1: Diagnostic Process

Constructing a differential diagnosis, choosing diagnostic tests, and interpreting the results are key skills for all physicians. The diagnostic process, often called clinical reasoning, is complex, and errors in reasoning are thought to account for 17% of all adverse events. Diagnostic errors can occur due to faulty knowledge, faulty data gathering, and faulty information processing. While this chapter will focus on the reasoning process, remember that the data you acquire through your history and physical exam, sometimes accompanied by preliminary laboratory tests, form the basis for your initial clinical impression. Even with flawless reasoning, your final diagnosis will be wrong if you do not start with accurate data. You must have well developed interviewing and physical examination skills.

Clinicians often use a combination of 2 reasoning processes: non-analytical/intuitive and analytical. The intuitive process is rapid and consists of an unconscious match to examples stored in memory, while the analytical process is slow, logical, and rule-based. Clinicians should be aware of common biases in clinical reasoning (Table 1-1) and reflect upon their reasoning processes, looking for potential errors. This chapter breaks down the reasoning process into a series of steps that can help you work through large differential diagnoses, avoid biases, and retrospectively identify sources of error when your diagnosis is wrong.

(Figure 1-1)

Step 1: Identify the Problem

Be certain you understand what the patient is telling you. Sometimes “I’m tired” means “I become short of breath when I walk” and at other times means “My muscles are weak.” Construct a complete problem list consisting of the chief complaint, other acute symptoms and physical exam abnormalities, chronic active problems (such as diabetes or hypertension), and important past problems (such as history of bowel obstruction or cancer). Problems that are likely to be related, such as shortness of breath and chest pain, should be grouped together. It is necessary to accurately identify the problem every time you evaluate a patient.

Step 2: Frame the Differential Diagnosis

The differential diagnosis should be framed in a way that facilitates recall. It might be possible to memorize long lists of causes, or differential diagnoses, for various problems. However, doing so would not necessarily lead to a useful organization of differentials that helps you remember or use them. Instead, it is preferable to use some kind of problem-specific framework to organize differentials into subcategories that are easier to remember and often clinically useful. Problem-specific frameworks can be anatomic, a framework often used for chest pain; organ/system, used for symptoms with very broad differentials like fatigue; physiologic; or based on pivotal points (defined below). Each chapter in Symptom to Diagnosis begins with a problem-specific framework for the differential. Using such frameworks has been shown to improve the diagnostic accuracy of medical students.

Step 3: Organize the Differential Diagnosis

Structuring the differential diagnosis into clinically useful subgroups can enable you to systematically work through the differential diagnosis. Sometimes the framework that is easiest to remember, such as grouping causes of dyspnea as cardiac or pulmonary, does not facilitate reasoning. Then, reorganizing the differential in a way that helps you understand the order in which to consider various diagnoses is necessary. The most clinically useful differentials are organized using pivotal points, one of a pair of opposing descriptors that compare and contrast diagnoses, or clinical characteristics. Examples include old versus new headache, unilateral versus bilateral edema, and right lower quadrant pain versus epigastric pain. When pivotal points are used to frame the differential in the first place, it is not necessary to reorganize the differential.

You can frame and reorganize the differential yourself or find a source that does so in a way that makes sense to you. Each chapter in Symptom to Diagnosis contains a diagnostic algorithm that uses pivotal points to highlight logical reasoning pathways for each symptom. Steps 2 and 3 need to be done only once for each problem you encounter; with experience, you will develop a repertoire of logically framed differentials and structured diagnostic approaches.

Step 4: Limit the Differential Diagnosis

Since every disease in a differential may not be relevant to an individual patient, using pivotal points to create a patient-specific differential diagnosis can help narrow the list. Extracting pivotal points from the history and physical exam enables the clinician to limit a large, complete differential diagnosis to a more focused set of diagnoses pertinent to that particular patient. This step, and steps 5 through 9, should be included in your clinical reasoning for all patients.

Step 5: Explore Possible Diagnoses Using History and Physical Exam Findings

The next step is to look for clinical clues that point toward the most likely diagnosis. Does the patient have risk factors for a particular diagnosis? Does the patient’s description of the symptom suggest a likely cause? What have you observed on physical exam? Focus on the positive—positive findings on history or physical exam are important (65% of positive findings have a specificity > 80% and 43% of positive findings have a specificity > 90%). One-third have an LR+ > 5, and 16% have an LR+ > 10. Some very specific findings strongly suggest a specific diagnosis because they are rarely seen in patients without the disease, just as fingerprints point to a specific person because they are not seen in more than 1 individual. Such “fingerprint” findings will be marked with the symbol “FP” throughout the book. On the other hand, do not be fooled by the negative; “classic” findings, especially individual findings, are often absent. Only 21% of negative findings have a sensitivity > 80%, and only 11% of > 90%; just 7% have an LR– of < 0.1.

Step 6: Rank the Differential Diagnosis

Rank the differential diagnosis using the results obtained in Step 5. Even in a limited differential, not all diagnoses are equally likely or equally important. There are 4 approaches to ranking, or prioritizing, the differential diagnosis for a given problem: possibilistic, probabilistic, prognostic, and pragmatic.

1. Possibilistic approach: Consider all known causes equally likely and simultaneously test for all of them. This is not a useful approach.

2. Probabilistic approach: Consider first those disorders that are more likely; that is, those with the highest pretest probability, the probability that a disease is present before further testing is done.

3. Prognostic approach: Consider the most serious diagnoses first.

4. Pragmatic approach: Consider the diagnoses most responsive to treatment first.

Clearly there are limitations to each of these individual approaches. Experienced clinicians simultaneously integrate probabilistic, prognostic, and pragmatic approaches when reorganizing and prioritizing a differential diagnosis in order to decide when testing is necessary and which test to order (Table 1-2). Clinicians use their knowledge of pivotal points; “fingerprints”; risk factors; typical or “textbook” presentations of disease; the variability of disease presentation; and prevalence and prognosis to select a leading hypothesis, must not miss hypotheses, and other active alternative hypotheses.

Step 7: Test Your Hypotheses

Sometimes you are certain about the diagnosis based on the initial data and proceed to treatment. Most of the time, however, you require additional data to confirm your diagnostic hypotheses; in other words, you need to order diagnostic tests. Whenever you do so, you should understand how much the test will change the probability the patient has the disease in question.

Step 8: Re-rank the Differential Based on New Data

Remember, ruling out a disease is usually not enough; you must also determine the cause of the patient’s symptom. For example, you may have eliminated myocardial infarction (MI) as a cause of chest pain, but you still need to determine whether the pain is due to gastroesophageal reflux, muscle strain, aortic dissection, etc. Whenever you have not made a diagnosis, or when you encounter data that conflict with your original hypotheses, go back to the complete differential diagnosis and reprioritize it, taking the new data into consideration. Failure to reconsider the possibilities is called premature closure (see Table 1-1), one of the most common diagnostic errors made by clinicians.

Step 9: Test the New Hypotheses

Repeat the process until a diagnosis is reached.

Step 1: Identify the Problem

Problem lists should begin with the acute problems, followed by chronic active problems, ending with inactive problems. Mrs. S’s problems are (1) painful left leg edema with erythema, (2) hypertension, (3) osteoarthritis of the knees, and (4) status post cholecystectomy.

Step 2: Frame the Differential Diagnosis

As discussed in Chapter 17, Edema, the problem-specific organization of the full differential diagnosis starts with the distribution of the edema: generalized versus unilateral and limb versus localized. The causes of edema are fairly distinct for each of these subcategories. For instance, heart failure and chronic kidney disease cause generalized not unilateral edema.

Step 3: Organize the Differential Diagnosis

Since the edema differential is framed using the pivotal point of edema distribution, it is not necessary to organize it—step 3 has already been done.

Step 4: Limit the Differential Diagnosis

Mrs. S has acute unilateral leg edema, a pivotal point that leads to a limited portion of the edema differential.

Diagnostic possibilities are now narrowed to a distinct subset of diseases that can be organized using an anatomic framework:

1. Skin: Stasis dermatitis

2. Soft tissue: Cellulitis

3. Calf veins: Distal deep venous thrombosis (DVT)

4. Knee: Ruptured Baker cyst

5. Thigh veins: Proximal DVT

6. Pelvis: Mass causing lymphatic obstruction

Step 5: Use History and Physical Exam Findings to Explore Possible Diagnoses

Consider the risk factors for each of the diagnostic possibilities as well as their associated symptoms and signs. For example, venous insufficiency is a risk factor for stasis dermatitis, and there may be hemosiderin staining along the malleolar surface on physical exam. Cellulitis often follows skin injury, and physical exam shows erythema and tenderness. DVT is more frequent in patients with underlying malignancy or recent immobilization, and there may be shortness of breath if the clot has embolized.

Step 6: Rank the Differential Diagnosis

Mrs. S has a constellation of symptoms and signs supporting the diagnosis of cellulitis as the leading hypothesis: fever; an entry site for infection on her foot; and a red, tender, swollen leg. Even without risk factors for DVT, the active alternatives are proximal and calf DVT, being both common and “must not miss” diagnoses. If cellulitis and DVT are not present, ruptured Baker cyst and a pelvic mass should be considered. Finally, stasis dermatitis is excluded in a patient without a history of chronic leg swelling.

Step 7: Test Your Hypotheses

Once you have generated a leading hypothesis, with or without active alternatives, you need to decide whether you need further information before proceeding to treatment or before excluding other diagnoses. One way to think about this is in terms of certainty: how certain are you that your hypothesis is correct, and how much more certain do you need to be before starting treatment? Another way to think about this is in terms of probability: is your pretest probability of disease high enough or low enough that you do not need any further information from a test?

Determine the Pretest Probability

There are several ways to determine the pretest probability of your leading hypothesis and most important (often most serious) active alternatives: use a validated clinical decision rule (CDR), use prevalence data regarding the causes/etiologies of a symptom, and use your overall clinical impression.

1. Use a validated CDR

   1. Investigators construct a list of potential predictors of a disease, and then examine a group of patients to determine whether the predictors and the disease are present.

      1. Logistic regression is then used to determine which predictors are most powerful and which can be omitted.

      2. The model is then validated by applying it in other patient populations.

      3. To simplify use, the clinical predictors in the model are often assigned point values, and different point totals correspond to different pretest probabilities.

   2. CDRs are infrequently available but are the most precise way of estimating pretest probability.

   3. If you can find a validated CDR, you can come up with an exact number (or a small range of numbers) for your pretest probability.

2. Use information about the prevalence of etiologies for a symptom.

   1. You can sometimes find this information in textbooks or review articles.

   2. You can find studies providing this information by searching the symptom in question, combined with the term “differential diagnosis.”

   3. It is important to assess the quality of the studies you find before using the data. Guyatt’s Users’ Guides to the Medical Literature provides criteria for evaluating articles about differential diagnosis and disease frequency.

3. Use your overall clinical impression.

   1. This is a combination of what you know about disease prevalence and the match between the expected history and physical with that of the patient, mixed with your clinical experience, and the ever elusive attribute “clinical judgment.”

   2. This is just as imprecise as it sounds, and it has been shown that physicians are disproportionately influenced by their most recent clinical experience.

   3. Nevertheless, it has also been shown that the overall clinical impression of experienced clinicians has significant predictive value.

   4. Clinicians generally categorize pretest probability as low, moderate, or high. This rather vague categorization is still helpful. Do not get distracted thinking a number is necessary.

Consider the Potential Harms

Consider the potential harms of both a missed diagnosis and the treatment.

1. It is very harmful to miss certain diagnoses, such as MI or pulmonary embolism, while it is not so harmful to miss others, such as mild carpal tunnel syndrome. You need to be very certain that life-threatening diseases are not present (that is, have a very low pretest probability), before excluding them without testing.

2. Some treatments, such as thrombolytics, are more harmful than others, such as oral antibiotics; you need to be very certain that potentially harmful treatments are needed (that is, the pretest probability is very high) before prescribing them without testing.

The ends of the bar in the threshold model represent 0% and 100% pretest probability. The treatment threshold is the probability above which the diagnosis is so likely you would treat the patient without further testing. The test threshold is the probability below which the diagnosis is so unlikely it is excluded without further testing (Figure 1-2).

For example, consider Ms. A, a 19-year-old woman, who complains of 30 seconds of sharp right-sided chest pain after lifting a heavy box. The pretest probability of cardiac ischemia is so low that no further testing is necessary (Figure 1-3).

Now consider Mr. B, a 60-year-old man, who smokes and has diabetes, hypertension, and 15 minutes of crushing substernal chest pain accompanied by nausea and diaphoresis, with an ECG showing ST-segment elevations in the anterior leads. The pretest probability of an acute MI is so high you would treat without further testing, such as measuring cardiac enzymes (Figure 1-4).

Diagnostic tests are necessary when the pretest probability of disease is in the middle, above the test threshold and below the treatment threshold. A really useful test shifts the probability of disease so much that the posttest probability (the probability of disease after the test is done) crosses one of the thresholds (Figure 1-5).

A perfect diagnostic test would always be positive in patients with the disease and would always be negative in patients without the disease (Figure 1-6). Since there are no perfect diagnostic tests, some patients with the disease have negative tests (false-negative), and some without the disease have positive tests (false-positive) (Figure 1-7).

The test characteristics help you to know how often false results occur. They are determined by performing the test in patients known to have or not have the disease, and recording the distribution of results (Table 1-3).

Table 1-4 shows the test characteristics of duplex ultrasonography for the diagnosis of proximal DVT, based on a hypothetical group of 200 patients, 90 of whom have DVT.

The sensitivity is the percentage of patients with DVT who have a true-positive (TP) test result:

Since tests with very high sensitivity have a very low percentage of false-negative results (in Table 1-4, 4/90 = 0.04 = 4%), a negative result is likely a true negative.

The specificity is the percentage of patients without DVT who have a true-negative (TN) test result:

Since tests with very high specificity have a low percentage of false-positive results (in Table 1-4, 2/110 = 0.02 = 2%), a positive result is likely a true positive.

The sensitivity and specificity are important attributes of a test, but they do not tell you whether the test result will change your pretest probability enough to move beyond the test or treatment thresholds; the shift in probability depends on the interactions between sensitivity, specificity, and pretest probability. The likelihood ratio (LR), the likelihood that a given test result would occur in a patient with the disease compared with the likelihood that the same result would occur in a patient without the disease, enables you to calculate how much the probability will shift.

The positive likelihood ratio (LR+) tells you how likely it is that a result is a true-positive (TP), rather than a false-positive (FP):

Positive LRs that are significantly above 1 indicate that a true-positive is much more likely than a false-positive, pushing you across the treatment threshold. An LR+ > 10 causes a large shift in disease probability; in general, tests with LR+ > 10 are very useful for ruling in disease. An LR+ between 5 and 10 causes a moderate shift in probability, and tests with these LRs are somewhat useful. “Fingerprints,” findings that often rule in a disease, have very high positive LRs.

The negative likelihood ratio (LR–) tells you how likely it is that a result is a false-negative (FN), rather than a true-negative (TN):

Negative LRs that are significantly less than 1 indicate that a false-negative is much less likely than a true-negative, pushing you below the test threshold. An LR– less than 0.1 causes a large shift in disease probability; in general, tests with LR– less than 0.1 are very useful for ruling out disease. An LR– between 0.1 and 0.5 causes a moderate shift in probability, and tests with these LRs are somewhat useful.

The closer the LR is to 1, the less useful the test; tests with a LR = 1 do not change probability at all and are useless. The threshold model in Figure 1-8 incorporates LRs and illustrates how tests can change disease probability.

When you have a specific pretest probability, you can use the LR to calculate an exact posttest probability (see Box, Calculating an Exact Posttest Probability and Figure 1-9, Likelihood Ratio Nomogram). Table 1-5 shows some examples of how much LRs of different magnitudes change the pretest probability.

If you are using descriptive pretest probability terms such as low, moderate, and high, you can use LRs as follows:

1. A test with an LR– of 0.1 or less will rule out a disease of low or moderate pretest probability.

2. A test with an LR+ of 10 or greater will rule in a disease of moderate or high probability.

3. Beware if the test result is the opposite of what you expected!

   1. If your pretest probability is high, a negative test rarely rules out the disease, no matter what the LR– is.

   2. If you pretest probability is low, a positive test rarely rules in the disease, no matter what the LR+ is.

   3. In these situations, you need to perform another test.