The great contribution: Index Medicus , Index-Catalogue , and IndexCat

Objective: The systematic indexing of medical literature by the Library of the Surgeon-General’s Office (now the National Library of Medicine) has been called ‘‘America’s greatest contribution to medical knowledge.’’ In the 1870s, the library launched two indexes: the Index Medicus and the Index-Catalogue of the Library of the Surgeon-General’s Office . Index Medicus is better remembered today as the forerunner of MEDLINE, but Index Medicus began as the junior partner of what the library saw as its major publication, the Index-Catalogue . However, the Index-Catalogue had been largely overlooked by many medical librarians until 2004, when the National Library of Medicine released IndexCat, the online version of Index-Catalogue . Access to this huge amount of material raised new questions: What was the coverage of the Index-Catalogue ? How did it compare and overlap with the Index Medicus ? Method: Over 1,000 randomly generated Index Medicus citations were cross-referenced in IndexCat. Results: Inclusion, form, content, authority control, and subject headings were evaluated, revealing that the relationship between the two publications was neither simple nor static through time. In addition, the authors found interesting anomalies that shed light on how medical literature was selected and indexed in ‘‘America’s greatest contribution to medical knowledge.’’


INTRODUCTION
William Henry Welch , the great pathologist and bibliophile, is famously quoted as saying that the systematic indexing of medical literature, begun by John Shaw Billings at what was then the Library of the Surgeon-General's Office, United States Army (known today as the National Library of Medicine [NLM]), was ''America's greatest contribution to medical knowledge.''* He apparently was sincere in this belief. In fact, he made this statement several times, to different audiences, and in such varied contexts that it is not easy to be certain just what Welch was praising: Billings's commitment to indexing or the two products of this commitment, Index Medicus and the Index-Catalogue of the Library of the Surgeon-General's Office, United States Army.
Index Medicus may be better remembered today (if somewhat hazily) as the forerunner of MEDLINE and now PubMed. But Index Medicus began its publication career (in 1879) as very much the junior partner of what Billings saw as his library's major publication: the Index-Catalogue, whose first volume appeared in 1880, but whose genesis preceded Index Medicus by several years. In the experience of the authors of this article, the Index-Catalogue had been largely forgotten by many older medical librarians and never encountered by younger ones, primarily because it just was not available in most collections (though it had continued to be a major tool for history of medicine specialists) until 2004, when, after years of work with a number of partners, NLM released IndexCat, the online version of the complete Index-Catalogue. { The release of IndexCat provided free and readily available access to a huge amount of material (over three million citations, many available or searchable nowhere else) but also raised a number of new questions. What had been the coverage of the Index-Catalogue? How did it compare with Index Medicus? How great was the overlap, and why did the overlap happen? Does the conscientious librarian or researcher need to search both? This paper seeks to answer these questions.{ In addition, the paper will discuss some interesting anomalies found in the process of addressing these questions that shed a great deal of light on how medical literature was selected and indexed in ''America's greatest contribution to medical knowledge.''

HISTORY OF THE INDEX MEDICUS AND INDEX-CATALOGUE
According to Miles, the origins of the Index Medicus and Index-Catalogue go back to the 1870s, when Billings first became interested in indexing the library's serials. He had already published catalogues of its books, but these were author-title catalogues, which could be very frustrating for the researcher. Billings himself had had a very difficult time in 1859-1860, compiling a bibliography for his thesis on the surgical treatment of epilepsy. He wrote about it years later, and it may have been the defining moment of his life [1].1 At any rate, Billings began to develop a prototype for his new catalogue. In 1875, he published a 316-page, 18,000-entry bibliography on cholera. It added great prestige to the library's name and gave Billings leverage to move on to his next project. He prepared a sample volume of his proposed index, the Specimen Fasciculus, sent copies to everyone who might support the idea, and went to Congress to ask for money [2]. In addition to obtaining funding from Congress, Billings requested and received an assistant from the US Army: Robert Fletcher, who would stay at the library until his retirement in 1911 at the age of 88. It was the team of Billings and Fletcher who would make the Index-Catalogue and Index Medicus realities.
Looking at the slender Fasciculus volume (it is only seventy-two pages long), it is easy to see that Billings had already determined what he wanted. The scope and format of the Index-Catalogue are already in place. However, it would not have been so obvious that the end product of their labors would be so large and that production would be so painstaking and slow. It took fifteen years (1880-1895) to produce the first series of Index-Catalogue. Billings celebrated the event by retiring and moving to New York, where he became the first director of the New York Public Library. Fletcher stayed behind, and work on the second Index-Catalogue series began.
The long time lag inherent in the production of the Index-Catalogue was always its curse. If an article on the abdomen appeared just after the ''A'' volume went to press, it would be a decade or more before the new citation would appear in print. Eventually, the backlog would prove to be the Index-Catalogue's undoing. By the 1950s, with the fourth series incomplete and the need for a fifth series looming, the controversial decision was made to cease publication of the Index-Catalogue. Even though the library stopped new indexing for Index-Catalogue in March of 1950, so huge was the backlog that print volumes continued to appear until June 1961 [3].
Billings did not anticipate this backlog when he presented the Fasciculus to Congress, but he and Fletcher quickly realized the issue when production began. Fletcher devised a plan to combat it: they would publish a second index, a ''monthly classified record of the current medical literature of the world'' (the official subtitle of the Index Medicus as it appears on the title page of volume 1 in 1879). It was Fletcher who suggested that it be called Index Medicus [4].
From the outset, the two publications had fundamental differences. First of all, Index-Catalogue was to be the ''official'' publication, detailing the actual holdings of the library, be they books, journal titles, individual articles, theses, or even portraits. Index Medicus would be far narrower in scope, focusing on new articles from selected journals, selected new books, and theses. A journal list was included in the first volume of the Index Medicus, but subsequent volumes generally lacked such a list.
Functionally, however, the greatest difference between the two publications was that the Index-Catalogue was a government publication and Index Medicus was not. For its entire run, the Index-Catalogue was published by the Surgeon-General's Office of the US Army, while Index Medicus was privately published by a series of small publishers, who had difficulty making the work profitable. In this period (1879-1926), the Index-Catalogue had a secure source of funding, while Index Medicus was expected to be selfsupporting. As Billings wrote in his introduction to the first volume of Index Medicus: It has often been suggested that it is highly desirable that [the Index-Catalogue] should be supplemented by some current publication, which should show all recent works, together with articles in periodicals, arranged by subjects [emphasis in the original], but until quite lately no proper means have been available for such an undertaking. Now, however, Mr. F. Leypoldt, of New York City, proposes to undertake the publication of such a current medical bibliographical serial. [5] N During the period under discussion, the overlap of articles appearing in Index Medicus and Index-Catalogue varied between 96.6% and 46.9%. The role of Leypoldt has been described in different ways,** but it is undeniable that, between 1879 and 1926, Index Medicus had a number of publishers, including Leypoldt in New York, George Davis in Boston, and the Carnegie Foundation in Washington. There was even a period (1899-1902) when publication of Index Medicus ceased and was briefly replaced by a Paris publication called the Bibliographica Medica. There were also years, such as 1895-1899, when the title page mysteriously read only ''Published by the Editors, New York and Boston.'' Sometimes (as in 1879), publishers were listed for London, Paris, Leipzig, Amsterdam, and St. Petersburg, and wouldbe contributors were advised to submit their publications to these Europeans offices for inclusion in Index Medicus. A certain collection development objective is implied here: the introductory letter already cited made it clear that, after indexing, the publications would be added to the collections of the library.
It should be noted that 1926 was the last year that Index Medicus was published in its original format and so is the last year used for comparison in the current study. For 1927, Index Medicus was merged with a publication called the Quarterly Cumulative Index to Current Literature, begun by the American Medical Association (AMA) in 1916. The new publication, named the Quarterly Cumulative Index Medicus, was published jointly by the Army Medical Library and the AMA, with some funding from the Carnegie Foundation, until 1956. In scope, structure, and format, it was a fundamentally different publication from its predecessor, as can be quickly seen by a sideby-side comparison.
Through these years, however, the Index-Catalogue marched on: series 1, 16 volumes, 1880-1895; series 2, 21 volumes, 1896-1916; series 3, 10 volumes, 1918-1932; series 4, 11 volumes, 1936-1955. As already noted, the increasing volume of medical literature eventually overwhelmed the library's processing capacity, but in the days of the original Index Medicus (that is, 1879-1926), the library prepared citations for both series, shouldering both tasks with a very small staff. The work must have been staggering. The authors have found that Billings solicited citations from publishers worldwide in the name of the Index-Catalogue, and he obtained permission from the US Army surgeon-general to send citation cards prepared for Index-Catalogue for use in Index Medicus [6]. Library staffers were paid by the Index Medicus publishers to make duplicate cards.
This processing workflow implies a considerable amount of overlap between the two publications and brings up the original questions: how much overlap was there? Must a present-day researcher check both indexes? NLM's decision to digitize the Index-Catalogue raised these questions anew and coincidentally provided the tools required to answer them by allowing a random sample of citations from Index Medicus to be extracted and searched for in the IndexCat database.
Before presenting the results of this analysis, it is helpful to review the known differences in material included in these two publications. Our examination of the Index-Catalogue and Index Medicus indicate that certain classes of materials were always included in Index-Catalogue, and largely excluded from Index Medicus. Monographs accounted for the chief difference: they showed up only rarely in Index Medicus and then only when they seem to have been of particular importance, while they were included as a matter of course in the Index-Catalogue. Journal titles were not indexed in Index Medicus, although their contents were. In other words, articles from The Lancet were indexed individually, but The Lancet itself had no entry. The Index-Catalogue included entries for both articles from a given journal and the journal itself.
The most vexing question seems to be how books, journals, or articles were selected for inclusion in either publication. On the surface, it would seem that for the Index-Catalogue, the level of selection was the availability of the periodical itself. If a publisher sent a copy of a periodical to Billings, the expectation was that all of its significant contents would be indexed. We have found no evidence indicating that Billings or his successors rejected any submission to the Index-Catalogue during the period 1879-1926. (Folklore has it that Billings, and possibly Fletcher, would go home in the evenings with baskets of periodicals under their arms and, after dinner, fortified with port and cigars, make selections and assign subject headings. Sadly, such tales cannot be verified.) All of this is a far cry from NLM's Literature Selection and Technical Review Committee (LSTRC), which since the 1960s has carefully examined and recommended journals to be indexed by NLM. However, our examination has made it clear that articles included in Index Medicus were sometimes excluded from the Index-Catalogue. It does not seem possible to determine why, but one thing can be said with some certainty: Index-Catalogue primarily indexed journal articles selected on an individual basis while Index Medicus eventually selected and indexed the entire contents of particular journal titles. Unfortunately, it seems impossible to track the exact process and inconsistencies abound.
Moreover, this pattern of selection would seem to contradict the traditional accounts by Miles and others, which state that indexing for the Index-Catalogue took priority. This certainly seems to have been the original intention, but one wonders if it remained so in practice, particularly as work on the Index-Catalogue fell farther and farther behind that for Index Medicus. Our examination strongly suggested that citation cards were produced by the Index-

Greenberg and Gallagher
Catalogue staff, never to be used in that publication, but solely for inclusion in Index Medicus. Journals lists were included in the first volumes of both publications (in later years, a list would be printed for each Index-Catalogue series, but only rarely for Index Medicus), but the scope was very different. The Index Medicus list was 25 pages long, while the Index-Catalogue list, printed in similar format, was 5 times longer at 125 pages. Coupled with the fact that the 2 indexes used different subject headings (a point addressed below), it seems likely that, over time, the workflow diverged considerably in method, though not in scope.

METHODS
The current study used a sample derived by selecting Index Medicus volumes at 5-year intervals between 1880 and 1925. The contents of Index Medicus expanded over time (the 1880 volume contains an estimated 22,248 citations; the 1925 volume over 55,000). So, to avoid biasing the sample in favor of later years, a fixed percentage of entries was selected from each volume. This resulted in 1,042 citations randomly selected from over 393,000 entries in the selected Index Medicus volumes that were then searched in IndexCat. It should be noted that, because the sample was derived from Index Medicus and then compared to IndexCat, rather than the other way around, no correction needed to be made for classes of material included in the Index-Catalogue but generally omitted from Index Medicus. Figure 1 shows the results. Overall, 77.2% of the citations selected from Index Medicus could be found in IndexCat, but there was considerable variation over time. The ''best'' years, the ones with the most overlap, were the early years, when the library presumably exercised the most control over the process. The ''worst'' were the years 1899-1902 when the process moved to Paris (the authors have been unable to find much about how indexing decisions were made in those years). Moreover, if we took those ''French years'' out of the equation, the overlap rose to 80%. Was Robert Fletcher, who stayed on the job throughout those troubled times when no US publisher could be found or his new, young assistant, Fielding Garrison, actually sending cards to France? Garrison, incidentally, was ''given the opportunity to revive the Index Medicus'' in 1902 [7]. The Carnegie years  are better, but they never achieve the results of the first few years. Perhaps they should not. After all, Index Medicus was not designed to be a temporary Index-Catalogue; this was not a Pre-MED-LINE-MEDLINE relationship. Index Medicus was the newsletter of the medical publishing world, while the Index-Catalogue was the research guide to a particular medical library that would grow to become the world's largest.

RESULTS
Evidence to this effect can be found in how the materials in these two publications were indexed, by which we mean the vocabulary of the subject headings. In the ''Prefatory Remarks'' to the first issue of Index Medicus, the editors (that is, Billings and Fletcher) wrote: ''The nomenclature and classification are essentially those adopted by the Royal College of Physicians, based on Dr. Farre's well-known system'' [8]. Frederic John Farre (1804-1886) was a distinguished British botanist, physician, editor, and historian. It would make sense to adopt an already accepted system for Index Medicus. But Index-Catalogue was another story. Because Index-Catalogue was a government document, and a military one at that, it included a letter of transmission from Brevet Lieutenant Colonel Billings, US Army, to his superior, General Joseph K. Barnes, surgeon-general, US Army. Nearly half of the letter was concerned with the new system of headings, of which the most important points were: I. Those [headings] have been selected for subjects for which it is presumed that the majority of educated Englishspeaking physicians would look in an alphabetical arrangement. II. Where there is a doubt as between two or more subjectheadings, cross references are given. III. Where both an English and a Latin or Greek word are in common use to designate the same subject, the English word is preferred, and references are given to the other. [9] Billings was thus consciously creating a new, nononsense, English-language guide to the international literature of medicine, as it existed in the library he ran: In conclusion, permit me to call attention to the fact that this is not a complete medical bibliography, and that any one who relies upon it as such will commit a serious error. It is the Catalogue of what is to be found in a single collection-a collection so large and of such a character, that there are few subjects in medicine with regard to which something may not be found in it, but which is by no means complete. [10] Of course, things did not quite remain that way. As Billings's library grew from a military medical library to a national medical library (in truth, he seemed to have always had that in mind: On the title page of the Specimen Fasciculus, Billings calls his institution the National Medical Library) to the National Library of Medicine, its controlled vocabularies and classification system would grow in complexity, depth, and consistency, far outstripping any other [11,12]. However, it is worth taking a moment to note the results of having two parallel subject vocabularies and two very divergent printing and production schedules, during the period under discussion. Table 1 presents subject headings for single articles from each of the Index Medicus volumes examined for this paper. Included are the year the article was originally published, the year it appeared in the Index-Catalogue, its Index-Catalogue subject heading, the year it appeared in Index Medicus, and its Index Medicus subject heading. It becomes immediately apparent that the differences were so great that no more systematic sample was called for. The headings are entirely different in virtually every case. As for the year that indexing was available, the differences are equally dramatic.
As we compared entries in Index Medicus and Index-Catalogue, some other peculiarities in format became evident. One particular difference was the indexing of the authors' names. In the Index Medicus, while most of the entries were the traditional last name, first and middle initial, some entries included the author's first name. Those names were usually female. For the current paper, a random sample was not undertaken (the question of teasing out all of the female authors hiding behind initials was too daunting), but the Index Medicus volumes used for the project were examined to determine if this was just an oddity in a few years of the index or if it reflected the all of the indexing from 1880 through 1925. The latter was, in fact, the case.