EVERY DECISION WE MAKE IN CLINICAL DENTISTRY INVOLVES SOME DEGREE OF UNCERTAINTY WHETHER
WE RECOGNIZE IT OR NOT. MOST PATIENTS DON'T APPRECIATE THE UNCERTAINTY WE FACE AND OFTEN
DEMAND DOCTORS WHO ARE CONFIDENT AND CERTAIN, WHO ALWAYS HAVE THE CORRECT ANSWERS.
It is important to understand that many errors in judgment
are a result of our human thinking processes—processes that are
common to all rational beings. We must also recognize and embrace
the inevitable uncertainties in clinical decision-making by
applying principles of probability in ways most of us were never
exposed to in school. Finally, we need to understand the nature of
endodontic diseases and why our endodontic tests often fail us.
Part 1: Judgment under uncertainty
In the early seventies, many of the weaknesses of the human
machinery used for making judgments in the face of uncertainty
were delineated by Amos Tversky and Daniel Kahneman. They
explained that people tend to utilize "heuristics," which are rules
of thumb, and these heuristics can lead us to biased decisions.
Heuristics are built-in, experience-based shortcuts we all rely
upon to quickly process information.
However, because heuristics are so commonly used (subconsciously),
they easily convince our brain that the conclusions they
point us towards are automatically valid, and this is certainly not
true.
Some well-known heuristics have been described in the cognitive
science literature:
Anchoring Heuristic
For example, a patient reports of pain on the upper right with
tooth #3 displaying a large area of coronal destruction due to caries
may influence how we interpret a periapical radiograph of #3. As a
specialist, I often get referrals where the patient or referral slip reports
a periapical radiolucency ("my dentist says there's a ‘dark spot'
on the X-ray") when none of my 2D or 3D images corroborate this.
In these cases, the clinician probably allowed early information to
"anchor" his/her thought process, leading to the overly concrete
and specific visualization of a radiolucency on the imaging.
This is especially problematic for younger and less experienced
clinicians. When I was in my residency, I would frequently
discuss cases with undergraduate dental students who would tend
to "see" radiolucencies on periapical radiography when I could appreciate
no such radiographic feature. By way of clinical advice, on
cases where you suspect subtle radiographic perturbations, scan
other areas of the image and see if similar areas don't show up on
adjacent teeth.
Availability Heuristic
We should be careful when we are interpreting the subjective
and objective information we've collected from patients and at
least recognize the availability heuristic. For example, pain on a
tooth with recent restorative work very well could be pain of endodontic
origin that requires endodontic therapy. However, in my
practice, occlusion and overall life stress of the patient are large
factors that I always consider, despite the very plausible explanation
of pulp trauma from dental treatment. Failure to include other
hypotheses in the diagnostic work-up generally leads to people
getting root canals they didn't need.
Representativeness Heuristic
A good example of this is the terrible-looking root canal treatment.
Often we see short root fills, missed canals, weak or incomplete
fi lls and we infer disease from these features even in the
absence of clear clinical findings or patient reports of symptoms.
Combined with anchoring, we might even infer a periapical radiolucency
on the image when in fact none exists.
I had a patient referred to me one year ago with complaint of
pain on #9. Her dentist had recently performed root canal therapy
on that tooth. When I examined her, it was clear her problems were coming from #10. However, she had two
other root canal treatments that were supposedly
causing her pain even though I couldn't elicit any
responses from them with biting or percussion.
It took 5 months of calcium hydroxide therapy
on #10 for her symptoms to subside. Everything
about this patient pointed to her being "a little
crazy." However, I decided to perform a test by
prescribing amitriptyline for one month. All her
symptoms subsided.
Escalation of Commitment
I frequently encounter patients in my practice
who have, in the recent past, invested in a new
full or partial veneer crown, with the tooth now
exhibiting signs of pulpal/periradicular disease.
During my assessment, I might determine that the
structural integrity of the tooth, among other factors,
creates a poor prognosis for the tooth. I may
even recommend extraction of the tooth and the
placement of a titanium implant. Patients often
choose to ignore this advice and move forward
with endodontic therapy "because I've already put
so much into this tooth." The faults of this line of
reasoning are self-evident and center around what
is called "escalation of commitment."
I remember an accounting professor in college
state this bold axiom: "Sunk costs are never
relevant in future decisions." His classic example
involved the decision of whether to continue a relationship
with one's girlfriend/boyfriend or not.
To consider how much time had already been put
into the relationship serves only to confuse the
situation and shouldn't have any weight on the
decision of whether to move forward with the relationship
today.
Part 2: The probability of
diagnostic science
As part of our dental educations, we all took a
course or had a section on biostatistics. We were
taught the basics of frequentist statistics, championed
a century ago by RA Fisher and later by
Neyman, Pearson and others. These calculations
assume that we "observe" large repetitions of scenarios
that we actually don't observe. I remember
my professor saying, "If we ran this experiment
thousands of times . . . " But we didn't run it thousands
of times. This brand of statistics doesn't
necessarily characterize the types of things we, as
clinicians, want to know.
For example, in the real clinical world, I'm not
so interested in the probability of a patient with
pulpitis experiencing cold sensitivity; I'm interested
in knowing the probability that a patient who
reports cold sensitivity has pulpitis and needs interventional
pulp therapy. It is critical to understand
the differences in direction of logic between
these examples. Mathematically we write: P(A|B)
≠ P(B|A). This would read: The probability of
event A given observation B is NOT the same as
the probability of observation B given event A.
To clarify and drive home this point, consider
this simple example: Suppose you see an American
citizen. What is the probability that citizen
is a U.S. Senator? Now, suppose you see a U.S.
Senator. What is the probability that Senator is an
American citizen? Huge difference in the answers.
The type of probability we're interested in as
clinicians is what is known as conditional probability.
A good example of conditional probability can be
shown with this short story.
There was a hit-and-run accident that happened in the late evening hours. The only witness
reports to the police that is was a blue taxicab that
struck a pedestrian. As the law enforcement officials check out her report and test her reliability at
cab color identification, they find that her accuracy
is only about 80 percent. What are the odds that
it was a blue taxi that hit the pedestrian?
We can't actually answer this yet. We need a
key piece of information: are there any other kinds
of taxis in the town? It turns out there are two cab
companies, one green and one blue. The green taxis
make up 85 percent of the taxis in town, blue the
other 15 percent. It's critical to note that this distribution
of taxis in town is a very important part of
the equation. This is known as the base rate.
Clinicians often fail to take into account base
rates. I've frequently had dentists show me radiographs
of maxillary molars and ask if the mesiobuccal
root is cracked. I usually point out that the
"crack" they are "seeing" is actually the superposition
of the periodontal ligament of the palatal
root. Then I talk about the frequency, or base rate,
of cracks of that shape and location, which rate is
close to zero. "That's not how teeth crack."
The English friar, Thomas Bayes, introduced
Bayes' Theorem in the 18th century. It involves
calculating posterior probabilities (the chance
of an event after applying new observations and
data) using prior probabilities and accumulated
evidence. Prior probability is the chances of an
event before it has occurred. What is the likelihood
that the sun will come up tomorrow? It
hasn't happened yet. But the prior information
we have available to us (our life experience and
scientific knowledge of the solar system) indicates
that the sun will rise tomorrow with exceedingly
high probability, almost complete certainty. We
can't have absolute certainty since it still hasn't
happened yet!
I encourage the reader to spend some time on
the Internet researching the principles of Bayes'
Theorem. As you do, you won't be surprised to
learn that in our hit-and-run example above, the
chances of it being a blue taxi, as the eyewitness
stated, aren't 80 percent. In fact, the probability
that it was a blue taxi, utilizing Bayes Theorem
to calculate it, is actually only 41 percent. So this
80 percent accurate witness is more likely to be wrong than right!
When I lecture on this topic, I get a lot of
questions from audience members that sound like
this: "What's this got to do with endodontic diagnosis?"
Go to Dentaltown.com and search for the
thread in the EndoFiles section titled: "So what
does 80 years of Lit have to say for Ms. Smith."
This illustrates beautifully what the real-world
problems are stemming from the failure to understand
conditional probability. School gave us a lot
of answers that served us well during written examinations.
The clinical reality provides endless
examples of where these answers are insufficient.
I love this quote from the late physicist, E.T.
Jaynes: "In scientific inference our job is always
to do the best we can with whatever information we have;
there is no advance guarantee that our information will
be sufficient to lead us to the truth. But many of the
supposed difficulties arise from an inexperienced
user's failure to recognize and use the safety devices
that probability theory as logic always provides.
Unfortunately, the current literature offers no
help here because its viewpoint . . . directs attention
to other things . . . (emphasis mine)."
Part 3: Diseases of endodontic origin
and diagnosis
Endodontology as a science teaches us many
things about the dental and periradicular tissues
that are valuable in our efforts to solve the problems
patients present to us. Now that I have been
practicing clinical endodontics full time, I've
come to realize that this training fails me often.
It's overly simplistic and tries to fit everything
into neat and tidy boxes that simply don't exist in
reality.
Patients often present with complex problems
that require more than the simple tests we were all
taught in school. The body can produce a tremendous
amount of noise that confuses the signals we
as clinicians seek when unraveling complaints of
pain, sensitivity, pressure, and even esoteric reports.
I'm not talking about the straightforward
cases where the patient points to the tooth with
the large cavity and reports temperature sensitivity.
I'm also not talking about the painful tooth with
corresponding unambiguous radiographic lucency.
Those are the easy ones clinically, but they can actually
harm our diagnostic abilities. They mislead us into thinking our diagnostic tests should produce
specific results.
We've all had cases where it seems that none
of our tests are helpful. These are the tough ones;
these are the cases where we must be cautious
to avoid errors in judgment and carefully weigh
probabilities.
I recently saw a retired Navy SEAL who was
referred for endodontic treatment on tooth #2.
Upon interviewing the patient, he confirmed that
he was experiencing pain in the area of #2. Upon
further questioning, he admitted that the pain was
actually vague and he wasn't certain it was #2 but
that his general dentist “confi rmed” it was #2. My
tests were very inconclusive. In my mind, candidates
for disease included #2, #3 and #31. The
posterior probability from my testing was pretty
much even for those three: 33 percent each. So I
decided to run one of my most powerful tests: the
“come back next week” test.
If you take away nothing from this article, at
least consider this: everything is a test. Tapping
on teeth is a test. So is biting on a ToothSlooth, a
wooden stick or a cotton roll (each can be useful
in its own way). Initiating treatment is also a test.
Our pre-operative diagnoses of vitality or necrosis
are really just tentative diagnoses until we open
the tooth and verify.
Legitimate diseases of endodontic origin are
not self-limiting. They may take months or years
to manifest in some cases, but they nearly always
get worse. So waiting—observing—is a test. Now,
when to observe and when to be more aggressive
is usually dictated by risk tolerance of the patient.
An astronaut heading to the ISS for 9 months has
no tolerance for risk. If I'm not certain and if the
patient isn't in a lot of pain, I wait until I have
more information.
So back to the SEAL . . . I had him return in
one week and the symptoms had pointed to #31,
the testing now implicated #31 (percussion and
bite), and initiation of treatment of #31 revealed
a necrotic distal pulp horn, necrotic distal canal,
and hyperemic tissue in the mesial canals. Waiting
was the test that made all the difference.
Summary
You and I have each sat through many lectures
on endodontic diagnosis. It's almost always useless
information because they focus on things that
help us with the cases we don't need help with: the
easy ones. Moving into the big leagues requires a
big-league mentality.
We must understand the pitfalls of our cognitive
machinery if we are to take rational approaches
to diagnostics. We also have to spend
some time crawling the Internet for information
relative to probability theory, specifically conditional
probability. Conditional probability is not
a topic that comes naturally to most people. The
whole point of a differential diagnosis is to enumerate
possibilities with associated probabilities.
Finally, we need to realize that the tests we use
in endodontic diagnosis aren't always that good at
discriminating, especially for subtle disease. Applying
the give it time test can help avoid costly and
confusing problems.
|
Dr. Scott Weed received his DDS from the University of the Pacific in San Francisco, California. Upon graduation from dental school, Dr. Weed was commissioned
a lieutenant in the Navy and completed an Advanced Education in General Dentistry at the United States Naval Dental Center, Okinawa, Japan. He then
practiced general dentistry for two years with his father, Dr. Robert F. Weed, in Fallon, Nevada. Dr. Weed left general practice and then completed a two-year
specialty program in endodontics at the University of Southern California in Los Angeles, California. It was in Los Angeles that Dr. Weed conceived of the Reno
Endodontic Continuum.
|
|