by Nicole Contaxis

In season three of Doogie Howser, MD, Cecilia, a medical librarian, instructs Doogie on using computers for clinical research. With his newfound computer literacy, Doogie is able to solve a medical mystery.

Just out of sight in this scene is the National Library of Medicine (NLM) and the excited medical librarians that consulted on the episode. NLM began as the Library of the Surgeon General’s Office in 1838 and has become an international resource with materials catering to doctors, nurses, biomedical researchers, and the public. The process of transforming a military hospital library into what NLM is today was equally informed by technological advancement and changes in institutional goals and ethics.


Image by Flickr user NIH Image Gallery

At the time of the Doogie Howser episode, NLM had developed and implemented Grateful Med, an oddly named but important software program that provided a user-friendly interface to NLM’s bibliographic data, including medical journal citations. It allowed for near instantaneous search of biomedical research all over the world. The creation of Grateful Med, as with all technology at NLM, depended on the skills of staff members, the current state of computer technology, and perhaps most importantly, a morally informed conception of who their users where and what they needed. While many people have capitalized on the resources at NLM, the type and number of people who used the library’s technology over the years has changed.


Image by Flickr user Pacific NW Region National Network of Libraries of Medicine

As a library that develops its own software, NLM acts as an interesting case study of how a public resource defines itself, understands it users, and addresses those users through technology over time. Technology may be deeply woven into this story, but it is not the story. Instead, what NLM illustrates is the way that institutional goals and ethics can shape how, why, and when technology is developed. Particularly because NLM is a library and a public resource, their goals are deeply informed by their relationship with and understanding of their users.

Between 1960 and 2001, there are three dominant computer systems that aided in bibliographic search in the history of NLM: MEDLARS I, MEDLARS II, and Grateful Med. Each of these systems was comprised of many software programs and administrative activities, and other search systems ran concurrently. However, MEDLARS I, MEDLARS II, and Grateful Med demonstrate three separate advancements in computer technology and in understandings of library and computer users.

Each of these systems acted as technological embodiment of user-based institutional goals, but the relationship between technology and these goals is far from linear. The possibilities provided by technological advancements affected how broadly the institution could consider their users. One cannot aim for an international audience if users can only access materials while in the library’s reading room. Technology assisted in a creative approach to the user, but the way the institution understood the user informed the way that technology was implemented.

Throughout NLM’s history, pilot projects and experimentation informed these larger systems. These experiments, some of which were eventually included into MEDLARS II and Grateful Med, demonstrate how early adoption of technology can inform the way that upper-level management approaches new ideas and new directions for an institution.

Before delving into the history and structure of each of these systems, it is helpful to talk about the data that each system provided access to. The data collected throughout NLM’s history grows in conjunction with the growth of medical literature. Essentially, each new system constitutes a migration of the same bibliographic data that NLM began compiling in 1879 with the publication of Index Medicus. Index Medicus was a list of current medical literature, published regularly, that was intended to aid medical research in the late 19th and early 20th centuries. While the data first looked like a book, then like punched paper tape, and eventually magnetic tape, it is essentially the same data through time—medical literature citations. Medical literature citations assist in medical research and patient care by helping doctors learn about current research. These citations allow doctors to provide patients with the most up-to-date care. The beginning of NLM’s legacy computing systems does not start with a computer; instead it starts with the information held inside Index Medicus.


Image from NLM Digital Collections, NLMUID: 101648151


MEDLARS, an abbreviation for the Medical Literature Analysis and Retrieval System, was the first attempt to build a computerized system for bibliographic search at NLM. NLM had begun trying to mechanize the process of producing Index Medicus in the 1940s. By the time the 1960s rolled around, NLM was beginning to experiment with ways to computerize the process of creating and using Index Medicus.

The creation of MEDLARS was informed by two separate but linked goals. The institution wanted to improve worker efficacy while producing Index Medicus and also allow for computer-driven search of their bibliographic data. This demonstrates two things. First, the library still considered the existing published bibliography an important tool for medical research. Second, the creation of MEDLARS I was about diminishing labor costs, both for the institution and for its users. By decreasing the amount of time spent creating Index Medicus, the library saved on labor, and by providing for computerized search, the user could, hypothetically, save on time. The development of technology and concerns about labor were connected from the start.

Although the system was designed to diminish labor, creating and maintaining MEDLARS I required considerable effort. Bibliographic data needed to be inputted using punched paper tape, and then the data needed to be added to the magnetic tapes that stored the majority of the records. This was additional work and a new type of work, requiring new expertise at the library.

The system itself ran on a batch-processing computer, a Honeywell-800. This batch-processing computer functioned by performing several jobs at the same time. Instead of seeing errors in computing as they happened, errors would only become apparent at the end of the process. This means that users would submit a program, in this case a search query, the computer would run those programs, and if there was an error the users would have to start from the beginning and locate the error on their own. This was a time-consuming process, but for a medical researcher, it could significantly reduce time spent searching for new citations and could locate citations that the researcher would have never found in the first place.


Image from NLM Digital Collections, NLMUID: 101446893

At NLM, the researcher would not interact directly with the computer. Instead, they would need to submit researcher requests through a medical librarian, the technical use of this NLM system. The librarian, specifically trained to interact with MEDLARS I, would properly format the query using the controlled vocabularies that had been programmed into the computer.

A controlled vocabulary is basically a set of words that standardizes language for a computer. A ‘heart attack’, in this case, needs to always be referred to as a ‘myocardial infarction’. The search query would then be run through one of the MEDLARS Centers. This was a library that held copies of the NLM bibliographic data. NLM, to this day, licenses its data for free to help facilitate access. The computer at the MEDLARS Center would then search the data for the particular subject of interest to the researcher. When running at optimal speeds, this process, equivalent in some ways to today’s internet search, would take two weeks.


Example form used to perform MEDLARS search, taken from MEDLARS 1963-1967

It is interesting to note that this structure, even if not technologically advanced, is a decentralized system. As opposed to having all search requests go through NLM, searches were performed at a MEDLARS Center, located internationally, closest to the research. This cut costs on postage and facilitated quicker responses, particularly because NLM would not have been able to handle all search requests with its own Honeywell-800. This decentralized structure is mirrored in the structure of the later search systems at NLM, even as they grew more complex and technologically advanced.


As turnover time for a search query expanded to far over two weeks for MEDLARS I, it became apparent that a new system was necessary to meet user needs. The design history of MEDLARS II is not straightforward, and the system went through the design stage twice. The first time MEDLARS II was designed, it was designed as an improved version of MEDLARS I. NLM planned on getting a faster batch-processing computer to improve search times and thus improve user experience.

While a system was being designed for these faster batch-processing computers, a smaller project blossomed at NLM: *Abridged Index Medicus *– Teletypewriter Exchange System (AIM-TWX). AIM-TWX built on an earlier project, the State University of New York (SUNY) Biomedical Communications Network, built in 1968. Both of these projects provided interactive search, a step past the batch-processing model of MEDLARS I and the first design of MEDLARS II. Interactive search provided instantaneous results, and didn’t require the extensive wait time of batch-processing computers. AIM-TWX as so successful that MEDLARS II was re-designed to match its model.


Image from NLM Digital Collections, NLMUID: 101446013

Upon the re-design, MEDLARS II became MEDLINE, an abbreviation for MEDLARS Online, which began operating in 1971. The system functioned through teletype machines placed at libraries internationally. Teletype machines allowed users to dial into a computer from a distance, and a teletype printer could print out results nearly instantly. Teletypes have a long history, and were used as early as 1933 with the Telex network, a network meant for business communication that was maintained through a variety of companies internationally.

Thirty years after its origin, this networked style of communication was combined with time-sharing computers. A time-sharing computer allowed for many users to interact simultaneously with a computer over individual terminals, like teletype machines. As opposed to batch-processing computers, time-sharing computers allow users to adapt and change queries more quickly and do not require users to wait for all previous programs to be run before they change theirs.

As with MEDLARS I, there were international centers that hosted MEDLINE data, thus distributing the computational and human labor associated with performing these searchers.

Like many systems, naming and understanding the various components used here can be difficult. The name “MEDLINE” was used to refer to the entire system, but also to the data that the system searched. The software that actually performed the searches of the data was called ELHILL. ELHILL went through many iterations and continued to function in some form or another until 2001. While the interface for access to NLM data changed over time, ELHILL powered searches in the background for many years.

Allowing for instantaneous search expanded the number of people who could use these NLM technologies. With MEDLINE, more people than ever were performing searches on NLM data and interacting with the terminals, computers, and software programs. As with MEDLARS I, MEDLINE still required that researchers interact with specially trained medical librarians to search NLM data. These librarians were trained in Boolean logic and MeSH, the Medical Subject Headings produced at NLM, to perform as wide and as accurate of a search as possible. Training sessions could be two weeks long, and librarians would fly to NLM in order to complete this training. These librarians were expert searchers, and trained specifically to interact with NLM technologies.

What this means is that these librarians were the user community for NLM technology. Although NLM was serving biomedical researchers around the globe, it was the librarian who interacted with the computer. Thus, the programs were designed to be understandable and accessible to those librarians, not the researcher and not to the general public. This occurred in part because the technology was not yet advanced enough to allow for non-trained professionals to use it. Rather than the technology accommodating the user’s needs, the users accommodated the technology’s needs. This relationship was about to change.

Grateful Med

Before Grateful Med, NLM understood that their technology needed to be accessible to librarians trained specifically on their systems. But in 1986 with the release of Grateful Med, the relationship between NLM and the people who used its technology changed. This new software allowed NLM to talk directly to doctors, nurses, and biomedical researchers. Librarians, while still deeply important to the process of searching, were no longer needed to interact directly with the data and NLM technology.

Grateful Med itself was a user-friendly interface that relied on ELHILL to power searches on NLM data. As a system, it interacted with and relied on many software programs, pieces of hardware, and administrative staff in order to function as it was intended. Grateful Med was the most visible aspect of this system because it was an end-user interface, but that does not make it intrinsically more important than the other parts of the system. However, designing technology that was meant to be easy to use was an important advancement in technological design and in the institutional goals of NLM. By designing technology intended for non-expert users, NLM was broadening its mission and undertaking more responsibility to the biomedical community.


Image from NLM Digital Collections, NLMUID: 101648264

This system was a success and non-expert use of NLM data increased quickly across the globe. In 1987 alone, the number of authorized users for Grateful Med increased by over 50%. The rate of increase grew as the project gained momentum and more researchers gained access to personal computers. By 1994, Grateful Med searches account for 80% of the four million ELHILL searches of NLM databases. Because NLM chose to design technology that spoke directly with researchers, they were able to expand access to their resources and assist with biomedical research globally.

It would be foolhardy, however, to not recognize that Grateful Med also relied on technological and market forces outside of the control of NLM management. In order to run Grateful Med, a researcher needed a personal computer, at the time referred to as a “micro-computer.” Without personal computers becoming more accessible and more affordable, Grateful Med could not have been successful and NLM would not have been able to reach so wide of an audience.

The most important aspect of Grateful Med is that it helped NLM expand the number and types of people who were interacting directly with its data. While the institution made the initial decision to begin the development of Grateful Med, its continued success helped determine the future of the institution, its goals, and who it considered its users. This success was dependent on a variety of other things, including the affordability of personal computers and the administrative systems put in place with MEDLARS I and the history of library services at NLM.

The creation and popularization of the internet further changes the shape of services of NLM. The library offered Internet Grateful Med in 1996 to provide an interface to the services of Grateful Med through the internet, but the institution officially retired Grateful Med in 2001 in favor of internet-only services, like PubMed. Interestingly, the adoption of PubMed led to a further expansion of the NLM user base. Services for the general public, not trained as doctors, nurses or researchers, began to expand and NLM started to concern itself with helping to provide information to the lay public. Understanding the foundation of networked search is helpful in making sense of these recent expansions.

Over this forty year period, NLM expanded its technology and re-defined its user base according to its own institutional goals and technological capabilities. While one of these influences cannot be said to directly cause the other, the history of computerized search at NLM helps demonstrate how and why new programs and new systems are implemented at an institution.

More recently, as genomic research exploded throughout the 1990s, NLM worked to provide search systems to help researchers match protein shapes and speed up the matching of the human genome. Technology cannot exist separately from the needs and requirements of a user base that is always trying to address the newest aspects of biomedical research. Aiding in the research process means helping doctors communicate both nationally and internationally and adjusting to their needs as medicine and technology changes and adapts.

Nicole Contaxis is an archivist and technology historian living and working in New York City.

We have a print edition too! Find this issue in the shop.