THE GLAUCOMA RECORD AND THE COMPUTER
(Notes for a Lecture)
I have been asked to say a few words about the medical
record of the glaucoma patient. What should such a record
contain? How should it be organized? How are its content
and organization related to computers and computing
techniques? What can you expect of computerized data banks
about glaucoma? What should one expect from computerized
decision-making?
The answers to some of these questions will become
clear if one takes a closer look at the medical record of
the glaucoma patient. The glaucoma record is the site where
the shortcomings of our work are documented, and it is the
point where plans for betterment must start.
Glaucoma, whatever acute manifestations it may have, is
a chronic disease, and I know of no clinical situation where
one may properly say to a glaucoma patient that no further
thought need be given to his glaucoma problem because it has
been conclusively resolved. Thus whenever you are
confronted with a patient who has, or is suspected of
having, glaucoma, you must think of a process beginning more
likely in the distant than in the recent past, a process
which extends far into the future, literally to the end of
the patient's life.
We understand and treat glaucoma as a progressive disease.
Therefore the interpretation of any given examination depends on
the findings of past examinations. And since you will not be
able to remember them, even if you were at all times the pa-
tient's ophthalmologist, you depend on the record, your own or
that of one of your colleagues for the details. Glaucoma is a
disease which usually extends over many years, and the record of
the glaucoma patient is likely to be correspondingly voluminous.
A reasonably complete understanding of the patient's problem re-
quires the interpretation of that record. This is no simple
task, but in my experience, is is a very rewarding one; and I can
think of few better ways to learn about the multifaceted behavior
of glaucoma than to sit down with a difficult patient's record,
for an hour or two, and reconstruct what the physicians who were
taking care of the patient thought was going on and what they
thought that they were doing.
What then should the glaucoma record contain. Perhaps most
important: it should contain what comes to your mind as unique,
surprising, unusual, and hence worth recording: and that, para-
doxically, is something the computer can never understand. The
record which is most valuable, will contain your thoughts, your
uncertainties, your deliberations. In a more mundane perspec-
tive, it should contain, in addition to the date, also the time
of the examination. It should include the time, because of the
frequency with which diurnal variations in the pressure occur,
and the obvious desirability to correlating pressure with the
time of day in which it was measured. It should contain the med-
ication which the patient is receiving for his glaucoma, the
dosage, the most recent occasion of administration, and perhaps
also the patient's words about side effects and about the regu-
larity with which he takes it. It should contain the applanation
pressure. It should contain the corrected visual acuity, and the
patient's refractive error as well. It should contain a descrip-
tion of the angle, of the depth and clarity of the anterior cham-
ber, of the transparency of cornea, lens and vitreous, and it
should contain also a description of the optic disc. It should
also contain a description of the visual fields. All these pa-
rameters may be expected to vary from examination to examination
and it is the function of the record to make these data accessi-
ble to the physician at the time that he examines and advises his
patient. Since it is too time-consuming to review the patient's
record and to reconstruct his medical history, and since, in any
event, such reconstruction could become very complex, the obvious
solution might seem to register and store the data on electronic
computers, and then to program the computer to produce various
statistical summaries and graphic displays which would presumably
constitute a much more solid and reliable foundation for clinical
judgment than the haphazard reviews of the handwritten chart by
means of which we customarily acquaint ourselves with the pa-
tient's problem.
From time to time such programs are proposed, but I am not
aware of any that has proved a practical success. If it seems
worthwhile discussing this issue nonetheless, that is for two
reason, the first being, that in the coming years you will un-
doubtedly be plied with offers to purchase various computers and
their operating systems, and it will be helpful to you to have
some perspective on the opportunities and problems of computer
assisted record keeping, and it might, incidentally, also save
you a lot of money. The second reason, which is less practical,
but ultimately probably more cogent, is that in the attempt to
design computer assisted records and computer assisted decision
making systems, one encounters certain very fundamental issues
that bear on the creating and interpretation of the glaucoma
record.
Most of our clinical descriptions are approximations. The
computerized record is characterized by the circumstance that it
is inherently mathematical: both its difficulty and its great po-
tential value stem from the cirmcumstance that mathematics does
not admit of the same uncertainty, does not tolerate the same am-
biguities that other forms of communication can tolerate. This
demand for exactness limits its use but makes it more powerful.
It also explains why thus far the computer has been of use in
medical practice in keeping financial accounts, this being the
only area in the physician's practice where ambiguity is not tol-
erated.
The computer, in other words, forces us to be precise in our
descriptions, the computer makes fun of us if we are not. And if
then, mindful of the requirements which the computer imposes on
us, we return to our practices, we find that it is impossible for
us to be precise, that with all the imprecision of our descrip-
tions, we do very well, and that ill-advised efforts at precision
may become very real obstacles to the performance of our work.
With this introduction, let me then turn to discuss three
uncertainties with which we deal when we examine a glaucoma pa-
tient and translate our findings into a written record: namely,
the visual field, the appearance of the anterior chamber angle,
and the appearence of the disc.
Progression of visual field loss in a glaucoma patient who
is receiving maximum tolerated medical therapy is the single most
important criterion by which the need for surgery is determined.
The field test is subjective. The patient's responses are not
absolutely consistent and the isopter that is so sharply demar-
cated on the plot of the visual field actually represents a fre-
quency distribution that must be defined in statistical terms.
The lines that are customarily inscribed on the field chart are
but intuitive approximations of such statistical calculations.
The glaucoma record usually contains a series of visual field
plots. It is a simple matter to enter such plots as their carte-
sian coordinates into a computer either by digitizing the televi-
sion image or by tracing its outline on a digitizing pad. It
would be simple then to write computer programs to calculate the
reliability of any given isopter, to determine the absolute area
of the field, to calculate the rate of change of the total area
or of any specific segment of the field. Thus one could state
quickly and accurately whether the field was stationary or
shrinking. Such quantitative analyses of the visual field, how-
ever, are not undertaken, and the reason they are not undertaken
is that one can see, simply by inspecting a series of field
charts, that they are not sufficiently accurate to make mathemat-
ical interpretation meaningful. And yet, inspite of its obvious
inaccuracy, we find the visual field examination very useful.
This is something to think about.
What makes the visual field so readily susceptible to mathe-
matical analysis, specifically, to computer analysis, is that the
plotting of the field, as we customary do, already represents a
coding of data, and thus the most important and difficult part of
the interpretation of the medical record will already have been
done: Coding is the translation of qualitative observations, in
this case, the depression of the sensitivity of the visual field,
into quantitative terms. The visual acuity, the intraocular
pressure, the refractive error, and the prescription of specific
medications, to be used at specific hours, are also already in
code, and makes these data eminently suitable for computer stor-
age and analysis.
The same cannot be said about the description of the anteri-
or chamber angle. We use gonioscopy to view the trabecular mesh-
work, to identify normal and abnormal pigmentation, normal and
abnormal vascularization, inflammatory deposits, and angle reces-
sion and peripheral anterior synechia. But perhaps the most com-
mon, and possible overall the most important use of the gonio-
scope is to determine the width of the angle in order to ascer-
tain whether or not it has closed or might be about to close.
Angles vary greatly in width, and it is quite important to be
able to say for any given eye whether or not the angle was so
wide that it could not possibly close as the pupil dilated spon-
taneously or whether it was already at the point of closure. And
one would wish not only to describe these two extreme situa-
tions, but one would also wish to describe various intermediate
conditions, such as, for example, that the angle was definitely
open but so narrow that it might be expected to close sponta-
neously, or that it was unlikely to close spontaneously but might
close with mydriasis. Obviously, it is difficult to be confident
of such judgments. It would be much better if one could measure
the angle and enter its width in degrees or radians into the
record. There are no instruments to make such measurements. But
consider the geometric problem of defining an angle whose sides,
the iris below and the meshwork above are not plane but curved
surfaces. Finally remember that an estimate of angle width in
only one meridian is of relatively little value, for what we need
to know is what the angle looks like around all twelve hours of
the clock. There are preprinted diagrams on which one may chart
the appearance of the angle. Such diagrams are, however limited
in usefulness in that they were designed, if I understsnd them
correctly, for recording the extent of actual synechial closure.
They were not designed for recording the narrowness - or width of
an angle that was still open. Verbal descriptions of the appear-
ance of the angle are fine if one has time to read through the
chart to find them, think about them, interpret them, and compare
them. But such free text descriptions are incomprehensible to
the computer, and if we wish to rely on the computer for
analysing and retrieving our findings for us, we must encode
them, i.e., we must translate them into unambiguous mathematical
formulas. Let me describe for you an improvised coding system
for angle width that I myself used for some years and that I
found rather useful.
I undertook to examine all angle that I gonioscoped at eight
principal points on the face of the clock, specifically at 12,
1:30, 3:00, 4:30, 6:00, 7:30, 9:00 and 10:30 o'clock. Thus for
each angle that I gonioscoped, I made eight observations suitable
for coding. At each of these eight points, I asked myself, how
wide is this angle, and I classified what I saw into six cate-
gories: If I considered the angle definitely closed, I assigned
the point to class 0. If the angle was open but slit like and at
the point of closure, I assigned it to class 1. If the angle was
narrow so that I could imagine with mydriasis to five millimeters
it might close, I assigned it to class 2 and considered it at
risk for spontaneous closure. If the angle was narrow so that
full mydriasis might close it, I considered the angle subject to
mydriatic but perhaps not to spontaneous closure, and assigned it
to class 3. If the angle was slightly narrower than average but
unlikely to close at that point even with mydriasis, I assigned
it to class 4. Angles of normal or greater than normal width, I
called class 5, and angles that I considered pathologically re-
cessed, I designated as class 9. I could then record the de-
scription of the angle simply by listing eight digits, which I
grouped into two sets of four digits each, and interpreted as de-
scribing the angle clockwise, beginning at twelve o'clock. Thus
the value 2234-2110 would describe an angle which I thought was
closed at 10:30 o'clock, was slitlike and at the point of closure
at 7:30 and 9:00 o'clock, might close with slight mydriasis at
12, 1:30 and 6:00 o'clock and was relatively wider at 3 and 4:30.
It goes without saying that I could not test or confirm such de-
scriptions empirically, but I was able to test myself by serial
calibration of the same angle on successive examination over pe-
riods of months and years, and I found my descriptions to be suf-
ficiently consistent that the effort they required seemed well
worthwhile. Calibration and coding of this sort is made to order
for computer storage and analysis, and I was, in general, rather
satisfied with it.
I wish I could say the same for the descriptions of the op-
tic disc which I attempted, proceeding somewhat in the same man-
ner. I selected the same eight points of the clock face as for
the angle and I attempted to estimate for each of these positions
the proportion of the radius of the disc that had not been dis-
solved in the central cup. Thus where the disc was totally
cupped, I would assign a value of 0, and where there was no cup-
ping at all, I assigned a value of 9, If the thickness of the rim
was 1/10 of the radius, the assigned value would be 1, if it was
3/10, the assigned value would be 3. I grouped the coded values
in groups of four, just as I had done for the angle. Thus a disc
cupped to the rim at 3 and 4:30 o'clock might be described as
2100-2332. While the absence of excavation and total cupping
seemed easy enough to identify, the intermediate values gave me
much trouble; and this for two reasons: In the first place, there
were the saucerized cups, where the excavation sloped gradually
up to an intact rim. These discs were definitely abnormal, but I
could not calibrate the edge or the severity of the excavation.
The second source of confusion was the color of the remaining rim
which, difficult enough to estimate, had no place in the scheme
that I had devised. Thus a thin sharp rim with excellent capil-
lary perfusion and a bright pink color would most likely be much
less advanced a glaucomatous change, - might even be normal, than
a broader, pale and atrophic rim. Finally, and not least in im-
portance, I found the outer boundary of the disc often very dif-
ficult to identify, and although there were many instances where
one might see that the excavation had extended significantly to-
ward the disc periphery, one would yet be at a loss to say exact-
ly or to estimate what proportion of the disc remained unaffect-
ed, or, in other words, how wide the remaining rim in fact was.
I was unable to devise any approach to calibrating the undermin-
ing of the disc margin. I do not think that my efforts were en-
tirely useless, but I found them so unreproduceable, that I de-
cided to suspend this effort until I had devised some better
scheme.
The foregoing examples make clear that the task of coding
the information that goes into the glaucoma record is formidable
and is far from solved. Notice also that I have not even begun
to mention descriptions of the cornea, of the depth or configura-
tion of the anterior chamber, or of the presence in it of white
cells, red cells or protein or pigment, of atrophy of the iris
and of its transillumination, of the various distributions of
cataract, of pigmentation, inflammatory debris, neovasculariza-
tion or pigmentation of the meshwork. The vitreous remains unde-
scribed and uncoded, as does the retina, choroid, and the retinal
vasculature. I mention these only to enumerate the substantial
problems that stand in the way of even a moderately complete en-
coding of the medical record.
Two questions present themselves. The first of these ques-
tions is: Is coding of the glaucoma record of any benefit? Yes,
for purposes of research. If it does nothing else, the attempt
to be specific in your descriptions of what you see and conclude
from your examinations will force you to reexamine, to be criti-
cal, to consider the significance of what you see.
The second, and even more important question is: If the en-
coding of our clinical data is so very difficult and has largely
eluded our efforts, if we cannot be definite about our findings,
how are we able to diagnose and treat glaucoma at all? Since our
knowledge does not satisfy the criteria of the biostatisticians,
how do we discharge our duty of taking care of our patients? We
are, after all, confronted with patients who have immediate
pressing problems that cannot wait for the implementation of as
yet unwritten computer algorithms. Given the approximations of
which our medical record consists, how can we fulfill this task
at all? Some reflections on this paradox in a subsequent issue.
* * * * * * *
I think that we can, and the reason, paradoxical as it may
seem, why we are able to proceed, and in many instances make the
correct decision, and help our patients, inspite of the fact that
the visual fields are inaccurate and unreliable, inspite of the
fact that we cannot, in many cases, reproduceably describe the
optic disc; inspite of the fact that often we cannot be sure
whether an angle is closed or merely in apparent apposition; the
reason why we are able to proceed in the face of such uncertainty
is that the hypothesis upon which the mathematical approach to
knowledge rests, the hypothesis that our understanding is built
up of individual separable building blocks, this hypothesis is
probably invalid.
Let me give you an example of what I mean. Take a photo-
graph of a disc, study it, and you are likely to form some very
good idea about whether or not the disc is normal or whether it
has been severely damaged by glaucoma. The inductive, mathemati-
cal theory of knowledge would hold that your interpretation of
that disc is the summation of your interpretation of discrete
parts of the disc. The computer is expected to simulate that
summation by a mathematical function, integrating the value of
each separate item into the whole. But what if in reality no
such summation occurs? To demonstrate that this theory is wrong,
you need merely take a pair of scissors, and cut the photograph
of the disc, let us say into fifty small fragments, labeled if
you will, so that you know precisely which part of the disc each
segment represents. Try as you will, you will be unable to in-
terpret each isolated segment by itself. Your interpretation and
understanding, therefore, is of the entire structure, and your
references to its appearance at 12:00 or 4:30 o'clock are possi-
ble only in consequence of your prior interpretation of the
whole. This process of intuitive apperception of the configura-
tion applies not only to the disc; it applies to the angle, to
the lens, to the field; it applies to the historical sequence of
events. Thus there appears a compelling explanation for the fact
that some of the most competent of clinicians will have no truck
with the computer business; they distrust it instinctively, and
to a large extent they are correct.
It is foolish, and you can now see why, to expect that the
computer should replace this intutive judgment; but it is not at
all foolish or improbable that the computer will supplement it;
and supplement it in ways that we cannot now predict. The com-
puter is a new instrument, like a telescope or a microscope if
you will, and what it shows us is qualitatively entirely differ-
ent from what we see with the unaided eye or what we think with
the unaided mind. If the intuitive approximations on which we
now rely were perfect and infallible, we could dispense with data
processing technology. But they are not, and we need all the
help we can get.
* * * * *
Back
Next
Glaucoma Letter Index
Website Index
Copyright 2006, Ernst Jochen Meyer