Greg R. Notess Reference Librarian Montana State University

ON THE NET Free MEDLINEs on the Web DATABASE, June 1998 Copyright © Online Inc.

For the past couple of years, free access points to the venerable MEDLINE database have been proliferating on the Web. The well-known health sciences abstracts database became the magnet for attracting customers to Web sites' other offerings, such as document delivery or virtual medical communities. The earliest form of these free MEDLINEs usually had a single search box with very few, if any, advanced features-no limits, field searches, or ability to browse MeSH. Questions about the comprehensiveness, currentness, and search features kept most information professionals wary of these free offerings.

The National Library of Medicine's (NLM) high-profile announcement of the free availability of PubMed in June 1997 was followed shortly thereafter by the announcement of free access to Internet Grateful Med without a password, thus legitimizing the use of the Web for free MEDLINE access.

MEDLINE AUDIENCE

The audience for almost every free MEDLINE on the Web is primarily end-users-healthcare practitioners, patients, and the general public. Thus, the emphasis in almost all versions is to make it easy to use, accessible, and not too complex. For the information professional and the frequent MEDLINE searcher, the lack of advanced searching options, limits, field searching, MeSH term and subheading searching, sort options, and dependable response time have made these free offerings much less attractive than commercial online and CD-ROM choices.

Do the free Web versions offer anything to the information professional? Are there advantages to some of these no-cost databases? Can they completely replace commercial versions? While there are many search features, database reliability issues, and other points of comparison, this article will focus on several issues including database coverage, registration requirements, advanced searching capabilities, document delivery options, follow-up searching, and sorting. For an online comparison which includes even more MEDLINE versions, see the MEDLINE Resource Center (http://omni.ac.uk/general-info/internet_medline.html).

AVICENNA

Avicenna requires registration, but there is no charge associated with registration. For advanced search options, Avicenna uses a form. This approach has separate lines to combine terms along with check boxes and drop-down choices for limiting, field searching, and Boolean operations. As with many of these systems, its advanced search form offers far more options than the basic search screen. Limiting to English language and only citations with abstracts is simple with the check boxes. The date limit is available in a drop-down menu and gives a clue to Avicenna's main defect. As of early 1998, the drop-down list of publication years did not yet include an option for 1998.

To search MeSH, use Avicenna's phrase search syntax of double quotes. Subheadings can be searched as well, but use the full name of the subheading instead of a code, enclose it in double quotes, and precede the subheading with a forward slash. For example, use "Behavior Therapy" "/psychology". Note that the subheading will not necessarily be tied to the MeSH heading, and there is currently no way to designate major or minor headings.

The dates of coverage is an important criteria when comparing these different access methods for MEDLINE. Of the free options reviewed here, Avicenna has the smallest MEDLINE database. It claims to cover 1990 to the present. But what is the update frequency? With no Dialog Bluesheet or detailed documentation specifying the update frequency, other techniques must be employed. Searching for the most current records from a major weekly publication like JAMA is one method. In early 1998, Avicenna included very few citations from any 1997 issue of JAMA. Closer investigation showed that the few issues included were about a year old, which means that the database in general is about a year old and badly in need of updating.

One of the great advantages of a Web-based search system is the ability to do follow-up searches directly from a bibliographic citation. In other words, when viewing a citation, a click on an author's name or a MeSH term should result in a follow-up search for that specific name or subject heading. Avicenna fails to take advantage of this option. Nor does Avicenna offer any connection to document delivery of full-text articles. Citations are listed in some perceived order of relevance, with no option to sort by the more commonly requested reverse chronological sort order.

BIOMEDNET

BioMedNet is another service that requires free registration. It offers Evaluated MEDLINE, which includes recommended articles that are selected and annotated. Occasional links are available from citations to full-text articles in the BioMedNet library, for a fee.

Advanced searching with field searching and Boolean operators is available in BioMedNet via its single search box. Fields are designated with a colon and a two letter code. For example, using asthma and 1997:dp limits the results to the publication date of 1997. The full list of field names is available by clicking the List of Fields link.

One of the nicest features of BioMedNet is its ability to combine sets and remember searches. This is an advantage of requiring registration. BioMedNet remembers search statements from previous visits. These are listed below the search box along with the number of hits and a set number. The set numbers can be combined with each other or with new terms. While this feature is standard in most commercial online and CD-ROM database systems, it is all too rare on the Web.

MeSH headings can be entered directly in the search box, or choose the Search MeSH button to find a MeSH heading which can then be used as a limit. The MeSH tree structure can also be browsed to select a heading. While these options work for finding a single MeSH heading, combining multiple headings works best by using the full heading with the :mh field qualifier. Subheadings can be searched in a similar fashion with the :mhql field qualifier and the two letter subheading code. Multiple word MeSH headings and any other phrases need to be surrounded with single quotes. Thus, a search on surgical treatment of retinal perforations would be 'retinal perforation':mh and su:mhql. There is no direct capability to be sure the subheading is attached to a specific MeSH heading, although some creative proximity searching may come close. BioMedNet does not have the ability to limit to major MeSH headings only.

Like most other MEDLINE options, BioMedNet's coverage goes back to 1966. A limit is available for specifying a lesser year span. As to the currentness of the database, a search for issues of JAMA found much more recent issues than Avicenna, but still about five months behind publication.

Of all the free MEDLINE options, BioMedNet is the only one that offers follow-up searching from citations. While it does not include all fields, the authors and MeSH heading are hyperlinks to new searches. Click on a MeSH heading to see the scope note and to limit searches to just that heading. In addition, BioMedNet uses PubMed's related record system to identify other similar citations.

BioMedNet offers a link in each citation for ordering the full text of documents from the British Library Document Supply Center (BLDSC) via BioMedNet. Also, on the initial search screen, the sort order of results can be specified as either relevance or date. While most of the free Web MEDLINEs only offer the relevance sort, which may or may not actually bring the most relevant documents to the top, this ability to sort by date (in reverse chronological order) is a welcome exception.

HEALTHGATE

HealthGate requires no registration and has similar date coverage to BioMedNet, from 1966 to about 5 months ago. In addition to MEDLINE, HealthGate can also search AIDSDRUGS, AIDSLINE, AIDSTRIALS, BIOETHICSLINE, CANCERLIT, and HealthSTAR. The Advanced Page offers more search options than the initial screen, but both offer easy check boxes to limit results to English only, citations with abstracts, and the most recent two years' publication dates.

On the Advanced Page, a form is used for combining terms with Boolean and proximity operators and a few field searches are available: title, abstract, MeSH, and major MeSH. Other limits available on the Advanced Page include specifying the year range, limiting to one specific age range, choosing one publication type, and journal subsets. No options for specifying attached or free floating subheadings is available.

HealthGate uses a system known as ReADER that attempts to map search words to MeSH. For example, a search on lung cancer would actually be searched as lung neoplasms and treatment for lung cancer would be searched as lung-cancer-th. Once again, this is a technology designed for end-users, and may be effective for them. The information professional can use the option on the Advanced Page to turn ReADER off.

While HealthGate does offer document delivery, it does not offer any kind of follow-up searches from citations. The search results are sorted by relevance with no other sort option available.

INFOTRIEVE

Infotrieve is primarily a document delivery company. Thus, their offering of free MEDLINE is designed to make it easy to find citations to order. No registration is required to search MEDLINE, but it is required for document ordering. Infotrieve's MEDLINE covers from 1966 to about five months from the present (based on the JAMA search). Advanced searching is available using their form and/or Boolean operators to combine field names in the same style as BioMedNet ('choanal atresia':mh and 1997:dp). Phrases should be in single quotes and the operator can be in uppercase or lowercase. No follow-up searching is available, and the only sort is by relevance.

There is no option for browsing MeSH nor is there even a field for MeSH subheadings. The only field for MeSH does not distinguish between major and minor headings. However, the search form does make it easy to limit by language, article type, age group, and human subjects.

MEDSCAPE

Medscape requires registration, but it also can save registration information on an individual's computer as a cookie, making it easier to connect each time. Medscape offers AIDSLINE and TOXLINE. Its MEDLINE coverage dates from 1966 up to about five months ago, based on the same search for most recent JAMA issues. While natural language searching is the default, advanced searching is possible by using Boolean operators and field tags like BioMedNet and Infotrieve, although phrases must be designated with double quotes rather than single. No field for subheadings, major MeSH, or minor MeSH is available. Medscape automatically truncates to the root of a term, so that a search on infection is actually a search on infect*. Medscape offers document delivery, but it does not have any follow-up searching and only sorts by relevance.

PUBMED AND INTERNET GRATEFUL MED

PubMed grew out of the National Center for Biotechnology Information's development of the Entrez databases for researchers. These genetics databases, including GenBank and the Online Mendelian Inheritance in Man, continue to be connected to some citations in PubMed. Internet Grateful Med (IGM) is the Web successor to Grateful Med, the end-user front end to MEDLARS. Neither of NLM's offerings require registration.

In addition to being official NLM products, PubMed and IGM are the most current free MEDLINEs. Both go back to 1966, and include access to the very current PREMEDLINE. In addition, PubMed includes all articles from several journals that are generally only indexed selectively for MEDLINE. On the JAMA test, PubMed and IGM were by far the most current of all the MEDLINEs. Rather than the five-month lag time, they both included citations to articles from the previous week's issue of JAMA. Note also that while most MEDLINE incarnations give a starting date of 1966, they may well include older publications. Both PubMed and BioMedNet include citations in the early 1960s and a few published in 1959. In IGM, choose the Search Other Files button and pick OLDMEDLINE for citations from 1962-1965.

Other databases available in IGM include HISTLINE, AIDSLINE, HSRPROJ, AIDSDRUGS, AIDSTRIALS, PREMEDLINE, DIRLINE, SDILINE, HealthSTAR, and surprisingly, PubMed. PREMEDLINE is rolled into MEDLINE on PubMed, but it also offers search access to its genetic databases under the names of GenBank DNA Sequences, GenBank Protein Sequences, Biomolecule 3D Structures, and Complete Genomes.

The advanced search features in PubMed differ markedly from the other MEDLINE versions. The Advanced option allows field searching from field labels available in a drop-down list. To combine multiple terms or limits, PubMed uses a sequential approach rather than a form or a single search box. Available fields include MeSH terms, major MeSH, and subheadings. PubMed even makes it possible to search subheadings of specific MeSH. Choose MeSH terms and then select a Search Mode of List Terms rather than Automatic. Then after entering the subject term, PubMed displays a list of MeSH terms with associated subheadings and even the number of postings for each combination. PubMed will also accept field names, using the full name, in square brackets after a search term: for example, Human[MeSH Terms]. These can be combined, but the Boolean operators must be in all uppercase. Most limits are available using these methods, but the simple check boxes available on forms would make it much easier. The forms approach is exactly what IGM uses. It offers subject, author, and title word searches that can be combined with an AND. Another button is used to add an OR. At the bottom of the form is one of the more complete sets of limits including language, publication type, study group, gender, age group, journals, and publication year.

Grateful Med uses the UMLS Metathesaurus to map subject terms to appropriate MeSH terms. If this fails, the Find MeSH/Meta Term button can help provide access to the proper MeSH headings as well as access to the tree structures. To search subheadings, add the full subheading in a separate subject box.

The speed, or lack thereof, of IGM is one of its major problems. If response time on IGM is too slow, try logging in again. Sometimes that can result in speedier access. Note that occasionally IGM will not respond well if you use the Web browser back button. Use the IGM buttons instead, when possible.

Both PubMed and IGM offer document delivery, but since it is via Lonesome Doc, it is not as convenient for the general public as some of the other services. Neither PubMed nor IGM yet offer direct follow-up searching. PubMed does feature an opportunity to find related articles, but the criteria are PubMed defined rather than user defined. The ability to click on an author's name and find all other articles by that name and the option to click on a MeSH heading to find other postings for that term are both features that ought to be easy to add to any Web-based system.

To their credit, both IGM and PubMed sort their output in reverse chronological order. Neither use the "relevance" sorting so common among the other systems. Given all of the preceding criteria, which of all of these MEDLINEs is best for the professional searcher and how do they compare to their commercial counterparts?

DUMP THE COMMERCIAL VERSIONS?

As of yet, there are no easy answers. Personal preference will remain a strong determinant. Free MEDLINE is definitely an important resource for our patrons. Many of the free versions offer some interesting and unique features even for the information professional. Except for Avicenna, all the MEDLINEs reviewed here are at least relatively current and seem to be regularly updated, although for any search requiring the most current information possible, PubMed and IGM are the obvious choice and can well be more current than local CD-ROM or even online commercial versions.

As to advanced search features, most of these systems can be coaxed into giving appropriate results, but it takes time to find the hidden keys. The ability to combine MeSH terms directly with subheadings is generally not available; the exception is PubMed. The ability to combine sets is only available on BioMedNet. And there are plenty of other advanced search, display, and output features that are completely absent from the free Web versions. In general, the commercial versions are all still more powerful, have more advanced search features, and offer more reliable response time.

Any library trying to connect MEDLINE citations to local holdings messages will need to continue using a commercial version for at least the near term future. A searcher needing to search multiple files, using complex search statements, and demanding a reliable response time, will continue to be better served by the online commercial systems. Commercial Web-based MEDLINEs like those from the Community of Science (http://www.cos.com) and Knowledge Finder (http://www.kfinder.com) also have marketplace niches to fill.

On the other side, our users and clients are likely to rely more and more on the free MEDLINEs and come to us with questions and problems. With the rapid change in search interfaces made possible by the Web and today's computing platforms, it is a fascinating time to watch the changes in MEDLINE access. At the same time, it can be extremely frustrating for those who rely on the database on a daily basis when they try a free Web version.

With so much interest in the MEDLINE database, we should see steady improvements in these interfaces. By sending in our feedback, suggestions, and complaints, information professionals can help influence and guide those improvements.

Comparison of Features of the Free Versions of MEDLINE

Communications to the author should be addressed to Greg R. Notess, Montana State University Libraries, Bozeman, MT 59717-0332; 406/994-6563; align@montana.edu or http://imt.net/~notess/.