Cases & Commentaries

Crushing Chest Pain: A Missed Opportunity

Spotlight Case
Commentary By Mark Graber, MD

Case Objectives

  • Appreciate the challenges of diagnosing
    aortic dissection
  • Describe a Bayesian approach to
    diagnosis
  • Understand the benefits and limitations
    of heuristic thinking
  • List the cardinal dimensions of clinical
    decision-making

Case & Commentary: Part 1

A 62-year-old female presented with 12 hours
of crushing chest pain. Her physical exam revealed a blood pressure
of 140/90, a heart rate of 110, and a respiratory rate of 16. An
electrocardiogram revealed left ventricular hypertrophy with
strain. Review of the chest x-ray in the emergency department (ED)
revealed no abnormalities. The patient was treated for an acute
coronary syndrome (ACS) with heparin, aspirin, morphine, and a
nitroglycerin drip. Cardiac enzymes were drawn. She was admitted to
the cardiac care unit (CCU).

Seven hours after admission, the patient
became hypotensive, with a systolic blood pressure in the 80s and a
heart rate in the 120s. A repeat electrocardiogram revealed no
significant changes. Right-sided leads showed no evidence of right
ventricular infarct. The first set of cardiac enzymes was
equivocal, and a CPK-MB was minimally elevated.

Chest pain is a common complaint in the ED,
increasingly so as patients heed advice to find the closest
hospital for evaluation. Correct diagnosis is critical in this
setting: the patient’s survival may hinge on making a timely
and accurate diagnosis.

In a 62-year-old female with crushing chest pain,
most physicians would choose an acute coronary syndrome as the most
likely diagnosis, as in this case. They would arrive at this
diagnosis knowing that coronary artery disease is common in the
aging population and that unrelieved "crushing" pain often
indicates myocardial ischemia. In the lingo of medical
decision-making, this diagnosis emerged subconsciously from the
'availability' or the 'representativeness' heuristics.(1) Availability
implies that the diagnosis springs to mind, likely because ED
physicians often see patients whose chest pain is due to ACS.
Representativeness implies a mental match between the patient's
symptoms and the characteristic symptoms of ACS stored in the
clinician's memory. In the absence of expert skills, heuristics are
remarkably effective in helping us reach a correct diagnosis
rapidly, accurately, and with little conscious effort.
Unfortunately, they also lead to occasional diagnostic errors, for
example, when the correct diagnosis is not considered.

Ideally, the ED clinicians would have used a
Bayesian
approach, such as 'expected value' decision-making, to derive
the most probable diagnosis.(2) A Bayesian
approach cannot guarantee success, but in theory has the highest
likelihood of selecting the correct diagnosis. This approach begins
by listing all the diagnostic possibilities and the likelihood of
each. In one study, the most common cause of chest or back pain was
acute coronary syndrome, present in 24.4% (Table
1
).(3) The second
step in the Bayesian approach is to adjust the initial
probabilities using Bayesian calculations to incorporate any
information uncovered in data gathering.(2) This
patient described her chest pain as "crushing." Although this
description has long been considered a hallmark of myocardial
infarction (MI), studies suggest it is a weak predictor, with a
positive likelihood ratio less than 2.(4) In
addition, this patient lacked physical findings that increase the
likelihood of MI (diaphoresis, a 3rd heart sound, hypotension, or
rales). Finally, her ECG had no features suggestive of MI, a
feature that modifies the possibility of ACS by a likelihood ratio
of 0.1 to 0.3. Combining these likelihood ratios with the pre-test
probability of 24.4% yields an overall likelihood of cardiac
ischemia of less than 17%. These calculations can be performed
simply by using the 'odds' form of Bayes theorem (Table 2) or a
simple nomogram.

An alternative approach is to use an algorithm
that simulates expert thinking. Many such aides are available, and
these improve the sensitivity and specificity of diagnosing cardiac
ischemia compared with 'clinical judgment.'(4-8)
For example, one study used a formula based on seven clinical
variables to predict cardiac ischemia,(8) and
another derived a prediction rule using four clinical variables:
past history of MI, presence of diaphoresis with chest pain, ST
elevation, and the presence of a Q wave.(5) By
these formulae, this patient's likelihood of having a myocardial
infarction is less than 7%, or less than 2%, respectively.

The ED team correctly decided that the patient
would benefit from observation in the CCU. Their initial diagnosis
of ACS was reasonable in view of the history, and the fact that
this diagnosis, in terms of base rates (Table
1
), is orders of magnitude more likely than alternatives, such
as acute aortic dissection (AD) (80 times less likely) or pulmonary
embolism (60 times less likely). However, had the team incorporated
all available data, they might have realized that the likelihood of
cardiac ischemia was substantially less than initially assumed.
This might have prompted a search for an alternative diagnosis.

Case & Commentary: Part 2

The team re-reviewed the chest x-ray and
discovered an abnormality in the aorta: a 1-cm separation between
the intimal calcification and the adventitial outline of the
descending aorta (Figure
1
).

The mortality rate of undetected AD approaches 1%
per hour, and the diagnosis is missed in 25%-50% of
patients.(9,10) Death
from a missed AD is preventable, because AD can be easily and
definitively identified by appropriate imaging (CT scan, MRI, or
trans-esophageal echography), and survival exceeds 90% with prompt
diagnosis and management.(11) However,
the challenge is to consider the diagnosis in the first place.

The classical presentation is chest or back pain
of acute onset, severe from the outset, tearing or ripping in
quality.(10,12)
Unfortunately, atypical presentations are common, because symptoms
will vary depending on the exact anatomic location of the
dissection and the secondary vasculature involved. Pain may not be
present at all; in one series, 15% of patients with AD reported no
pain.(9) Physical
findings that may support the presence of a dissection include
pulse deficits (greater than 20 mm differential), new aortic
regurgitation, signs of pericardial tamponade, and focal
neurological deficits (Table 3). The
majority of patients, however, will have no specific physical
findings.

In 90% of patients with AD, the admission chest
x-ray is abnormal.(10) Abnormal
aortic contour (sensitivity = 71%) and a widened mediastinum
(sensitivity = 64%) are the most common findings. As in this case,
the x-ray may show the 'calcium sign'—a separation by greater
than 1 cm between the intimal calcification and the outer
adventitial border. Problems degrading the test characteristics of
the chest x-ray in this setting include vagaries in the test itself
and problems with interpretation. For example, the apparent width
of the mediastinum can be increased by a poor inspiration or supine
positioning of the patient. Substantial inter-observer variability
is encountered in the reading of chest x-rays.(13)
In summary, a normal chest x-ray should not be used to exclude the
possibility of AD.

A reliable blood test would be ideal to simplify
the diagnosis of AD. Currently the use of D-dimer is being
evaluated for this purpose. A recent report found a 100%
sensitivity for D-dimer in detecting AD in a small series of
patients.(15) Although
this test lacks specificity (it may also be elevated in pulmonary
embolism, cancer, DIC, sepsis, etc.), a negative result may help
physicians reliably rule out AD. The specificity problem may be
solved by measuring myosin heavy chain, released from vascular
smooth muscle.(16) However,
the utility of either test in the evaluation of AD has yet to be
validated.

Case & Commentary: Part 3

A transesophageal echocardiogram revealed an
ascending aortic dissection (Figure 2).
Anticoagulation therapy was discontinued, beta-blocker therapy was
initiated, and cardiothoracic surgery was called. The patient was
transported to the operating room. Upon arrival in the operating
room, the patient became progressively hypotensive, coded, and
died. Post-mortem autopsy revealed hemorrhage into the
pericardium.

What led to the fatal diagnostic error in this
case? Delays in diagnosis of AD have been associated with
incomplete historical questioning and atypical
presentations.(14) This
patient's death is the result of errors in each of the cardinal
dimensions of clinical decision-making: data gathering, hypothesis
generation (synthesis), and verification.

Data Gathering: The most critical error in
this case was interpreting the chest x-ray as normal. This is a
cognitive error, which may be knowledge-based (if staff were never
trained to recognize abnormalities of the aorta) or skill-based (if
staff were knowledgeable of the x-ray changes of AD but did not
appreciate these on the films). In addition, the misread chest
x-ray reflects a widespread "systems" flaw, where ED staff read
chest x-rays instead of expert radiologists. This practice
contributes to an untold number of diagnostic errors in medicine
and illustrates the challenge medicine will face in reducing
diagnostic error in the future. The error can be prevented, but at
the substantial cost of having expertise available when needed.
Teleradiology may be a solution for some institutions, but
credentialing radiologists in another city, state, or country, has
yet to be standardized.(17)

Additional errors in data gathering may have
contributed if the history was incomplete. Rosman and colleagues
found that clinicians correctly identified AD in more than 90% of
patients if the history included three essential questions
regarding the chest pain: the quality of the chest pain, its
severity, and its location.(18) In
contrast, the diagnosis was correct in less than half if all three
questions were not asked.

Synthesis: The ED physicians assigned a
diagnosis of ACS in this patient, without seriously excluding other
possibilities. If the initial presentation does not "trigger" the
correct diagnosis, it is unlikely to ever be considered.(19,20) The
Bayesian approach would guarantee that the diagnosis of AD would
have been considered.

Verification: Once a diagnosis is reached,
clinicians have a tendency to stop thinking. An existing diagnosis
has almost infinite inertia. This phenomenon of "premature closure"
is possibly the most common cognitive error in internal
medicine.(21) The CCU
team in this case made another cognitive error when they accepted
the ED diagnosis without re-examining the facts and independently
re-thinking the case. Related cognitive biases include "framing"
(we are overly biased by the way in which a case is presented; we
tend to blindly accept a previous diagnosis established by others,
or even ourselves), and "anchoring" (fixating inappropriately on an
early diagnosis). To their credit, the CCU team eventually did
re-think the case when the patient’s condition changed. If a
patient’s course takes an unexpected turn, it is important to
re-think initial assumptions and consider recruiting a colleague or
consultant to evaluate the case anew.

Cognitive errors are all too common in medical
decision-making. Although they can never be eliminated, they can be
reduced by learning optimal decision-making strategies,
understanding the intrinsic biases of using heuristics, and
improving metacognition (the ability to monitor the accuracy of our
own thought processes).(22-24) For
example, the tendency to premature closure can be offset by
conscious efforts to keep an open mind. Clinicians should routinely
generate a complete differential diagnosis in every case. Consider
applying the "crystal ball experience"—after reaching a
diagnosis, pretend you can look into the future with your crystal
ball and see that your initial diagnosis is wrong. What
alternatives should be considered?

Take-Home Points

  • "The first priority in
    differential diagnosis is to think about the diseases you can't
    afford to miss."(2) It is always
    appropriate in ED settings to rule out the "worst case
    scenario."
  • Keys
    to detecting acute aortic dissection are a complete history
    (identifying the quality, severity, and location of the chest pain)
    and attention to the chest x-ray.
  • Always strive to
    achieve expert skills through learning. Non-experts can supplement
    diagnostic skills using expected value decision-making or
    established algorithms.
  • It is often wise to consider tests that might help
    "rule in" an alternative diagnosis than to pursue tests for a
    diagnosis already in doubt. Be prepared to re-think a case if an
    inconsistent finding or unexpected turn of events
    arises.
  • Learn to use
    cognitive forcing strategies to reduce diagnostic error (23); understand
    the common cognitive biases and use metacognition to avoid these
    traps. Keep an open mind.

Mark Graber, MD
Chief, Medical Service, VA Medical Center, Northport, NY
Professor and Vice-Chair, Department of Medicine, SUNY Stony
Brook

Faculty
Disclosure: Dr. Graber has declared that neither he, nor any
immediate member of his family, has a financial arrangement or
other relationship with the manufacturers of any commercial
products discussed in this continuing medical education activity.
In addition, his commentary does not include information regarding
investigational or off-label use of pharmaceutical products or
medical devices.

References

1. Elstein AS.
Heuristics and biases: selected errors in clinical reasoning. Acad
Med. 1999;74:791-4.[ go to PubMed
]

2. Sox HC Jr, Blatt
MA, Higgins MC, Marton KI. Medical decision making. Stoneham, MA:
Butterworth-Heinemann; 1988.

3. von Kodolitsch Y,
Schwartz AG, Nienaber CA. Clinical prediction of acute aortic
dissection. Arch Intern Med. 2000;160:2977-82.[ go to PubMed
]

4. Panju AA,
Hemmelgarn BR, Guyatt GH, Simel DL. The rational clinical
examination. Is this patient having a myocardial infarction? JAMA.
1998;280:1256-63.[ go to PubMed
]

5. Tierney WM, Roth
BJ, Psaty B. Predictors of myocardial infarction in emergency room
patients. Crit Care Med. 1985;13:526-31.[ go to PubMed
]

6. Diamond GA,
Forrester JS. Analysis of probability as an aid in the clinical
diagnosis of coronary-artery disease. N Engl J Med.
1979;300:1350-58.[ go to PubMed
]

7. Goldman L, Cook E,
Brand DA. A computer protocol to predict myocardial infarction in
emergency department patients with chest pain. N Engl J Med.
1988;318:797-803.[ go to PubMed
]

8. Pozen MW,
D'Agostino RB, Selker HP, Sytkowki PA, Hood WB Jr. A predictive
instrument to improve coronary-care-unit admission practices in
acute ischemic heart disease. A prospective multicenter clinical
trial. N Engl J Med. 1984;310:1273-8.[ go to PubMed
]

9. Spittell PC,
Spittell JA Jr, Joyce JW, et al. Clinical features and differential
diagnosis of aortic dissection: experience with 226 cases (1980
through 1990). Mayo Clin Proc. 1993;68:642-51.[ go to PubMed
]

10. Klompas M. Does
this patient have an acute thoracic aortic dissection? JAMA.
2002;287:2262-72.[ go to PubMed
]

11. Nienaber CA, von
Kodolitsch Y, Nicolas V. The diagnosis of thoracic aortic aneurysm
by noninvasive imaging procedures. N Engl J Med. 1993;328:1-9.[ go to PubMed
]

12. Hagan PG,
Nienaber CA, Isselbacher EM, et al. The International Registry of
Acute Aortic Dissection (IRAD): new insights into an old disease.
JAMA. 2000;283:897-903.[ go to PubMed
]

13. Gregorio MD,
Baumgartner FJ, Omari BO. The presenting chest roentgenogram in
acute type A aortic dissection: a multidisciplinary study. Am Surg.
2002;68:6-10.[ go to PubMed
]

14. Alsous F, Islan
A, Ezeldin A, Zarich S. Potential pitfalls in the diagnosis of
aortic dissection. Conn Med. 2003;67:131-4.[ go to PubMed
]

15. Weber T, Hogler
S, Auer J, et al. D-dimer in acute aortic dissection. Chest.
2003;123:1375-8.[ go to PubMed
]

16. Suzuki T, Katoh
H, Nagai R. Biochemical diagnosis of aortic dissection: from bench
to bedside. Jpn Heart J. 1999;40:527-34.[ go to PubMed
]

17. Pollack A. Who's
reading your x-ray? New York Times. Nov 16, 2003: BU 1, BU
9.

18. Rosman HS, Patel
S, Borzak S, Paone G, Retter K. Quality of history taking in
patients with aortic dissection. Chest. 1998;114:793-5.[ go to PubMed
]

19. Elstein AS.
Clinical reasoning in medicine. In: Higgs JM, Jones M, eds.
Clinical reasoning in the health professions. Oxford, England:
Butterworth-Heinemann Ltd; 1995:
49-59.

20. Kassirer JP,
Kopelman RI. Cognitive errors in diagnosis: instantiation,
classification, and consequences. Am J Med. 1989;86:433-41.[ go to PubMed
]

21. Graber M,
Franklin N, Gordon R. Diagnostic error in internal medicine. Paper
presented at: 24th Annual Meeting of the Society for Medical
Decision Making. October 20, 2002; Baltimore, MD.
2002.

22. Croskerry P.
Achieving quality in clinical decision making: cognitive strategies
and detection of bias. Acad Emerg Med. 2002;9:1184-204.[ go to PubMed
]

23. Croskerry P.
Cognitive forcing strategies in clinical decision making. Ann Emerg
Med. 2003;41:110-120.[ go to PubMed
]

24. Graber M, Gordon
Ruthanna, Franklin N. Reducing diagnostic error in medicine: what's
the goal? Acad Med. 2002;77:981-92.[ go to PubMed
]

Tables

Table 1. Diagnosis
of ED Patients with Chest Pain (3)

ED Diagnosis

Percentage (%)

Acute coronary syndromes

24.4

Neuro-radicular pain

17.1

Pulmonary disease

15

Cardiac arrhythmia

6

Vasovagal

6.5

Congestive heart failure

5

Hyperventilation

3.3

Hypertensive crisis

2.2

Chest wall syndrome

1.8

Gastrointestinal disease

1.2

Pneumothorax

0.6

Pulmonary embolism

0.4

Aortic dissection

0.3

Pleuritis

0.2

Pericarditis

0.1

No clear diagnosis

5.9

Other

10.1

Table 2. Using
Bayes Theorem to Adjust Pre-test Probabilities for New Data
(2)

Step

Process

Specific Calculations

1

Identify a likely diagnostic
possibility.

Does this patient have an acute coronary
syndrome?

2

Estimate the pre-test probability of this
disease using known prevalence rates, a clinical prediction rule,
or your local experience.

Pre-test probability = 24.4%;
Probability = 24.4% = Odds/(1+odds)
Odds = approximately 1/3 (one chance in three; one person will have
an acute coronary syndrome for every 3 that have some other
explanation).

3

Gather additional data or perform a test
that will help support or refute your initial hypothesis.

Crushing chest pain: Likelihood ratio

4

Uses Bayes' theorem to adjust your
pre-test probability, incorporating this new information. The
post-test probability is calculated most easily using published
nomograms. Alternatively, the post-test probability can be
calculated by multiplying the pre-test odds of disease by the
relevant likelihood ratios.

Post-test odds = Pre-test odds x
Likelihood ratio
Post-test odds = 1/3 X (

5

Keep gathering new data until you have
reached your threshold for going to the next step, or dropping the
disease from consideration.

At some point, the likelihood of an acute
coronary syndrome becomes low enough that other diagnostic
possibilities are more seriously considered.

6

Repeat this process for any other
diagnostic possibilities.

 

Table 3.
Sensitivity of the Physical Exam and Chest X-ray in the Diagnosis
of Acute Thoracic Aortic Dissection (10)

Finding

Sensitivity (%)

95 CI* (%)

Elevated blood pressure

49

41-57

Diastolic murmur

28

21-36

Pulse deficit

31

24-39

Pericardial rub

6

3-13

Congestive heart failure

15

4-33

Focal neurological deficit

17

12-23

Shock

19

15-26

New MI by ECG

7

4-14

Chest x-ray findings

 

 

   Abnormal aortic
contour

71

56-84

   Wide mediastinum

64

44-80

   Pleural effusion

16

12-21

   Displaced intimal
calcification

9

6-13

*CI indicates
confidence interval.

Figures

Figure 1. Chest
X-ray with Calcium Sign (arrow).


Figure 2.
Transesophageal Echocardiography of Aortic Dissection. Short Axis
View of Proximal Aorta.