Healthcare Computer Applications and  The problem of language: A Brief Review*

Jeffrey S. Rose, MD and Debra Kirkley, PhD, RN

"The significant problems we face cannot be solved at the same level of thinking we were at when we created them."

Albert Einstein

Dr. Rose is the former Chief Medical Officer at Health Language Inc. where he was responsible for strategic leadership in the company and supervision of the Clinical Team. Prior to joining HLI, Dr. Rose was Director of Clinical Information Systems for the Rocky Mountain Division of Kaiser Permanente and co-manager of their computerized health record endeavor (winner CPRI 5th Annual Nicholas E. Davies Recognition Award). He served on the Kaiser Permanente Clinical Information System Advisory and the Convergent Medical Terminology Group, and authored Medicine and the Information Age, (ACPE). He is a Clinical Assistant Professor at the University of Colorado Health Sciences Center where he completed medical training and fellowships in Medical Imaging. He is presently the Informatics Editorial Group Chair of The Doody Review and  an instructor of informatics for the American College of Physician Executives.

Introduction

In this article we discuss  a core problem in healthcare.  We indicate the technology necessary to make the transformation from partially informed ‘financial’ decision-making (managed costs) to better informed  ‘clinical’ decision making (true managed care). We believe that using tools that address this core problem enable the delivery of the highest quality, lowest cost and most evidence based healthcare possible.

‘Language’ as a Core Problem

To obtain information linking clinical actions to the costs of health care, clinicians and health enterprise managers require very specific tools: the computerized patient record (CPR) chief among them.  But the CPR alone is insufficient, because standard methods of representing and communicating health data are missing.

The complexity of the English language, which is formidable indeed, pales in comparison to the eclectic, ambiguous, redundant idiom of clinicians.  The down side of the ‘uncontrolled’ parlance of medicine has been sharply accentuated by the computer age, because without standard vocabulary the ability to acquire knowledge about healing professions through information technology is painfully limited. The clinical applications available today have restricted utility because they cannot ‘understand’ each other. 

Today’s information technology tools fail without ‘controlled medical terminology’ (CMT). Christopher Chute of the Mayo clinic pointed out with elegant simplicity that for “clinical information systems to be broadly useful…comparable information…is needed.’[i]  Central to obtaining ‘comparable information’ is standardized language.  Over the past two decades academicians have grappled with issues surrounding clinical terminology.[ii],[iii]  We have learned from these efforts that there are several areas of our medical language and codes that must be integrated and managed if we are to leverage our health system information technology to benefit our profession and the health status of American society.

The Heart of the Problem

Despite the wide availability and acceptance of computer applications in some niches of medicine—such as lab systems that print clear reports and ICU monitors that check for irregular heart beats—the majority of these systems "know” very little about medicine. Critical knowledge, such as what should be done about a cardiac arrhythmia or unexpected lab results, still exists in the minds of practitioners and in the notes they write in patient charts. To date, healthcare IT has failed to get the knowledge into the machine.

The irony is that many of the mundane, repetitive tasks involved in caring for people can be more efficiently and securely handled by computers. Computers can do things more reliably than humans: they rarely make mistakes when they are given precise, correct instructions, and they never forget. And when they are given improved instructions they totally forget the old, inadequate methods of doing things (which we humans call habits) and move ahead.  As recently highlighted by the Institute of Medicine of the National Academy of Sciences, human error and the lack of basic information are harming many people who seek aid in our healthcare system. A strong recommendation for greater reliance on technology arose from these findings.[iv]                                                                                                                                                           

Physicians are the consummate users of medical information and health technology, so why do we fail to effectively harness the power of computers?  One reason is that despite a firm understanding of how to gather clinical information and care for patients, we have used computers almost exclusively to deal with illogical rules on how to submit bills, order lab tests or document treatments in compliance with governmental or payor mandates.  Each of these practical tasks revolves around separate systems with their own codes or languages, none of which actually enhance the delivery of care nor communicate with each other.  Although physicians are trained to document care and propose proper courses of treatment in ‘the chart’, we are unprepared to craft clinical notes to match-up with the more than 20,000 idiosyncratic billing codes. 

This detraction from pure clinical thinking is but one of many reasons physicians are so unhappy with the changes in medical practice today.  Add to this the fear of ungrounded malpractice claims, the barriers to efficient information transfer, the need to re-collect evidence for clinical decision making in an almost ludicrous fashion from encounter to encounter, the often poorly constructed ‘report cards’ based on utilization patterns rather than clinical effectiveness measurements, and one has just a preliminary grasp of the elements driving physicians to lash out at or leave their profession in droves[v].  At the same time, we, and the health plans we practice with, remain rightfully wary of the possibility of Medicare fraud charges because of coding errors.  And despite all of the data collection and submission efforts we attend to, we still have no reliable way to recover accurate information about our efforts and efficacy because of the inherent faults of billing data sets.

Controlled Medical Terminology

The Computer Based Patient Record Institute suggests that clinical terminology is: ‘Standardized terms and their synonyms, which [allow one to] record patient findings, circumstances, events and interventions with sufficient detail to support clinical care, decision support, outcomes research and quality improvement; and can be efficiently mapped to broader classifications for administrative, regulatory, oversight and fiscal requirements.’ [vi]   Work with existing terminologies has allowed several important observations:  First, simple ‘nomenclatures’ (lists) of terms have proven inadequate as CMT schemas because they are difficult to manage and nearly impossible to use for knowledge derivation.  Second, ‘classifications’, which are better because they relate terms to each other in various ways, are still less robust than optimal for efficiently managing, updating and tracking term and concept changes.  And third, ‘knowledge based ontologies’*, in which concepts and their terms are strictly defined and related to each other based on their definitions, seem to be best suited to accomplishing the controlled medical terminology task.[vii],[viii],[ix]

Ontologies provide a human readable yet machine processable conceptual framework that diverse computer systems may utilize for purposes of interoperability.[x] A seamless bridge is built from our ‘words’ to the abstract concepts they represent in the computer.  Note how closely such an approach approximates the dictum that ‘the core utility of any language is to represent concepts by terms or words with meanings that are commonly understood.’[xi]

Software Tools to Manage CMT

Controlled medical terminology is optimally modeled and updated in knowledge-based ‘ontologies’.  A 'lexical runtime engine', formerly called a vocabulary server (VOSER), then serves up the relevant vocabulary to users of applications in the clinical environment.[xii] [xiii]  The ontology, housed in a lexicon model, implies the existence of a data structure that supports both the maintenance and use of controlled vocabulary data.  The model is accessed via a modeling environment, which can be simple or complex.  The measure of a modeling environment is simply its ability to convey the information contained within the model to its users. 

Some of the core facilities provided by a modeling environment include:

   Browsing – A user interface for looking through the lexicon and moving from concept to concept in a logical fashion.

   Searching – Finding and navigating directly to a desired spot in the knowledge model without having “know where it is” at the start.  A robust search interface allows rapid navigation to needed information.

   Editing – Making manual and or/bulk changes to the existing vocabulary for the purposes of local requirements.  Editing systems should have some form of security to ensure the integrity of the underlying knowledge model.

   Configuration Management and version control– As changes are made to the lexicon, careful tracking of the changes is important.

   List management– While the vocabulary must be very rich in content as a whole, the user or application may only need or want a select portion of the 100,000+ concepts at a time.  Lists are used to accomplish this task.

What is Wrong with the Billing Codes We Use Now?

Attempts to cross map, catalogue, and unify medical terminology have occurred historically, resulting in multiple ‘right’ clinical vocabularies.[xiv]  Medicine ultimately found itself infiltrated by proprietary and diverging sets of ‘standard vocabularies’, some in the form of nomenclatures and others as classifications. [xv]  Today there are over 100 vocabulary sets defined internationally for some purpose, and in the United States institutions must deal with from 5-10 of them just to meet reporting or reimbursement requirements![xvi]                                                                                                 

In 1968, at the behest of the Department of Health and Human Services, the United States adopted ICD as the ‘official code set’, and realizing it was woefully inadequate, set about creating ICD-CM (Clinical Modifications). Third party insurers followed suit, but stirred in the equally inadequate CPT codes as requirements for clinical service reimbursement. These chosen schemes, detailed below, lack accuracy, precision and cognitive coherence.  But since the actual communication and documentation of care were largely unaffected by the new add-on coding regulations, clinical professionals swallowed hard and closed their eyes to the downstream effects such derisory coding schemes would have on their professional lives.  An overview of currently used code sets is presented in Table 1.

Progressive healthcare information system developers knew that billing codes were inadequate for electronic clinical documentation and that precise medical words were needed if analysis of information from automated medical records was to improve quality and service in healthcare.  Rather than gravitate toward any existing clinical vocabulary standard, they elected to create their own ‘dictionaries’ or ‘vocabulary sets’.   Their methods represented an improvement over completely un-encoded or free text documentation (which is still the norm in many computer based patient records), but each vendor, working in isolation, created a terminology which could not be read or understood by other vendor systems.  Thus the potential of data exchange and aggregation was impossible.  Managing these highly proprietary code sets, and then trying to compare information gathered in one system to data from another proved quite expectedly problematic.[xvii]

Controlled Medical Terminology and Clinical Infonomics

The term clinical infonomics implies a new science in which care practices, health informatics and national economics are finally recognized as inseparable.  Healthcare delivery is reliant on information technology because clinical decision-making, as well as the reengineering of business and clinical processes, depends on the immediate availability of current accurate information about patients, diseases and populations.  In fact, information technology, along with strong clinical leadership, are key to bringing about a badly needed shift in the entire care delivery system from one based on cost, to one based on value.[xx]  Never before have tools existed that could verify, validate, or debunk treatments, narrow practice variation to a more cost-effective mean, allow discovery and communication of optimal methods of diagnosis or therapy and literally prove the value of clinical interventions. [xxi]

Iatrogenic injury has received a great deal of attention lately[xxii]. During hospitalization in the US it is estimated that as many as a third of patients suffer from complications related to their care [xxiii]; between 5% and 13% of hospital admissions result from the adverse effects of diagnosis or treatment [xxiv], and some 70% of iatrogenic complications affecting more than 1.3 million hospitalized patients annually are preventable![xxv] There are from 44,000 to 180,000 treatment related deaths in hospitals every year [xxvi], [xxvii], [xxviii] and of life threatening/serious adverse drug events, the most common failures relating to these errors are system errors involving drug knowledge dissemination, drug dosing, patient identity checking and patient information availability, suggesting inadequate systems, rather than individual inadequacy, as causes for error.[xxix]  In a survey of attendees conducted at the 2000 Healthcare Information Management and Systems Society, it was discovered that a startling 98% of respondents believed that controlled medical terminology would be important to mitigation of iatrogenic mishaps.[xxx]

As Care Providers how does CMT Benefit US?

Benefits include faster, more accurate billing, more informed patient management, faster, clearer and more precise documentation (the ultimate in ‘decision support’), lower malpractice premiums and improved knowledge.  We cannot develop nor exercise the ability to move healthcare forward, as clinicians or administrators, without the informatics tools, language and communication standards outlined above.  Without technology that “listens to and understands us” we cannot act in accordance with evidence based guidelines and principles, narrow practice variation and improve knowledge of appropriate and cost-effective techniques for managing health. 

So what can one do?

The tools and language content sets exist today to help with this problem, but unless controlled medical terminology and the informatics infrastructure suggested above are widely embraced, we cannot possibly make good decisions about whether interventions are in a range of from ‘possibly beneficial’ to ‘highly indicated’, determine physiologic from biologic outcomes based on measurements, or assess clinician assessment versus patients perceived health outcomes (including pain, suffering, disfigurement, disability).[xxxi] Without a common language we cannot improve the quality of healthcare, no matter how many times we say the words, nor even approach evidence based medicine on anything but a microcosmic scale.  To get there, the steps one can take as an individual are relatively simple:

1.       Insist that clinical computing systems to be purchased for your ambulatory or in-patient environments be based upon a controlled medical terminology, (not ‘our’ standard terminology); and one which meets the requirements of good CMT.

2.       Seek greater understanding of the clinical systems that are being purchased for use by your group, because they are not going away; they are enormously helpful once you get over the initial fear of using them, and if you insist that they collect good information they will be tools you can use rather than weapons that can be used against you.

3.       Become familiar with E-mail interactions and the formalities surrounding them because they will become a norm in the future, and even a billable one at that [xxxii], but do NOT become comfortable with the fact that these ‘un-encoded’ communications are sufficient, for all the same reasons un-encoded information in the medical record are insufficient.

4.       Don’t be afraid of the change, in either the technology or the language. Having lived through it, and having helped many clinicians through the transition, we know it can be done to our own great advantage, and to the benefit of healthcare in the United States population in entirety.

Without CMT- enabled health informatics technology we are ‘sewing in the dark’; we have no standard methods for evaluating either our work or the preferences of our patients. Controlled medical terminology essentially ‘turns on the lights’. Failing to move this direction with our language and information technology dooms us to control by those who have more data, but far less knowledge.