Clinical Reference Guide

Two medical students hated throwing their notes away at the end of each year. Why not put them on computer, and use them as the nucleus of a constantly updated textbook? After graduation, three other doctors joined the cause, forming the Clinical Reference Guide, an electronic textbook aimed at students for its digestibility, and GPs with its convenience. One of them, Dr Damian Crowther, writes about the company past, present and future.

The Beginnings

The Clinical Reference Guide (CRG) is a concise database of medical knowledge which has been developed over ten years by a team of doctors. The CRG began back in 1990 when a couple of outstanding medical students entered the Oxford and Cambridge University medical schools. Brothers, Stewart and Jim McMorran were fresh from Ph.D. research projects in computer science, and so started their medical studies with a unique perspective on the process of learning medicine. Their aim was to change the way they collected and archived the information provided in their lectures.

Perhaps you recognise my experience of undergraduate lectures as a process of transferring knowledge from the lecturer's head onto a piece of A4 paper and seamlessly into the bin at the end of the year. The McMorran brothers wanted to really harness this valuable information and, instead of taking pen and paper to the lecture theatre, they took laptops. This behaviour was considered eccentric, but probably harmless, by many of their fellow medical students.

Jim McMorran says "We didn't want all our lecture notes to get binned at the end of the year, instead we wanted them archived and cross-linked in a way which permits knowledge to be accessed at various levels of complexity and detail".

Jim remembers writing his medical notes in a shorthand which he subsequently adapted to provide the logical structure to the CRG: "When I started the clinical course I would write notes in my own medical shorthand". For example:

jaundice MBCB cirrhosis	meant	jaundice may be caused by cirrhosis
hepatitis TYPE hepatitis A	meant	one type of hepatitis is hepatitis A

During his computer science degree, Stewart McMorran had written programs that created networks of medical information, he used this experience to transform Jim's notation into a new database. The database contained a set of nodes each of which described single concepts such as "liver cirrhosis", "jaundice", "hepatitis" and so on. The nodes were joined by semantic links with the values "MBCB", "TYPE" and so on; there are now over thirty different semantic link types now describing in detail the strength and the type of link between two nodes.

With a blank canvas, the McMorran brothers initially spent most of their time mapping out the terrain of medical knowledge. With time the details began to be filled in. From a beginning based on the undergraduate clinical course the system has maintained its "generalist" approach - seeking to be sensible and comprehensive without hoping to provide the depth of more specialist texts. We now maintain the data using the major medical journals paying particular attention to review articles but also reporting summaries of key primary research material. The natural market for such a product has continued to be the medical student, but of major interest to us at present is the general practitioner. Two of the current team are general practitioners: Jim McMorran is now a GP principal, and Dr Clive Prince is a principal in the award winning Bewdley Medical Centre.

The project begins to grow

Clearly a work of such breadth requires a team of authors to maintain the data. Stewart McMorran is the co-creator of the system and now a specialist registrar in accident and emergency medicine. He is involved in developing the CRG as a resource for the NHS National electronic Library of Health. Steven Young-Min is a research registrar in the University of Newcastle department of Rheumatology. Ian Wacogne is a specialist registrar in paediatrics, currently working in Australia. Jon Pleat is a surgical trainee. Finally, there's me, Damian Crowther. I am an Honorary Registrar in Neurology and a Wellcome Trust Training Fellow in the Wellcome Trust Centre for Molecular Mechanisms of Disease on the Cambridge Addenbrooke's Hospital Site. I've been involved in writing the system and also in the internet implementation on the www.epulse.co.uk site. We have developed a wide range of clinical and research interests but are committed to maintaining a level of general medical knowledge which enables us to write for the CRG. We have a pragmatic approach to authorship - we write as the educated, and hopefully intelligent doctor, who although not a specialist has the latest literature to hand. The result is practical and useful coverage of specialist areas.

Keeping a team of doctors working on a single database has required some thought and organisation. Our approach has been to provide each author with the current version of the database and ask them to go away and work for a month or two. At the end of this period everyone submits their work and the various contributions are integrated into a single database. This works well because it allows the authors to work on their desktop machines without the need for internet connection, which makes everything faster and more reliable. We have the problem of possible duplication of effort - occasionally if there is a hot topic in the journals we find that two authors will make contributions or alterations to the same subject. Happily this is not a common problem because of the regular integrations and the breadth of the database. Another potential drawback is the effort required to integrate the various contributions, however the integration process does allow Stewart McMorran to exert a final editorial influence on what is incorporated into the database.

The impact of the Internet

The CRG provides a concise and comprehensive medical knowledge base. The challenge posed by the internet is that as more specialist databases become available we will be competing for an audience who could easily go to more "authoritative" sources. However what an internet-full of specialist databases does not provide is the consistency, integration and easy searchability which we offer with the CRG. We are confident that the concise comprehensively of the CRG will provide enduring value.

The CRG is well-suited to implementation on the internet; it has been designed from the start as a web of medical knowledge. This contrasts with some other similar works, which are conversions from printed books to CD-ROM or internet. The standard texts such as Harrison's online are clearly excellent information resources; however the only advantage of putting them into electronic format is that they are quicker to search. In contrast the database is written as a web of concise nodes with links of explicit meaning. This provides information which is consistent, and leaves the user with a clearer impression of the structure of the knowledge.

Design issues

The idea that knowledge is an interlinked web of discrete concepts has fostered a concise writing style - each page is usually 40-100 words and usually covers a single main topic. Broader fields of knowledge are constructed by linking various pages together. The result is a concise editorial style with a structure, which facilitates easy maintenance of the information. Maintenance is of utmost importance when the text now stretches to over 27,000 pages. We have a policy of dealing with a concept only once and referencing this page whenever the concept comes up. This approach gets round the problem faced by the multi-author monoliths on our bookshelves which often deal with a concept several times, in several places in several slightly different ways.

An example of the concise style of the CRG:

PANCREATIC CARCINOMA

A disease of later life, pancreatic carcinoma is the third most common cancer of the gastrointestinal tract and the sixth most common cause of cancer deaths.

70% of tumours occur in the head of the pancreas in proportion to the amount of tissue in different parts of the pancreas.

Progress is insidious. There is early regional lymph node involvement and spread to the liver. The tumour may metastasize to bone and lung.

Less common endocrine tumours of the pancreas such as glucagonomas and insulinomas may occur.

The search strategy is also important in a database of this complexity. The task faced by the internet search engines is unenviable and unnecessary when the database is designed, like the CRG, to be searchable from day one. When creating a new page the author must consider carefully how the user might access the information. Titles for pages should contain the keywords, which might appear in a search query. If a single title cannot cover the range of possible access routes then an unlimited number of "synonymous" titles may be provided which incorporate important alternative keywords, for example the page on "breast cancer" will have a synonymous title of "breast carcinoma". Our experience is that searching titles and their synonyms provides the best approach to generating a short, but focused, list of "hits". When a page is created the author must also decide whether the page will be visible to the search engine. Making a page invisible permits the author to cut down the clutter in search results pages. For example the page on breast cancer will be visible to the search engine but the linked page on "aetiology" will not be visible - this later page is accessed by first jumping to the indexed page on "breast cancer" and clicking on the "aetiology" hyper link.

We are also considering a more exhaustive search that would look for search terms in the whole 2.5 million words in the database. This will be reserved as an "advanced option" because it has the potential to yield a lot of unfocussed hits. We are not keen on an exhaustive searching approach because we believe that the logical structure of the CRG allows rapid navigation from the search "hits" to the specific page of information required.

How things have changed

Initial implementation of the database was in DOS with the 4 megabytes of text supplied on multiple floppy disks. We subsequently moved into the CD-ROM era, and now pride ourselves in being one of the leading providers of online medical information to UK general practice. This is via the implementation of the database as epulse Clinical Reference Guide on the ePulse site. The site is available using a password to all UK registered doctors. We are keen to continue this relationship with Miller Freeman Ltd, and provide a more up-to-date, comprehensive and graphically rich database.

So what next? Desktop and laptop computers will continue to be the best platform for the full version of the database, however we are exploring an implementation of the text-only database on a personal digital assistant (PDA) [such as Palm Pilot, PocketPC or Psion]. The text requires only 25 Mbytes storage for the full version. With PDAs' RAM approaching this size, we want to provide the ability to download subsections of the CRG onto a PDA. Having a database covering the whole range of clinical medical knowledge is only a benefit to doctors if they can get their hands on that information when they really need it. We see the PDA as an important platform, particularly for community physicians, who are often away from the office or hospital ward.

Whatever the next incarnation of CRG turns out to be, we're looking forward to it.

Dr Damian Crowther