Can Health Care Information Technology Adapt?

 

 

Prepared for

 

Enterprise Strategy

Veterans Health Administration

Department of Veterans Affairs

810 Vermont St., N.W.

Washington, DC.  20420

 

 

Tom Munnecke

Science Applications International Corporation

10260 Campus Point Ct.

San Diego, Ca.  92121

(858) 756 4218

[email protected]

Version 1.0 Jan 30, 2002

 Available at http://www.munnecke.com/papers/D23.doc

 

Complexity, Information, and our Ways of Understanding. 2

How Adaptive Have Our Systems Been?. 4

Lessons Learned from Y2K Issue. 5

The Need for Adaptability. 5

The Transition to Genomics and Proteomics. 6

Transition from Biological to the Genetic Era of Medicine. 8

Science and Biological Medicine. 9

A Crude Look at the Whole. 12

Approaches for Adaptability. 14

Appendix B: L-Systems. 17

 


Complexity, Information, and our Ways of Understanding

Imagine a scientist trying to understand a symphony played by an orchestra.  The scientist might start by putting a microphone in the audience, recording the sound waves as they come from the orchestra.  These would be digitalized into bit streams to be analyzed by a computer to look for patterns.  When the first attempt failed, the scientist might increase the sensitivity of the microphone and increase the sampling rate, generating even more data.  When this attempt failed, the scientist might add 15 more microphones in different sampling positions, hoping to finally get enough data to understand the music.

If the scientist wandered up on stage, however, the score used by the conductor would be obvious. The notes of the symphony could be represented with kilobytes of information. The gigabytes of data collected by the array of microphones made it difficult to understand what was obvious and simply represented on the musical notation. 

The conductor’s score represented a language which was interpreted to become the symphony.  The technique of recording all the emanations of the instruments as discrete events and digital “snapshots” lead to an ever-increasing labyrinth of complexity.  More data created more complexity.  Having a musical language, however, creates a simpler way of representing what otherwise would be an enormously complex undertaking.

            Our current situation in health care can be likened to that scientist in the auditorium. We are already receiving an overwhelming array of data and information, and we know that with the advent of genetically based medicine this flow will increase dramatically, perhaps by orders of magnitude.  This information may be of a fundamentally different nature than what we are receiving today.  Our current models of understanding health and medicine may undergo fundamental revisions.

            Perhaps this new technology will appear gradually, merely being minor additions to the formulary and some additional lab tests.  Current physicians will be able to study papers and take some CME courses to understand it.  Perhaps some new specialties will arise within the existing framework of health care delivery.

            On the other hand, these changes may have far greater scope than is currently imagined.  Issues of privacy, politics, fear of the unknown, and media frenzies may swamp scientific evidence and clinical research.  New knowledge may emerge from the lab and be driven by direct to consumer marketing activities faster than our current knowledge system, to the extent that it exists, can assimilate them.

            Efforts to automate the medical record go back at least 30 years, yet there is no wide-spread success.  Clinical knowledge can take 17 years to disseminate for general use.  In today’s world of “Internet time” these numbers are amazingly long.  Medicine and our health care system stands on the brink of waves of rapid change, yet its information and knowledge infrastructure stands as one of the longest running failures in the information technology industry.

            Like our symphony scientists getting overwhelmed by the data generated by their array of microphones, medicine and health care are being overwhelmed by inappropriate information and knowledge structures.  Our way out of the exploding complexity we face is through smarter information structures, and perhaps wandering out of the audience to discover a higher level language – a “score” which simplifies our quest.

Waves of Accelerating Change

 

There is a huge gap between technology and our ability to apply it in health care, much of which reduces to our ability to handle information:

 

“Health care today is characterized by more to know, more to manage, more to watch, more to do, and more people involved in doing it than at any time in the nation’s history.  Our current methods of organizing and delivering care are unable to meet the expectations of patients and their families because the science and technologies involved in health care – the knowledge, skills, care interventions, devices, and drugs – have advanced more rapidly than our ability to deliver them safely, effectively, and efficiently.[1]

 

We can expect these technological changes to continue at an increasing rate from many different directions.  Technology in general is accelerating:

[We are approaching] the "perfect storm" of the converging exponentials of bio-X, nanotech, and information technologies/telecommunications. They will cause more change in less time than anything humankind has ever witnessed.[2]

Specific advances in proteomics will have dramatic effects on clinical systems:

The next technological leap will be the application of proteomic technologies to the bedsideThis will directly change clinical practice by affecting critical elements of care and management.  Outcomes may include early detection of disease using proteomic patterns of body fluid samples, diagnosis based on proteomic signatures as a complement to histopathology, individualized selection of therapeutic combinations that best target the entire disease-specific protein network, real-time assessment of therapeutic efficacy and toxicity, and rational modulation of therapy based on changes in the diseased protein network.[3]

Our understanding of interactions between drugs and genotype-specific activities will also trigger tremendous changes in health care:

Pharmacogenomics requires the integration and analysis of genomic, molecular, cellular, and clinical data, and thus offers a remarkable set of challenges to biomedical informatics. These include infrastructural challenges such as the creation of data models and data bases for storing this data, the integration of these data with external databases, the extraction of information from natural language text, and the protection of databases with sensitive information. There are also scientific challenge in creating tools to support gene expression analysis, three-dimensional structural analysis, and comparative genomic analysis.[4]

How Adaptive Have Our Systems Been?

Given these dramatic and accelerating forces on our health care system, it is instructive to look at how well the current system adapts to change.  Past history does not indicate a particularly adaptive response to even simple issues:

 

·        An average of 17 years is required for new knowledge generated by randomized controlled trials to be incorporated into clinical practice.[5] 

 

·        Changing our computer systems to deal with the Year 2000 (Y2K) problem cost the United States an estimated $100 billion and the federal government $8.5 billion.[6]  Yet the basic problem, changing a date field from 2 to 4 digits, was at core a simple programming problem.

 

·        The feedback loop between treatment and its effectiveness has not always worked well:

“By the time Moniz and Hess shared the Nobel Prize in 1949, [for inventing the frontal lobotomy] thousands of lobotomies were being performed every year.  Yet by the end of the 1950s, careful studies revealed what had somehow escaped the notice of many practicing physicians for two decades: the procedure severely damaged the mental and emotional lives of the men and women who underwent it.  “Lobotomized” became a popular synonym for “zombie,” and the number of lobotomies being performed dropped to near zero.”[7]

·            Despite 30 years of aggressive attempts to create an electronic medical record, this goal is still elusive.  For example, in 1991, the Institute of Medicine’s Committee on Improving the Patient Record set a goal of making the computer-based patient record a standard technology in health care by 2001.[8] Given the pressures of cost cutting, continuous changes in the industry, and increasingly complex issues relating to privacy, liability, bioterrorism, and genetic information security, it is likely that our ability to achieve this goal is diminishing, rather than increasing.  One reason for this continued failure is the brittleness of the technology we are attempting to use.  It is simply not adaptive enough for the task.  A Critical Time to Act

  •        

Lessons Learned from Y2K Issue

The calendar change to the new millennium triggered a Y2K problem of immense magnitude.  Some pPredictions ofed a global recession as computer systems, electronic funds transfers, and transportation systems shut down did not occur.  The fact that the world could bewas brought to the brink of such The global response to checking for errors for the change of century illustrates how brittle our software infrastructure is.  catastropheYet the Y2K problem was a relatively minor change to the system:s is remarkable due tofor the following reasonsissues:

1.            1.      The root problem was trivial – expanding a date field from two to four digits was something that could be accomplished by even the most inexperiencednovice programmers.  The problem was easily stated and recognized

2.            2.      We had perfect foreknowledge of the problem.  The fact that there would be a year 2000 was always known.  The arrival of Jan. 1, 2000 was not a surprise.

3.            3.      The problem was reversibleWith certain exceptions (for example, the safety of a factory control system), problems which may have been encountered during the changeover would have triggered delays in operation.  For example, Eeven if an airline reservation system failed, for example, service could be eventually restored and the system could returned to normal.

4.            It illustrated the network effect.  The problem did not only exist in isolated computer systems, but also in all of the interconnections between them.  Electronic funds transfer systems, for example, connected the world’s banking systems together, and a failure in a critical component could have cascaded into other systems.  What started out as isolated, enterprise-only applications had become globally connected.

 

Nevertheless, avoiding the this problemY2K problem cost the United States an estimated $100 billion and the federal government $8.5 billion to avoid the Y2K problem.

The Transition to Genomics and Proteomics

            The world is facingnow faces another mega-issue, based on our rapidly increasing understanding knowledge of DNAgenomicsAs w.  We are just beginning to unraveling the complex mysteries of the gene, .  O our understanding of genomics and proteomics could will have dramatic effects on our personal health and our health care system.   Compared to what we went through with Y2K:This problem has far greatert immediate and long term consequences:

1.            1.      The root problem issue is immense.  The field of bBioinformatics is one of the most challenging computer science problems today, pushing the state of the art in computer science, supercomputing, mathematics, biology, and complexity sciencesIt is pushing the technological limits of supercomputing, database storage, knowledge management, and standardization.Notions of privacy will extend beyond individuals to familial membersrelatives, not just individuals. 

2.            It opens up entirely new problems of privacy.  Notions of privacy will extend beyond individuals to entire families.  Relatives will become trustees of each other’s genetic information.  Information which was not sensitive in one era of knowledge may become highly sensitive with future discoveries.  Genotype testing could discover that a person is “difficult to treat” or “more expensive to treat” which could impair their future ability to get health insurance, a job, or other adverse events.  Furthermore, genetic samples released earlier could be reinterpreted with new knowledge, so that an informed consent at one time could lead to future negative effects in the future, beyond the expectations of the patient at the time of signing.  Thus, what is not sensitive today could become very sensitive tomorrow. New social questions and ethical problems will emerge regarding race and ethnicity.[9]. – what if a genetic privacy mechanism detects a biological father different from that person’s named father, for example?

3.            2.      We don’t know what we don’t know.  We can only expect surprises from our research and discoveries.  How discoveries will affect with existingcurrent medical practices, knowledge, and the public is unpredictable. “It is not entirely clear how many of the 35,000 genes assigned in the rough draft of the human genome are relevant to drug response (or even how to define relevance)[10]

4.            The tempo of knowledge creation is increasing.  Given that anIt now takes an average of 17 years is required for new medical knowledge generated from randomized controlled trials to be completely incorporated into clinical practice.  In the future, it is likely that new information will be created, and possibly madebecome obsolete in this time, by the time it is put into practiceThe Eeven what we think we do “know” may not be true, given the paradoxical nature of self-referential systems such as DNA..

5.            3.      The problemBioterrorism has entered the picture. may be irreversible.  Changing the evolution of the humans species, for example, is not something which can could be subjected to clinical trials.  New pathogens may might be created, either accidentally orperhaps by terrorists, which, once released, cannot be withdrawn.. Warfare has always attacked the means of a society’s production; the more productive the genomic revolution becomes, the more attractive it becomes for nefarious purposes.

6.            4.      The problem is continuous.  While the Y2K problem climaxed ended on a specific date, this problemthe coming changes in medicine will may be continuousaffect current and future generations on a continuingous basisRisks to future generations need to be balanced against benefits to the current one.There may not be a specific date on which will trigger action.

7.            5.      Our current scientific method may not be powerful enough. Current notions of causality, repeatability, and objectivity in scientific experimentation may not be capable of expressing cascades, singularities, and self-referential processes inherent in genetic and complex adaptive systems. At some point, the reductionistic model of scientific research will collide with the paradoxes of self-reference inherent in understanding DNA.  Biological notions of causal effects, “root cause analysis,” and other deterministic approaches may not be able to cope with feedback systems, parallelism, adaptation, and evolutionary processes.  Yet we have little or no information infrastructure to record or understand such effects.

8.            Basic notions of health and disease may change.  In the same way that it is impossible to create a one-to-one map between Chinese traditional medical concepts and Western biological allopathic medical models, it is likely that our genetic understanding of medicine will create new ways of understanding health and disease which cannot be mapped one-to-one to our current understanding of disease.  We may discover new cascades, networks, and metaphors which are simply beyond the ability of our current information systems to express.  The limited expressiveness of our current systems may in fact limit our ability to discover these effects.  We may find it necessary to represent notions of adaptation, evolution, learning, and feedback on the individual level, or even smaller components of an individual.  We may find a need to collect and treat families, communities, or collections of individuals with a common genotype, blending scales of intervention in ways not understood by current information systems.

1.                   

9.            6.      Network effects may dominate.  The World Wide Web has exploded into a leading global communications medium in just a decade, exploiting the network effect of connectivity.  Simple initial conditions can be amplified to create a huge cumulative effect.  This may create emergent properties which are not predictable from the outset, in the same way that it would have been impossible to predict the effects of the three initial web concepts of URL, HTTP, and HTML on global communications and commerce.  We have little understanding of how cascades operate, yet this is the core of the genetic process. We are connecting society, ourselves, our knowledge, and our health care systems in entirely new ways and subjecting them all to the network effect.  For example, all of our scientific knowledge about Cipro could not have predicted its emergence in the Anthrax scare in September, 2001, and its effect on microbial resistance. 

10.        The complexity of the current health care system is already near a breaking point.  “The U.S. health care system as currently structured is so complicated and rife with economic conflict that every attempt to simplify if actually complicates it further.”[11]  Given the perverse incentives rife within the industry, it is difficult to understand how the sweeping changes facing us can be introduced into the system.  “Over the years, certain elements have attached themselves to the health care system like a fungus…these permanent growths thrive on the health care system’s essential complexities and slowness to adapt.[12]

             The changes mandated to our information infrastructure cannot be accommodated as simply as adding a few characters to a field, as was done with Y2K.  It is problematical whether they can even be accommodated within the current information systems technology we use.  Accommodating these changes will require some fundamental rethinking of health care system, its science, and the media by which medical knowledge is communicated.

The Need for Adaptability

Adaptability of our information systems can be seen to be a core need of our information systems, particularly if it is viewed in a broader perspective:

 

Adaptability over

Equates to

Users

Security

Patients

Privacy

Time

Maintenance

Sites

Portability

Hardware

Hardware Independence

Data bases

Data independence

Medical Model

Transition from biological to genetic models of medicine

 

Brittleness in any of the above areas tends to indicate brittleness in the other categories.  Viewed from this perspective, CIO’s budgets are largely controlled today by the cost of adaptation.  Prospects for future activities will increasingly be controlled by concerns of adaptation: privacy and security being “sleeper” issues which are increasingly becoming drivers in both medicine and general information concerns.

 

 

Transition from Biological to the Genetic Era of Medicine

            We can look at three great waves of modern medicine and health:

 

1.      Sanitary – understanding germs, the role of public health in creating sanitary sewagage systems and water supplies.

2.      Biological – understanding the components of living things such as cells, organs, and biochemical processes.  Understanding diseases and symptoms of failure of these components, and applying scientific experimentation to discover their treatment and cure.  The success story of allopathic medicine.

3.      Genetic – understanding health and disease based on our understanding of the basic genetic makeup of living things as expressed in DNA codes.  This phase is just beginning; we are just now beginning to get the raw data with which to begin our deeper understanding of the genetic basis of health.

 

Medical Era

Scale

Key Theories

Validation of Theory

Sanitary

Cities, communities

Germs, antisepsis

Population statistics

Biological

Individuals, organs

Allopathic medicine, evidence-based medicine

Experimentation, clinical trials, outcomes assessment

Genetic

Genes, base pairs

DNA replication, genomics, proteomics

???

 

Science and Biological Medicine

            Claude Bernard was one of the early thought leaders in bringing about the biological age of medicine.  In the late 1800’s there was considerable turmoil in thinking about living things, including the groundbreaking theory of evolution by Charles Darwin.  There were those who thought that life was caused by an élan vital, a mysterious life force which could be contacted in the spirit world through séances.  Superstition and religious beliefs clashed with science.  The scientific method was a crucial ally in this effort.  It is interesting, however, to note that 150 years after the introduction of the scientific method in medicine that we are still calling for “evidence-based medicine.”

Bernard, faced with trying to create experimentally provable medical knowledge, used scientific laws and methods for proof:

“In living bodies, as in organic bodies, laws are immutable, and the phenomena governed by these laws are bound to be the conditions on which they exist, by a necessary and absolute determinism…if they [experimenters] are thoroughly imbued with the truth of this principle, they will exclude all supernatural intervention from their expectations; they will have unshaken faith in the idea that fixed laws govern biological science; and at the same time they will have a reliable criterion for judging the often variable and contradictory appearance of vital phenomena…for the facts cannot contradict one to another, they can only be indeterminate...  Facts never exclude one another, they are simply explained by differences in the conditions in which they were born. [italics added][13]

            It is interesting that Bernard is countering the vitalists’ belief system with his own admonition of “unshakable faith” in fixed laws of biology.  Would research conducted under principles of “absolute faith” ever be capable of refuting the belief system of those who professed it?  He was replacing faith-based vitalism with faith-based experimental medicine.

            A fundamental tenet of Bernard’s scientific faith was to slice the relationship between the observer and the subject.  “Experiment becomes the mediator between the objective and the subjective, that is to say, between the man of science and the phenomena which surround him.”[14]

            Subsequent discoveries by scientists and mathematicians, however, have made discoveries which undercut Bernard’s “unshakable faith.

Self-Referential Systems

The line between subject and observer is not as distinct as Bernard would have it.  DNA, for example, is a self-referential system.  DNA encodes a mechanism which interprets a code which constructs the mechanism.  Is DNA the subject or the object of an experiment?  If we “slice” the experimental process to assume that it is one, how do we know that we are not simultaneously changing the other?  This is a little like looking into a mirror which looks into a parallel mirror, creating perfectly parallel reflections.  If we stick our face between them to see what we see, we see the injection of our face in the mirrors, not the mirrors without our observing face.  This leads to paradoxical situations which can only be “objectified” by arbitrarily cutting off the self-referential loop.  We just define the observation to be a specific iteration of the loop, and are then able to carry on with a “consistent” theory of how things work.

Mathematician Kurt Goedel created a formal mathematical model of this problem in the mid 20th century, called Goedel’s Incompleteness Theorem.  Roughly stated, any language capable of referring to itself is capable of expressions which can neither be proven true or false within that language.  For example, the sentence, “This sentence is false.” is a paradox which cannot be proved true or false within the language in which the sentence is written.  Like the DNA code which creates both the machine to interpret the code and the code itself, we are dealing with the paradox of self reference.

These paradoxes can be resolved by creating higher level language, one which “looks down” on the lower level language which resolves the self-referential statement to be either true or false. Users of that language can then go about their business of maintaining consistency and truth within that language. 

However, this only bumps up the problem one level.  If this higher level language is self referential, then it can contain a self-referential statement, which can only be proven true by yet another, higher level language.  This leads to an infinite regress of higher level language. 

If DNA is the language of life, then it is tied to this infinite regress.  This regress can be ignored to some extent, but a full understanding of DNA’s meaning will eventually revolve around our ability to deal with the paradox of self reference.  It is doubtful that Bernard’s “unshakable faith” in objectivity, which served us so well in the biological era of medicine, will stand unshaken as we unravel the meaning of DNA and life.  In fact, probing the paradoxes of self-reference and object/subject “observation” may well be the path towards understanding.  Sooner or later, our information infrastructures will have to deal with issues of self-reference, feedback, and recursion.  Information systems to date have dealt with this problem by simply ignoring them.  Representing and understanding adaptive, learning, evolving systems is very difficult, yet it is the key towards a richer information infrastructure. If we are ever to discover the “score” in health care and medicine, it is likely to be closely tied to our understanding of the cascades of effects caused by DNA.

Fractals

Mandelbrot introduced the notion of fractal geometry in the 1970’s.  Fractal objects do not neatly fit into “normal” dimensions.  The scale at which we examine a fractal object affects the measurement.

For example, suppose we want to tie a rope around an island, touching the entire shoreline.  How much rope would we need?  This would seem to be a simple question, but what if the island is highly irregular, with deep bays and promontories? 

Do we wrap the rope from one promontory to the next, or have it follow the coast line.  If we follow the coastline, do we follow smaller indentations?  If we follow them, do we follow river inlets?  If so, do we follow the branches of the rivers into streams?  If we follow streams, do we follow individual rocks?  If we make the rope smaller, so that it is a string, do we follow smaller rocks?  If we make the string the thickness of a hair, do we follow even smaller pebbles?

If we are dealing with a fractal or self-similar object, the question of length requires that we also specify the scale at which we examine it.  A non-fractal object, such as a house, does not behave this way.  The more we measure the outside of a smooth house, for example, the more accurately we converge on the ‘true’ circumference.  We have well-developed statistical techniques to deal with measurement error and normal Gaussian distributions.  We don’t have well developed techniques to deal with fractal objects.  Yet examples of self-similar objects in medicine abound, such as dendrites of neurons, airways in the lungs, ducts in the liver, the intestine, placenta, cell membranes, and energy levels in proteins.  They also appear in the dimension of time, such as voltages across the cell membrane, timing of the opening and closing of ion channels, heartbeats, and volume of breaths.[15]

Bernard’s faith in single numerical values by objective observers is uprooted again.  The particular scale of observation chosen by the observer affects the measurement.  Despite this knowledge, we regularly accept statements such as “the length of the coast of England” without further qualification. Despite knowledge gained in the fields of mathematics and logic regarding self-referential systems and fractals, our systems continue to operate as if they did not exist.  Fractals introduce an entirely new appreciation of the issue of scale.  Appendix A illustrates some scientific discussion of this.

Media Driven Medicine

Challenges to our health care system are just from scientific discourse and discoveries of new technology.  There is an active “antiscience” community which challenges scientific facts, not on the basis of experimentation and proof, but underlying belief systems. 

For example, many believe that 60 hz powerline emissions cause cancer, despite scientific evidence to the contrary.  Parents will take their children out of schools near powerlines out of fear that they cause cancer.  The extra distance their children travel each day is a very real risk to their health, yet they choose to expose them to this rather in hopes of avoiding an imaginary one.  The White House Science Office estimates that the total cost of the power line scare, including relocating power lines and lost property value to be $25 billion,[16] none of which is supported by scientific research. 

Antiscience can be driven by the media.  For example, Nobel Laureate Irving Langmuir examined the experimental procedures of parapsychologist R.J. Rhine.  He discovered that Rhine had left out the observations of subjects who Rhine suspected of deliberately guessing wrong.  Rhine believed that these people actually had parapsychological powers and intentionally guessed wrong because they disliked his work.  Langmuir attempted to explain the obvious methodological problems in Rhine’s methods to a reporter, who wrote that a “famous Nobel laureate” was looking into ESP.  Rather than correcting Rhine’s error, Langmuir’s attention inflamed public acceptance of his work, unintentionally adding ‘scientific’ credibility to Rhine’s unscientific work.[17]

Even where there is clear, precise scientific information available, communicating this information to patients and providers in the context of media-driven medicine will become increasingly difficult.

Direct to Consumer Advertising

            Drug companies are well aware of the effects of media on the driving public demand.  Direct to consumer (DTC) advertising has proven to be a very effective method for drug companies to sell more of their products.  Given the cost of developing new drugs, there will be increasing pressure for these companies to increase their DTC advertising, further driving the media-driven medicine loop.  For example, in the weeks after the September 11 attacks and anthrax scare, CNN advertised an offer for a 30 day free supply of Prozac, available via the WebMD web site.  This media loop is based on information flows entirely independent of the traditional medical information domain, yet it is a very significant driver of health care activities.  The information infrastructure of tomorrow must be able to accept and interact with a much broader range of issues, which happen in “internet time” scales, not decades.

Paths Towards Adaptive Systems

            One of the most common health decisions in our society occurs perhaps 1 million times each morning – children complaining to their parents that they are “too sick to go to school.”  Parents must sort through a plethora of vague complaints about stomach aches, headaches, tiredness, and nausea.  They consider the children’s history of complaints, events at school, and the veracity of their claims.  They must come to a decision, “Is this child sick?”

            Having access to millions of medical terms, even if they had perfect knowledge of medicine, would not necessarily make the decision easier.  They must deal with a coarse-grained distinction, what may be called “a crude look at the whole.”

            This term is used by Dr. Murray Gell-Mann, Nobel laureate for his work in physics and the discovery of the quark.  He has focused in recent years on the issues of complexity and scientific thought, and was one of the founders of the Santa Fe Institute, in part as a “rebellion against the excesses of reductionism.”[18]  Rather than viewing a system strictly as a hierarchy of components-within-components, he explains an alternative form of self-organization:

“Scientists, including many members of the Santa Fe Institute family, are trying hard to understand the ways in which structures arise without the imposition of special requirements from the outside.  In an astonishing variety of contexts, apparently complex structures or behaviors emerge from systems characterized by very simple rules.  These systems are said to be self-organized and their properties are said to be emergent.  The grandest example is the universe itself, the full complexity of which emerges from simple rules plus the operation of chance.”[19]

            This approach to complex systems thinking perhaps offers an alternative to exploding levels of complexity.  Rather than trying to understand healthcare as an exploding catalog of emergent properties, we could uncover simpler order which are the generators of the multiplicities which become apparent in the taxonomies.

            The conductor’s score could be considered to be a “crude look at the whole” of the music being played by an orchestra.  The language is a compact representation of the complex interaction between conductor, musicians, musical instruments, and the auditorium. 

            We do not currently have a language which describes the biology of life at the same level of simplicity and compactness as a conductor’s score.  We can, however, do a thought experiment.  If such a language were to exist, what might it look like?  Are our scientific data that we are collecting today like the scientists in the auditorium collecting ever-increasing volumes of sound waves?  Is there a “score” somewhere waiting to be discovered which would explain with great simplicity the multiplicity of things we now see as separate, independent facts and data?  In the same way that Newton’s F=MA explained falling apples and orbits of the planets, are there linkages between things which otherwise may seem disparate?  What would be the building blocks of such a higher level linguistic “shell” within which to describe biological and health processes?  What can we do with our information systems today to help find these building blocks?

            Computer simulations and computational biology are exploring some of these issues:

 

We also know that agents that exist on one level of understanding are very different from agents on another level: cells are not organs, organs are not animals, and animals are not species.  Yet surprisingly the interactions on one level of understanding are often very similar to the interactions on other levels. How so?  Consider the following:

 

·         Why do we find self-similar structure in biology, such as trees, ferns, leaves, and twigs?  How does this relate to the self-similarity found in inanimate objects such as snowflakes, mountains, and clouds?  Is there some way of generalizing the notion of self-similarity to account for both types of phenomena?

 

·         Is there a common reason why it’s hard to predict the stock market and also hard to predict the weather?  Is unpredictability due to limited knowledge or is it somehow inherent in these systems?

 

·         How do collectives such as ant colonies, human brains, and economic markets self-organize to create enormously complex behavior that is much richer than the behavior of the individual component units?

 

·         What is the relationship between evolution, learning, and adaptation found in social systems?  Is adaptation unique to biological systems?  What is the relationship between an adaptive system and its environment?

 

The answer to all these questions are apparently related to one simple fact: Nature is frugal.  Of all the possible rules that could apparently be used to govern the interactions among agents, scientists are finding that nature often uses the simplest.  More than that, the same rules are repeatedly used in very different places:

 

·         Collections, Multiplicity, and Parallelism

·         Iteration, Recursion, and Feedback

·         Adaptation, Learning, Evolution[20]

           

Of particular interest is Flake’s comments about the places in which he finds universal rules of nature: Collections, Multiplicity, and Parallelism, Iteration, Recursion, Feedback, Adaptation, Learning, Evolution.  The describe features which are most difficult to describe within today’s standard database framework, the relational database.

Relational data base technology assumes a standard form of data layout, into well-defined two dimensional tables with specified rows and columns.  The meaning of a datum is closely related to its structural position within the table.  Databases are designed by “pigeonholing” data elements into tables, and relationships are expressed between the pigeonholes through additional tables.  These restrictions on freedom of expression allows the relational calculus to be used, but this comes at the cost of allowing individual data elements to have properties and characteristics which are not shared by other column “mates” in the structure.

Appendix B presents some illustrations of how a higher level language can be used to describe complex growth and adaptation issues.

Pigeonholing medical information into relational data structures, therefore, inhibits it ability to express the universal rules which Flake mentions above.

 

Conclusion

            Just as our symphony scientists were overwhelmed with data from their expanding array of microphones in the auditorium, our health care system faces the risk of being overwhelmed with a deluge of data in quantities and forms which our current information systems can record, but not comprehend.  And just as the conductor’s score provided a powerful, concise shorthand for understanding the symphony, our information technology must be used to express higher level languages and definitions of the patterns, adaptation, learning, evolution, and feedback mechanisms of health and medicine.


Appendix A: An Example of Biological Patterns

One set of patterns which may point to a higher level of language has been described by Nobel Laureate Gerald Edelman, who writes on the issue of degeneracy and complexity in biological systems.  Degeneracy, in his paper describes the ability of elements that are structurally different to perform the same function or yield the same output: 

 

Despite the fact the biological examples of degeneracy abound, the concept has not been fully incorporated into biological thinking.  We suspect that this is because of the lack of a general evolutionary framework for the concept and absence, until recently, of a theoretical analysis.[21]

 

He goes on to discuss the occurrence of degeneracy as occurring at many different scales:

 

·         Genetic code (many different nucleotide sequences encode a polypeptide)

·         Protein fold (different polypeptides can fold to be structurally and functionally equivalent)

·         Units of transcription (degenerate initiation, termination, and splicing sites give rise to functionally equivalent mRNA molecules)

·         Genes (functionally equivalent alleles, duplications, paralogs, etc, all exist)

·         Gene regulatory sequences (there are degenerate gene elements in promoters, enhancers, silencers, etc.)

·         Gene control elements (degenerate sets of transcription factors can generate similar patterns of gene expression)

·         Posttranscriptional processing (degenerate mechanisms occur in mRNA processing, translocation, translation, and degradation)

·         Protein functions (overlapping binding functions and similar catalytic specificities are seen, and ‘‘moonlighting’’ occurs)

·         Metabolism (multiple, parallel biosynthetic and catabolic pathways exist)

·         Food sources and end products (an enormous variety of diets are nutritionally equivalent)

·         Subcellular localization (degenerate mechanisms transport cell constituents and anchor them to appropriate compartments)

·         Subcellular organelles (there is a heterogeneous population of mitochondria, ribosomes, and other organelles in every cell)

·         Cells within tissues (no individual differentiated cell is uniquely indispensable)

·         Intra- and intercellular signaling (parallel and converging pathways of various hormones, growth factors, second messengers, etc., transmit degenerate signals)

·         Pathways of organismal development (development often can occur normally in the absence of usual cells, substrates, or signaling molecules)

·         Immune responses (populations of antibodies and other antigen-recognition molecules are degenerate)

·         Connectivity in neural networks (there is enormous degeneracy in local circuitry, long-range connections, and neural dynamics) , are all degenerate)

·         Sensory modalities (information obtained by any one modality often overlaps that obtained by others)

·         Body movements (many different patterns of muscle contraction yield equivalent outcomes)

·         Behavioral repertoires (many steps in stereotypic feeding, mating, or other social behaviors are either dispensable or substitutable)

·         Interanimal communication (there are large and sometimes nearly infinite numbers of ways to transmit the same message, a situation most obvious in language)

 

Edelman is noting the existence of a continuum of scale, coupled with scale-independent characteristics.  This phenomenon has been discussed earlier in this sequence of papers, calling the scale-independent properties intrinsics.[22]  Edelman’s work touches on notions of fractals, characteristic scale, and a general “framework” for dealing with complexity in biological systems.




Appendix B: L-Systems

            If there were a score which could simplify our understanding, what might it look like?  One glimpse of such a language was invented by Lindemeyer [23] to describe plant morphology.  This language can be examined as a model for other, more powerful languages which express the dynamics of growth, feedback, and adaptation in living things.

He created a concise language which describes a plant as an iterated sequence of segments, created according to a production grammar.   A formal description is:

L-System     A method of constructing a fractal that is also a model for plant growth. L-systems use an axiom as a starting string and iteratively apply a set of parallel string substitution rules to yield one long string that can be used as instructions for drawing the fractal. One method of interpreting the resulting string is as an instruction to a turtle graphics plotter.[24]

            The following table illustrates how an image of a tree can be drawn with progressively greater detail:      

 

Depth =1

Depth=2

Depth=3

Depth=4

Depth=5

Depth=6

 

These trees are generated by the L-Systems rule:

 

 f=|[5+f][7-f]-|[4+f][6-f]-|[3+f][5-f]-|f

 

The difference between the images is the depth of the application of the rule.  Where depth=1, the tree appears to be just twigs.  This is the basic pattern of the tree, which is used in successive depths.  Each segment of the tree for depth=1 has been replaced with a smaller version of the basic pattern for depth=2.  Each succeeding layer of depth repeats this process.

This approach to expressing plant morphology has several interesting properties.

 

1.      A simple formula, only 37 characters long, is able to describe the shape of the tree.

2.      The formula stays the same, even at different levels of detail.  The complexity of the figure drawn stays the same, only the depth of drawing changes.  Thus, an apparently complex tree structure (depth = 5) is really just a simple structure drawn to a greater depth.

3.      If we were to try to understand the tree by cataloging and studying each twig, the number of points to be studied becomes greater by about a factor of ten per depth.

4.      Looking at the expression at a lower depth allows us a “crude look at the whole” while iterating more deeply allows us to see greater detail.

 

 

Some examples of how L-Systems can be used to describe complex objects can be found across the web:

 

 

 

            An example of an L-systems growth of a plant may be found at:

http://www.cpsc.ucalgary.ca/projects/bmv/vmm/QT/Greenash/fullview.qt Understanding L-Systems is possibly best viewed graphically. An interactive L-system program can be viewed at http://mitpress.mit.edu/books/FLAOH/cbnhtml/java.html

 

 

 

 

[25]

L-systems can be used to illustrate adaptation between multiple growing things:

 

 

 

This computer simulation shows two L-Systems generated trees which grow next to each other, competing for space, illustrating the process of adaptation.

 

This simulation shows trees competing for both light and space…[26]

 

The above drawing illustrates how L-Systems can draw spiral phylotaxis in plants. [27]



[1] Robert Woods Johnson Foundation , 1996, quoted in Institute of Medicine’s Crossing the Quality Chasm, a New Health System for the21st Century, 2001, p. 25

[2] Interview with Larrry Smarr, director of the of the UCSC Supercomputer Center, San Diego, http://www.fortune.com/indexw.jhtml?channel=artcol.jhtml&doc_id=205301

[3] Liotta, Lance A., et al, Clinical Proteomics, Personalized Molecular Medicine, JAMA, Nov. 14, 2001, Vol 286, No. 18, pp 221-2214

[4] Altman, Russ B. and Klein & Teri E. Challenges for Biomedical Informatics and Pharmacogenomics, http://www-smi.stanford.edu/pubs/SMI_Reports/SMI-2001-0898.pdf

[5] Balas, E Andrew and Suzanne Boren, Managing Clinical Knowledge for Health Care Improvement, Yearbook of Medical Informatics, National Library of Medicine, Bethesda, MD:65-70, 2000

[7] Millenson, Michael, Demanding Medical Excellence, Doctors and Accountability in the Information Age, University of Chicago Press, 1997, p.108

[8] Institute of Medicine, The Computer-Based Patient Record: An Essential Technology for Health Care Revised Edition, 1997.

 

[9] Rothstein, Mark A. and Epps, Phyllis, Ethical and legal implications of pharmacogenomics, Nature, Vol 2, March 2001, pp.  228-231

[10] Altman, Russ B. and Klein & Teri E. Challenges for Biomedical Informatics and Pharmacogenomics, http://www-smi.stanford.edu/pubs/SMI_Reports/SMI-2001-0898.pdf

 

[11] Kleinke, J.D, Oxymorons, The Myth of a U.S. Health Care System, Jossey-Bass, 2001, p. 3

[12] ibid, p. 50

[13] Bernard, Claude, Experimental Medicine, translation by Stewart Wolf, Transaction Publishers, 1999, p. 69, of papers written between 1855 and 1865

[14] ibid, p. 31.

[15] Liebovich, Larry S.  Fractals and Chaos, Simplified for Life Sciences, Oxford University Press, 1998, p. 23

[16] Park, Robert, Voodoo Science, the Road from Foolishness to Fraud, Oxford University Press, 2000, p. 161.

[17] Ibid, p. 43

[18] Gell-Mann, Murray, The Quark and the Jaguar, Adventures in the Simple and the Complex, Freeman and Company, New York,  1994, p. 119

[19] ibid, p. 100

[20] Flake, Gary William, The Computational Beauty of Nature, Computer Explorations of Fractals, Chaos, Complex Systems, and Adaptation, MIT Press, 1999, p. 3

[21] Edelman, Gerald, et al, “Degeneracy and complexity in biological systems”, Proceedings of the National Academy of Science, Nov 20, 2001, vol. 98, no. 24, 13763-3768

[22] “Health and the Devil’s Staircase” http://www.munnecke.com/papers/D13.doc

[23] A. Lindenmayer. Mathematical models for cellular interaction in development, Parts I and II. Journal of Theoretical Biology, 18:280-315, 1968.