posted 1 Jun 1998 in Volume 1 Issue 6
Evolving Role of Knowledge Management in Medicine
Bryan Bergeron, MD at MGH Critical Care examines how knowledge management has a key role to play in the practice and management of modern medicine.
In medicine, knowledge management issues have long been relegated to the academic community, which was primarily focused on supporting the more theoretical, clinical aspects of medical care. Because of social, political and technical barriers to acceptance by practising clinicians, there have been few obvious commercial successes in the application of knowledge management in clinical practice. The majority of the clinical knowledge management systems that have been developed remain as mere academic curiosities, and do not influence the daily practice of most clinicians. However, contemporary pressures of economics, legal and expense accountability, as well as recent advances in information technology, have brought knowledge management techniques to centre stage for the vast majority of clinicians. Knowledge management is coming to be viewed by the entire medical community as a necessary and crucial component of any modern technology-enabled healthcare initiative.
The practice of medicine is at a crossroads. Clinicians, like physician assistants, nurses and other medical knowledge workers, are finding their time compressed, their responsibilities expanded, and their control dissipated by regulatory agencies, insurance companies, and other third party payers. In this time of turmoil, technology is often viewed as a lifeboat in a sea of uncertainty.1 Paradoxically, the same technologies that many regulatory agencies use to critique and control medical practices on a population and healthcare enterprise level make it possible for clinicians to control their world on a patient level.2 For example, clinicians are today saddled with the responsibility of exactly documenting the procedures and diagnostic studies that are used to diagnose their patients. In addition, they are expected to follow best practices - typically expressed as clinical guidelines that have been developed by any number of agencies, from local hospitals to national groups - in diagnosing and treating their patients. In private practice, clinicians are additionally faced with competition from the larger health maintenance organizations. Many of these smaller practice clinicians view applied knowledge management technologies as one means of competing effectively with these larger entities.
Clinicians are knowledge workers. Traditionally, their tasks include gathering patient data, analysing the data for patterns that they recognise, and then formulating a differential diagnosis, which in turn leads to a treatment plan. The differential diagnosis, a list of potential maladies that are associated with the patient presentation, represents the potential problem landscape that clinicians must traverse. Much of their work is involved with either refuting or affirming each of the potential maladies listed in the differential diagnosis. One of the traits of a skilled clinician is the ability to quickly create a short list of differential diagnoses that presumably contains the true diagnosis, while not having so many potential diagnoses as to create an economic and time requirements disaster for the patient. Unfortunately, much of the traditional clinical work - work that clinicians have trained for - is being quickly superannuated by the need to index and code every activity, procedure, and diagnosis for billing and administrative purposes. It is not uncommon for some clinicians to devote 40 percent of their time in these non-clinical activities.3
Technologic innovations aimed at curbing cost and time expenditures have fuelled considerable interest in knowledge management tools. For example, the electronic medical record (EMR), the repository for all patient information at a given hospital or health care enterprise, represents the new core knowledge management vehicle in clinical practice.4 The EMR, sometimes referred to as the longitudinal medical record, contains the complete medical history for a given patient, including not only information on the patients current complaints, but on the past medical history and treatment as well. Once a medical enterprise makes an EMR system available to clinicians, it becomes relatively straightforward to develop ancillary programs that feed off of the patient data.5 For example, decision support tools, including expert systems, can use the EMR as the basis for assisting in clinical decision support, including formulating a differential diagnosis. However, the attraction for most clinicians is the ability to identify potential avenues for increased reimbursement, and to quantify clinical outcomes for administration and planning purposes. In contrast, the records of paper-base practices, which include the majority of US medical practices, must be analysed by hand. This slow, tedious process is usually based on samples of patient records because of the sheer number of records involved. As such, the data are not as timely, are more open to variation in interpretation, and are more error-prone due to errors in transcription.
A Historical Perspective
As noted earlier, knowledge management techniques have been used in medicine, with varying degrees of success, for decades. The more notable projects have been primarily associated with medical expert system development, critiquing systems, in the creation of decision support tools, as well as visualisation techniques used to make information more understandable.6,7 In these and other areas, knowledge acquisition and management have been the weak links in development. Also, because clinicians have traditionally placed little emphasis on clinical decision support assistance, there has been relatively little commercial activity in this area. Most of the research and development has been limited to academic research centres.
One of the areas that knowledge management techniques have been used with some degree of success is medical education.8 For example, keeping track of multimedia teaching files for a radiology course and for a resident teaching file is no mean task. There are the rather easily addressed issues of identifying the most appropriate storage media and database structure, but these issues aside, there is an even more critical issue of indexing images so that they can be quickly be identified and, most importantly, repurposed in different projects. For example, consider the difficulty of indexing several thousand histology slides. The images could be indexed by cell type, gross location, anatomic origin, stain type, pathogen or disease state. Because creating a multimedia library is so time intensive, failure to be able to quickly locate any media element in a knowledge base translates to lost productivity, added expense, and, at best, faculty frustration.
The same knowledge management techniques that have been applied to medical education, especially in multimedia indexing, retrieval, and repurposing, have been applied to clinical information management as well. For example, the issue of how to best archive and later retrieved radiographic images has received considerable attention. There are the technical issues of care and marking of the films, together with indexing, transport from one laceration in the enterprise to another, and tracking. With the move to digital imaging, the relatively large (10MB or more per image) image size is a technical challenge, especially when it’s desirable to have tens of thousands of digital radiographic images on hand within a few minutes notice. Hardware issues aside, there are knowledge management issues of how to best store images for rapid access.
Contemporary knowledge management issues
Knowledge management has a prominent place in the modern medical practice. While research continues in how to best apply knowledge management tools and techniques to clinical decision support and medical education, most of the effort in the field is focused on data gathering, in defining best practices for maximum quality of care, in maximising reimbursement, in managing electronic medical records, in communicating with colleagues, and in predicting individual and a population outcomes. Today, the pressures of outcomes measures, determining best practices, maximising reimbursement, and quality controls, all require timely, accurate information. Clinician interest is firmly affixed on the economic survival of their practice. For hospitals, reimbursement from the government and other third party payers is keyed to diagnostic impressions and physician work. As such, there is significant motivation to understand, catalogue, and have readily available timely local patient demographics, disease prevalent figures, and other populations specific problems.
A prominent example of how knowledge management is influencing medicine is in the development and use of a standard medical vocabulary.9 A standard vocabulary for describing patient findings, clinician orders, and patient progress notes not only allows disparate electronic record systems in the same hospital or clinic to share information, but it also allows regional and national comparison of patient data. For example, clinicians in one department might use the term “high blood pressure”, whereas another group of clinicians might use the term “hypertension” to describe the same concept. Since this preference varies from physician to physician and from hospital to hospital, a virtual tower of Babel often results.
There have been numerous projects that have attempted to create either a universal medical language, or to provide a translation medium between systems. For example, there is the Unified Medical Language System (UMLS) project, and the older Medical Subject Heading (MESH) vocabulary, both funded by the National Library of Medicine.10,11 Both attempt to address the tower of Babel issue by using concepts that map onto a variety of naming schemes. For example, hypertension and high blood pressure map to the same meaning. The importance of this mapping is that it allows free, unencumbered access to thousands of medical record documents that would have otherwise been unexaminable. For example, when searching a knowledge base for “high blood pressure”, the user generally wants references to equivalent terms as well. The alternative, of course, would be to have the researcher perform the translation, supplying all possible related terms. However, such an approach would be time and resource intensive, and prone to errors of omission.
A related knowledge management issue that has recently become significant in the medical community is how to handle the myriad of medical resources now available through the World Wide Web.12 While some well-known, high-quality resources, such as MEDLINE, from the National Library of Medicine13, are relatively easy to locate, other resources are not.14 Even so, many clinicians and their patients are making use of medical resources on the Web at a surprising rate. There are resources for continuing medical education, for decision support, and for creating electronic medical records. Although there are issues of security and confidentiality, Web-based electronic medical records is certainly the wave of the future. Given the richness and variable quality of the medical resources available through the Web, there is considerable pressure to develop a workable knowledge management scheme that provides easy-to-use, efficient access to a variety of medical sites.
Another example of how knowledge management technologies are being applied to modern medicine is in the areas of medical voice recognition.15 With the increasing cost and delays associated with traditional transcription services, where clinicians and hospitals pay typists to transcribe their recorded transcripts of patient findings, there is renewed interest in alternative recording technologies. One of the most promising, voice recognition, requires that context-specific language models be created for each area of practice. For example, there are typically different language models for radiology, oncology, pathology, and primary care. Creating a language model, which describes both the vocabulary and probability of word phrases, requires processing thousands of typical medical records for a given speciality. These prototypical records must be selected and analysed carefully for the best results.
A third example of how knowledge management techniques are being applied in modern medicine is in maximising the reimbursement to clinicians from third-party payers. Because the rules of reimbursement are so convoluted and difficult to understand, many clinicians simply accept lower pay for more work. Others are looking actively to utilities that can help them decide how to best manage their time and their patient care practices to receive appropriate compensation. For example, there are systems that interactively process the EMR and inform the clinician, during the patient encounter, that how to increase reimbursement without compromising the quality of patient care or subjecting the patient to any unnecessary procedures. These knowledge-based systems suggest, for example, that the clinician might want to examine the ears in order to move from one payment level to another (which can amount to approximately $45 for an additional 30 seconds of work). The danger, of course, is that clinicians can use these programs to “game” the healthcare reimbursement system, increasing the cost of healthcare in the long term.
Knowledge management has a key role in the practice and management of modern medicine. Due to economic pressures, the clinical role is greatly overshadowed by systems and techniques that have a direct, immediate impact on the bottom line. Medicine has become big business, and knowledge management has become another tool in the healthcare administrator’s arsenal against the bottom line.
Bryan Bergeron is MD at MGH Critical Care. He can be contacted at: