Clinical decision making is the process by which we determine who needs what, when. While not exactly arbitrary, this exercise can be quite subjective. Each clinician compiles their own data (hence the emphasis on learning to perform an accurate H&P) and then constructs an argument for a particular disease state based on their interpretation of the “facts.” The strength of their case will depend on the way in which they gather and assemble information. There may then be no single, right way of applying diagnostic and therapeutic strategies to a particular case. Of course, not every situation is a clinical quandary. A patient with a known history of coronary artery disease presenting with 3 hours of crushing chest pain, an EKG with 4mm ST segment elevations across the precordial leads, and an initial CK of 3000 is having a myocardial infarction. That is a diagnostic slam dunk. More commonly, however, there exist elements of uncertainty. Medicine involves playing the odds, assessing the relative chance that a patient is/is not suffering from a particular illness. Codifying the way in which clinicians logically approach problems and deal with this uncertainty is a difficult task. What follows is my take on diagnostic and therapeutic decision making. It incorporates the following series of questions in a more or less step-wise fashion:
- Does this particular clinical situation, on the basis of the H&P, seem familiar to me (i.e. does it fit any pattern of disease that I have seen/read about)? Is there a single answer which explains even a multitude of complaints/findings? Referred to as Occam’s Razor this, in essence, is the search for the simplest possible explanation.
- What other explanations exist? This is known as the “differential.” Rather then long, it should be logical. The list is arranged from most to least likely and highlights those conditions that I absolutely do not want to miss (i.e. conditions that would result in significant morbidity/mortality if not promptly identified). When searching for explanations, remember that common things occur commonly. While patients do contract unusual illnesses, these are rather rare events. Thus, strange symptoms and findings are still more likely to represent an uncommon presentation of a common problem then to be due to an altogether uncommon illness. I didn’t make this up, but have found it to be a clinical truism. However, fear not Great Lupus Hunters, unusual illnesses do occur. Simply make sure that you really rule out the more run-of-the-mill stuff first!
- What (if anything) do I need to do to rule out the “really bad things” and how quickly does this need to be done? Can it be achieved as an outpatient or will hospitalization be required? This type of evaluation frequently produces a list of things that the patient does not have without ever reaching a definitive diagnosis, which is quite acceptable.
- Of the remaining potential explanations, do I need to make use of additional tests or am I comfortable enough with the available information to make a presumptive diagnosis and proceed on that basis alone?
Note: A test should only be obtained if the result will in some way affect your decision
making. That is, if you are going to embark on a particular strategy regardless of the results,
why obtain it in the first place? Furthermore, if you don’t understand the operating
characteristics of a test (e.g. it’s sensitivity and specificity and thus how the results
will affect your view of the candidate diagnosis), don’t order it.
- If the diagnosis is still unclear, can I use the passage of time as a diagnostic test (i.e. perhaps with time the process will more closely resemble a pattern of disease that will be recognizable to me or simply fade away without explanation)? I like to think of every patient as living on their own curve, with “wellness” measured along the Y axis, and time on the X axis. Curves, of course, cannot be plotted on the basis of a single point. Similarly, it may take several observations separated by time before I can begin to determine a patient’s clinical direction (i.e. getting better, worse or staying the same).
- Does this condition require specific therapy? If so, do I first need to have an established
diagnosis (see above)? Can the patient wait for this diagnosis to be made before initiating
treatment or does their clinical situation mandate the beginning of empirical therapy while
a diagnosis is simultaneously being sought? Can the treatment be administered as an
outpatient or will it require hospitalization (either because of its complexity, compliance
issues, patient’s compromised clinical condition, need to assess efficacy on a frequent basis,
etc.)? What will I do if the treatment fails to have the desired effect?
- Is the patient on board with this plan? Do they understand the rationale for the approach that I have chosen as well as their role?
Elements of uncertainty are wound into almost every case that you will see. Students and the public are surprised (and perhaps disappointed) to learn that, despite the abundance of technology that currently exists, physicians are still “reduced” to using their judgment when making clinical decisions (e.g. is a cold viral or bacterial? Should antibiotics be administered or withheld? etc.). The trick lies in knowing when its OK to be parsimonious with the use of testing and which situations demand a no-holds-barred pursuit of an answer.
Lets look at a few examples:
Case 1: A 24 year old otherwise healthy male presents with a 3 day history of cough productive of green sputum, fever, chills and slight shortness of breath associated with right sided chest pain. Exam is remarkable for a temperature of 102 F with otherwise normal vital signs. The patient looks well (i.e. not distressed). A few crackles are noted in the right base on lung exam. No other abnormalities are apparent.
Discussion: Following the above question template, we can reason through the case as follows:
1 & 2: This clinical situation seems most consistent with a well compensated bacterial pneumonia.
Other possibilities might include:
pulmonary embolism (P.E.)
Each, however, either cannot explain all of the findings present or is not supported by the objective
data. A viral infection, for example shouldn’t cause a focal lung exam; asthma can cause a cough and shortness of breath, but wheezing should be present; while a pulmonary embolism can cause shortness of breath, cough and chest pain, it should not result in fever, chills or sputum production. Furthermore, P.E.s generally occur in patient’s who have risk factors for this illness, none of which were present in this case. “Other” includes the list of unlikely diagnoses (e.g. eosinophilic pneumonitis, histoplasmosis, malignancy etc.) that would only be considered if the patient’s course deviated markedly from expected and/or could not be explained on the basis of those things higher on the differential.
3&4: Many clinicians would feel comfortable enough at this point (based on their clinical impression) to proceed without obtaining any additional tests to either support the candidate diagnosis or rule out other possibilities. Other approaches would also be acceptable. For example, another clinician may have seen a similar case in the past, treated the patient for a bacterial process, and found out later that they had actually had a P.E. Because of this experience, they might be uncomfortable proceeding without first obtaining a CXR (to confirm the presence of an infiltrate), CBC (to identify a leukocytosis c/w a bacterial infection) and an EKG (to look for stigmata of a P.E.). This approach would not necessarily be incorrect. It’s driven by a particular provider’s anecdotal experience, which for obvious reasons has a powerful impact on future decision making.
This is generally helpful, as long as it is based on logic and not fear. You might then wonder, “Why
not obtain confirmatory tests whenever possible?” Remember, tests come at a cost (in terms of
dollars, time and patient discomfort). You need to be able to justify, at least in your own mind if not that of the insurance company, the expense. Furthermore, the expected value of any test is dependent on the situation in which it is being applied. In general, as few tests in medicine have 100% sensitivity and specificity (i.e. correctly identify those with and without disease all of the time), the likelihood that a result is correct is dependent on how strongly you already feel about the candidate diagnosis. That is, if you are certain that someone is suffering from a particular disease (on the basis of history, exam and other findings) and you order an additional test “just to make sure” then the results of the test aren’t likely to have a significant impact on your decision making (i.e. if it confirms your suspicions, so what; if it conflicts, you’ll ignore it, treating the result as a false negative). The same principles apply in the reverse situation (i.e. if you are certain that someone does not have a particular illness). Tests have their greatest value when applied to situations where you’re truly on the fence about a particular diagnosis. An in depth discussion of this subject can be found in any text under Baysean Analysis.
5: In this setting, I would probably not be comfortable waiting for the process to “play out”
any further without initiating therapy. Bacterial processes tend to worsen unless they are treated,
even in otherwise healthy 23 year olds.
6: Therapy in this case could be initiated on an outpatient basis with an antibiotic directed against
Strep and H. Flu, the pathogens most commonly associated with respiratory infections in this age
group. Treatment would last for a total of 7 days (a somewhat arbitrary number) and the patient
instructed to return for re-evaluation on the last day of therapy to insure that the infection was
completely treated and that the antibiotics could be discontinued. In addition, they would be
told to contact me if they felt worse. If this, in fact, occurred I would have to consider why things did not go as I had anticipated. Did the patient have an atypical infection (e.g. Legionellosis)? Were they non-compliant with medications? Had they developed a complication (e.g. empyema)? Were they suffering from a particularly virulent strain of bacterium? Or was the initial diagnosis (e.g. infectious process) correct in the first place? The only way to make this determination (and catch the rare zebra) would be through re-evaluation of the patient, applying additional tests in a logical and ordered fashion.
Example 2, by design, is a bit more murky.
Case 2: A 55 year old male with history of Chronic Obstructive Pulmonary Disease (COPD), Coronary Artery Disease (CAD), and past Pulmonary Embolism (P.E.) presents with several hours of chest pain radiating to his left arm associated with shortness of breath and diaphoresis. This is somewhat reminiscent of his past myocardial infarction, but is also similar to past admissions for COPD and his P.E. . he’s just not sure. Exam is remarkable for a pale gentleman who looks quite distressed, sweating profusely. Vital signs remarkable for Temp 99 P 110 BP 180/100 RR 30 Sat 91%. JVP is at 8cm. Lung exam is remarkable for diffuse wheezing. Patient has bilateral lower extremity edema (1+ to the mid shin) with the right leg slightly more swollen then the left, which he says has been present since his DVT and subsequent PE several years ago.
Discussion: We will approach this case in the same way as the previous example.
1&2: An exacerbation of any of this patient’s underlying conditions could explain his presentation. On the basis of the history and examination, recurrent cardiac ischemia, a flair of his COPD or another P.E. are all possible. Additionally, this could represent new Congestive Heart Failure (CHF), perhaps associated with ongoing cardiac ischemia. Although he has never had this before, I know that CHF tends to occur in patients with CAD (which this patient does have) and can cause a clinical picture similar to that presented above. A bacterial infection (either bronchitis or
pneumonia) is usually accompanied by additional symptoms (e.g. fever, chills, sputum production), but remains a possibility, particularly as I know that COPD flairs usually occur in association with such an infection. Bringing up the rear would be “other” which would include, but not be
restricted to, the initial presentation of a lung cancer or a pneumothorax. These processes do
occur in patients with COPD but tend to present with other exam/historical findings (e.g. malignancy is often associated with weeks-to-months of weakness, fatigue, weight loss and a focal lung exam; pneumothorax causes decreased/absent breath sounds on the affected side). Ordering these possibilities from most to least likely, I would put coronary ischemia/CHF and P.E. 1 and 2, followed by COPD flair, pneumonia and “other.” In this case, I am impressed by the acuity of the presentation, which has increased my suspicion for the first 2 processes. The others, however, remain reasonable diagnostic possibilities which cannot be ruled out on a clinical basis.
3&4: All of the above conditions carry significant morbidity and/or mortality. In addition,
treatment strategies for each are quite different, and may themselves carry risk. Anti-coagulation
with heparin, for example, would be useful in patients with PEs but has no role in the
treatment of a COPD flair, and would unnecessarily expose the patient to the risk of bleeding.
There is also a sense of urgency that surrounds the need to make the diagnosis and begin
- the patient appears ill, with clear potential for further deterioration
- some of the treatments are only effective if applied within a narrow window of
opportunity (e.g. thrombolytics can open an occluded coronary artery and save
downstream myocardium only if they are given soon after the onset of
The following tests, essentially performed simultaneously, are necessary in order to rapidly
make a diagnosis:
- EKG to assess for evidence of acute myocardial infarction or stigmata of PE
- CXR to evaluate for signs of CHF, PE, infiltrate will also identity pneumothorax or evidence of malignancy
- CBC to asses for anemia which could be a precipitant for cardiac ischemia or shortness of breath; might also suggest a bacterial process if the white blood cell count is elevated.
- Chem 7 (includes electrolytes, BUN, Creatinine, Glucose) might be helpful in determination of volume status; also useful if patient will need diuretic therapy if diagnosed with CHF.
- Arterial Blood Gas to define Alveolar-arterial gradient and degree of gas exchange abnormality
- CK-MB, Cardiac Troponins will be elevated if patient has suffered a myocardial infarction.
The test results are as follows:
EKG remarkable for sinus tachycardia at 110, non-specific ST segment changes in the inferior
and precordial leads.
CXR consistent with moderate emphysema; no evidence infection, pneumothorax or malignancy;
? upper zone vascular redistribution consistent with CHF
Chem 7 and cardiac enzymes all normal.
CBC mildly elevated White Blood Cell Count (14 thousand) with normal differential;
ABG PO2 50, PCO2 30, PH 7.5 c/w hypoxemia and acute respiratory alkalosis
The diagnosis still remains in doubt. The data does not support an acute myocardial
infarction, though unstable angina without myocardial necrosis is still a possibility. There is
nothing to suggest a pneumonia, pneumothorax or malignancy, which we thought were
unlikely. I would now re-order my differential, placing pulmonary embolism at the top followed
by COPD exacerbation, CHF and unstable angina. I still need to press on and make use of additional tests in order to identify the correct diagnosis and institute appropriate therapy. In this
case I would obtain a radiologic test known as a Ventilation/Perfusion (V/Q) Scan in order to
determine if the patient has had a pulmonary embolism. A scan is obtained and interpreted as
“high probability for PE,” confirming the diagnosis. If, however, the scan had been interpreted as
“low probability for PE”, it would not necessarily rule out an embolous. This is due to the fact that
a V/Q scan is not 100% sensitive (i.e. there is a false negative rate). Thus, because I have a high
clinical suspicion that the patient has had a PE, a low probability scan would not rule out this
diagnosis. Were this to occur, I would proceed to a pulmonary angiogram (the “gold standard,”
which is not routinely performed as it is invasive and itself carries risks) to make a definitive
diagnosis. If this were negative, I would then be left with a diagnosis of COPD, CHF or unstable
angina. Further treatment would be based on the clinical course, response to therapy, and if, in
fact, there were any additional means of distinguishing between these possibilities.
5&6: Because this patient is rather ill, treatment should occur in concert with the diagnostic evaluation and would include:
- Oxygen: The patient is hypoxemic and would benefit from oxygen, regardless of the
- Heparin: As it may take some time before the V/Q scan is performed, I would elect to begin therapy with heparin while waiting to obtain the study. This decision is based on my high clinical suspicion that the patient has had an embolous. Because he already appears pretty sick and compromised, I would be uncomfortable withholding therapy that could prevent additional (and perhaps catastrophic) emboli. Furthermore, anti-coagulation would also be used as first-line treatment for unstable angina, which is #2 on my differential. Only in the setting of COPD flair (#3) would heparin be inappropriate. In this case, I have decided that the potential benefit of heparin out weighs the short term risk of bleeding.
- Aerosolized Albuterol (a beta-2 agonist): I would also give the patient a breathing
treatment with nebulized albuterol to see if it relieved any of his bronchospasm.
- Lasix: This would help improve hypoxemia in the event that some component of the
patient’s illness is caused by CHF. While I have no way of knowing definitively that
he does have pulmonary edema and would therefore benefit from diuresis, there are
few other therapeutic options that would have a rapid, dramatic impact on his gas
exchange. He certainly does not appear intravascularly depleted, a situation where
Lasix would be dangerous. By my calculations then, the potential benefits of this
treatment outweigh its risks.
- Steroids: Another therapeutic option would be to give him a dose of intravenous
steroids for his possible COPD. A single dose of steroids has little downside. I also
know that it takes a while (at least several hours) to have an effect. Thus,
particularly if it will take some time to get the V/Q scan, I would probably opt to
initiate steroid therapy, which could always be discontinued if the results were
consistent with a P.E.
- Antibiotics: Since COPD remains high on my list, and COPD flairs are most
frequently precipitated by bacterial infections, it’s logical to treat for this potential
problem as well.
Thrombolytics are not indicated. Nor would I initiate therapy with any of the more potent platelet inhibiting agents (the 2a/3b receptor antagonists), as these carry a higher risk of bleeding (at least until that point when unstable angina was more clearly the leading diagnosis). As additional data became available, I would need to continually re-evaluate all of these decisions.
There are a few themes that are common to both cases and help guide decision making in general:
- Patient substrate along with the clarity of presentation will have a significant impact on our willingness to accept clinical uncertainty. Healthy patients who present with “classic” complaints/findings can generally be managed comfortably on the basis of data acquired from the H&P alone. These patients have significant physiologic reserves and can tolerate incorrect treatments/diagnoses, affording us the opportunity to redirect our efforts in the event of a diagnostic/therapeutic misstep without having them suffer significantly. As patients accrue baseline illnesses, however, there is a marked increase in “background chatter,” making it difficult to pinpoint the etiology of an illness solely on the basis of the history and physical. Furthermore, the patient’s ability to tolerate additional insults becomes quite limited. Thus, we become both less capable of making diagnoses on clinical grounds alone and less comfortable proceeding without the utilization of diagnostic tests. That is, our willingness to accept uncertainty diminishes dramatically.
- Therapies that carry significant risk are generally reserved for more serious illnesses. They are
administered only when there is some reasonable certainty that the patient is suffering from the disease
or are doing so poorly that it’s felt to be worth the risk. Treatments that carry less risk are initiated
with relative impunity, although obviously even these options have downsides.
- Patients with significant baseline organ dysfunction often have little reserve and therefore tolerate additional insults poorly. This imparts a sense of urgency to the process which is generally absent when caring for healthier patients. While tests are being obtained to define the nature/extent of a clinical problem, empirical therapies directed against several of the most likely illnesses are initiated in an effort to prevent further deterioration. If no tests are available that can accurately distinguish between several diagnoses, a number of different therapeutic strategies may be pursued in parallel. These patients frequently require very close monitoring (i.e. in-hospital treatment) so that appropriate adjustments can be made and clinical down-turns rapidly addressed. This everything-but-the-kitchen-sink mentality is certainly less appealing and elegant then focused therapy where interventions are initiated individually, enabling the clinician to clearly gauge their efficacy and minimize any associated therapeutic risk. Unfortunately, it is often impossible to identify the exact limb of the system that is broken, forcing the use of multi-modality approaches.
- Clinicians listen, ask, examine and then try to use the information obtained to put patients into known diagnostic categories. The differential diagnosis is a list of these possibilities and can be thought of as a group of boxes of various shapes and sizes. The quantity and variety of these containers is a function of the way in which providers combine historical information and exam findings with clinical experience and their understanding of pathophysiology. Frequently there is an imperfect fit between what we find and what we know. Instead of trying to cram patients into ill suited spaces we should recognize that these instances indicate a need to either expand our knowledge base, change our approach to a patient’s complaint, or gather additional data. Remember also that there is no shame in admitting that we can’t explain a particular finding or symptom. In fact, knowing what something is not has as much value as providing a specific label for a complaint or condition.
- Clinical decision making is based on the expectation that the human body will respond to illness in a predictable way. Disease states which alter this behavior wreak havoc on logical decision making.
Illnesses which decrease normal immune function (e.g. HIV infection), medications that blunt the
response to infection (e.g. steroids, chemotherapy), or advanced age (which frequently leads to
an impaired physiologic response to stress) are a few examples. When caring for these patients,
providers are forced to cast a wide net, relying on the initiation of empirical therapies while multiple
diagnostic avenues are pursued.
- There will always be some element of non-modifiable uncertainty in clinical medicine (unless, of course, someone actually invents Star Trek-type Tricordors!). Technological advances have succeeded in improving both diagnostic and therapeutic accuracy. However, they are almost never correct all of the time. Reliance on the less-then-perfect information that these tests provide without in some way taking into account clinical judgment can have serious consequences for the patient.
Adapted, with permission from University of California, San Diego School of Medicine By Charlie Goldberg, M.D.