80 citations per year. An average percent change relative to the previous year of 4.02% was found for this search term. Additional data for this search term are available from Table 4.

Table 4

Citations Added to PubMed in the Period from 1995 to 2004 that Match the Search Statement Cancer

Year

Citations Added

% Change Relative to Previous Year

2004

78,968

3.92

2003

75,991

7.48

2002

70,703

5.27

2001

67,161

2.92

2000

65,255

6.20

1999

61,443

3.10

1998

59,597

3.64

1997

57,506

2.34

1996

56,189

4.10

1995

53,975

1.25

The search term "neoplasms"[MeSH Terms] returned an average of 60,483.80 citations per year. An average percent change relative to the previous year of 3.48% was found for this search term. Additional data for this search term are available from Table 5.

Table 5

Citations Added to PubMed in the Period from 1995 to 2004 that Match the Search Statement "Neoplasms"[MeSH Terms]

Year

Citations Added

% Change Relative to Previous Year

2004

72,057

2.22

2003

70,490

6.99

2002

65,885

5.24

2001

62,607

3.52

2000

60,479

5.39

1999

57,384

2.21

1998

56,141

2.97

1997

54,521

1.84

1996

53,537

3.48

1995

51,737

0.97

Finally, the search term "cancer” returned an average of 33,724.80 citations per year. An average percent change relative to the previous year of 6.96% was found for this search term. Additional data for this search term are available from Table 6.

Table 6

Citations Added to PubMed in the Period from 1995 to 2004 that Match the Search Statement “Cancer”

Year

Citations Added

% Change Relative to Previous Year

2004

45,962

8.20

2003

42,478

9.61

2002

38,754

6.42

2001

36,417

4.82

2000

34,744

8.87

1999

31,912

8.21

1998

29,491

7.40

1997

27,459

5.42

1996

26,048

8.61

1995

23,983

2.00

These results suggest that cancer researchers can obtain the greatest recall with the search term cancer[sb]. The search term “cancer,” which results in a low level of recall, should be avoided. Some researchers may also wish to combine the term cancer[sb] or “neoplasms”[MeSH Terms] with the term cancer using the Boolean OR operator to achieve an even greater recall. This advice is consistent with the suggestion by Jenuwine and Floyd (2004) that sleep researchers use both MeSH- based searches and free text searches for maximal retrieval. They observed a higher precision for subject searches (66% compared to 47% for text-word searches) and a higher recall for text-word searches (88% compared to 78% for subject searches) after searching a preselected set of research papers published in the journal Sleep (Jenuwine & Floyd, 2004).

Users who Practice Evidence-Based Medicine

A growing number of health care professionals have turned their attention to evidence-based medicine. Sackett et al. (1996) have provided the most commonly cited definition of evidence-based medicine:

Evidence based medicine is the conscientious, explicit, and judicious use of current best evidence in making decisions about the care of individual patients. The practice of evidence based medicine means integrating individual clinical expertise with the best available external clinical evidence from systematic research.

PubMed is the largest index of clinical evidence available. The randomized controlled trial has become the “gold standard” for clinical evidence used to judge whether a treatment does more harm than good (Sackett et al., 1996). Thus, changes to the number of reports of randomized controlled trials indexed in PubMed can have substantial effects upon the practice of evidence-based medicine. In the period between 1978 and 1985, 1.9% of all MEDLINE- indexed articles concerned randomized controlled trials (Druss & Marcus, 2005). In the period from 1994 to 2001, this proportion jumped to 6.2% (Druss & Marcus, 2005). A change in the top 5 MeSH terms has also been observed for these periods. From 1978 to 1985, the top 5 headings were mammals, proteins, diagnostic techniques/procedures, public health, and pathological processes (Druss & Marcus, 2005). From 1986 to 1993, the top Medical Subject Headings were mammals, proteins, public health, diagnostic techniques and procedures, and quality of health care (Druss & Marcus, 2005). Between 1994 and 2001, the top MeSH terms were public health, proteins, quality of health care, epidemiological methods, and mammals (Druss & Marcus, 2005). Druss and Marcus (2005) suggested that these changes reflect a shift “away from basic science headings toward topics related to clinical care and public health.”

In 1995, 10,365 reports of randomized controlled trials were added to PubMed. By 2004, this figure had risen 38% to 14,307. Over the period from 1995 to 2004, the average number of articles related to randomized controlled trials that were added to PubMed was 11,488.70 per year. In 1995, 50.55% of all clinical trial reports added to PubMed concerned randomized controlled trials. By 2004, this figure increased somewhat to 53.98%. For each year between 1995 and 2004, the average proportion of clinical trial reports that discussed randomized controlled trials was 51.57%. Thus, during most years, a slight majority of all clinical trial reports indexed in PubMed each year concern randomized controlled trials. Widening interest in evidence-based medicine may drive the proportion even higher. Further data about the addition of randomized clinical trial reports to PubMed for the period between 1995 and 2004 are available from Table 7.

Table 7

Citations Added to PubMed in the Period from 1995 to 2004 for Articles about Randomized Controlled Trials

Year

Citations Added for Articles about Randomized Controlled Trials

Percent of Clinical Trial Reports that Described Randomized Controlled Trials

2004

14,307

53.98

2003

12,610

52.34

2002

12,068

55.46

2001

11,881

56.06

2000

11,032

49.42

1999

11,446

48.90

1998

10,709

49.16

1997

10,313

48.39

1996

10,156

51.48

1995

10,365

50.55

To respond to the increasing interest in clinical trial reports, the NLM introduced a resource called PubMed Clinical Queries (http://www.ncbi.nlm.nih.gov/entrez/query/static/clinical.shtml). Here, users can easily search PubMed by clinical study categories such as etiology, diagnosis, therapy, prognosis, and clinical prediction guides. Users can choose a narrow, specific search or a broad, sensitive search. Users may also use PubMed Clinical Queries to search PubMed for systematic reviews or medical genetics literature. PubMed Clinical Queries is based on strategies devised by Haynes and colleagues (2005). Haynes et al. (2005) described two facts that were already known about search strategies for retrieving sound clinical studies: “Many clinicians and researchers conduct Medline searches independently but lack skills to do this well” and “A barrier to searching for evidence in Medline is the difficulty selecting an optimal strategy to search for information.” Interfaces that enable users to search PubMed more effectively are an important means to overcome these problems. Haynes et al. (2005) added three additional points of interest on the topic of optimal search strategies. First, they determined a number of strategies for finding scientifically sound articles about treatment studies, which have now been automated for use in the PubMed Clinical Queries service (Haynes et al., 2005). Second, they found that clinicians can utilize a very specific search to find a few sound articles on a subject. Third, Haynes and colleagues (2005) found that researchers can use a very sensitive search to carry out a comprehensive search for trials to be considered in systematic reviews.

Users Who Need Information about PubMed- Indexed Journals

Information about PubMed- indexed journals can have many uses. For example, an individual may wish to find quality journals that match some MeSH category. The NLM’s Entrez Journals database ( http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=journals) allows user to search for journals by title, MeSH term, language, or country of origin. Unfortunately, the Entrez Journals database includes entries for all journals that were ever indexed in MEDLINE. To ameliorate this issue, Shawn Thomas created the Neurotransmitter.net Journal Directory (http://www.neurotransmitter.net/journals/index.html). The Journal Directory includes the 4,302 journals that added a citation to PubMed during 2005. Journals included in the directory are organized by MeSH category and by name. The Journal Directory offers information about journals’ languages, abbreviations, NLM notes, and links to RSS and Atom feeds. Users can click on a journal’s NLM unique ID to view its NLM LocatorPlus entry. The NLM’s LocatorPlus service (http://locatorplus.gov) is the best way to find detailed metadata about a PubMed- indexed journal. Journal Directory users can also utilize the Neurotransmitter.net Abstract Retriever to find recent abstracts from each of the included journals.

Users who Need to Mine PubMed for Data and Information

Genetics researchers and other PubMed users who need to filter large amounts of data and information from the database have several available options. Options exist that allow users to mine data and information from PubMed, to cluster similar articles together, and to find potential links between disparate sets of literature. These resources often employ complex means to achieve their purposes, and interested users should first consult information about their methods to understand how they manipulate information extracted from PubMed. Articles and links to these resources are indexed at MetaDB (http://www.neurotransmitter.net/metadb/index.php?catid=65), a Metadatabase for the Biological Sciences created by Shawn Thomas.

Conclusion

An impressive number of developments related to MEDLINE and PubMed accumulated in the period from 1995 to 2004. In this period, MEDLINE access became free for the public, a number of resources meant to improve access to full text articles became available, and a variety of alternative means to access and retrieve information from the database were introduced. Clearly, the potential benefits of MEDLINE and PubMed for their users increased dramatically in the ten-year period. The challenge is now to insure that users understand how the new developments can help them to achieve their goals. MEDLINE and the various means to access it are incredible but increasingly complex resources; resources such as this document are ever more necessary to enhance the usefulness of MEDLINE for a variety of user groups.

References

Chang, J. T., Schütze, H., & Altman, R. B. (2002). Creating an online dictionary of abbreviations from MEDLINE. The Journal of the American Medical Informatics Association, 9(6), 612-20. Retrieved November 15, 2005 from the PubMed Central database.

Druss, B. G., & Marcus, S. C. (2005). Growth and decentralization of the medical literature: Implications for evidence-based medicine. Journal of the Medical Library Association, 93(4), 499-501. Retrieved November 15, 2005 from the PubMed Central database.

Garfield, E. (1998). The diverse roles of citation indexes in scientific research. (1998). Rev Invest Clin., 50(6), 497-504. Retrieved November 20, 2005 from http://www.garfield.library.upenn.edu/papers/rev_invest_clin_v50p497y1998.pdf

Haynes, R. B., McKibbon, K. A., Wilczynski, N. L., Walter, S. D., & Were, S. R. (2005). Optimal search strategies for retrieving scientifically strong studies of treatment from Medline: Analytical survey. British Medical Journal, 330(7501). Retrieved November 20, 2005 from http://bmj.bmjjournals.com/cgi/content/full/330/7501/1179

Loria, A., & Arroyo, P. (2005). Language and country preponderance trends in MEDLINE and its causes. Journal of the Medical Library Association 93(3), 381-385. Retrieved November 15, 2005 from the PubMed Central database.

Jenuwine, E. S., & Floyd, J. A. (2004). Comparison of Medical Subject Headings and text-word searches in MEDLINE to retrieve studies on sleep in healthy individuals. Journal of the Medical Library Association 92(3), 349-53. Retrieved November 15, 2005 from the PubMed Central database.

National Library of Medicine, Bibliographic Services Division. (2005). Number of Medline searches. Retrieved November 30, 2005 from http://www.nlm.nih.gov/bsd/medline_growth_508.html

NLM Technical Bulletin. (2004). PubMed's Automatic Term Mapping Enhanced. (341):e7. Retrieved November 20, 2005 from http://www.nlm.nih.gov/pubs/techbull/nd04/nd04_atm.html

Sackett, D. L., Rosenberg, W. M., Muir Gray, J. A., Haynes, R. B., & Richardson, W. S. (1996). Evidence based medicine: what it is and what it isn't. British Medical Journal 312(7023), 71-72. Retrieved November 20, 2005 from http://bmj.bmjjournals.com/cgi/content/full/312/7023/71

Sousa Escandon, M. A., Gonzalez Guitian, C., & Gonzalez Fernandez, M. M. (2000). ¿Que´ lengua hablara´ MEDLINE en el pro´ximo milenio? Arch Esp Urol 53(2), 93–99.