Diagnosis: Differential Reasoning and Test Interpretation
What the diagnosis domain tests
Diagnosis items reward structured clinical reasoning. Generate a differential before narrowing: rule out the dangerous possibilities first, then weight the most probable. Know how to distinguish look-alike conditions - asthma versus COPD versus heart failure, cellulitis versus stasis dermatitis, migraine versus tension versus cluster headache, DKA versus hyperosmolar hyperglycemic state - by the discriminating features rather than the shared ones.
Test interpretation is heavily tested. Understand sensitivity (the proportion of diseased patients who test positive) and specificity (the proportion of disease-free patients who test negative). Positive and negative predictive values depend on prevalence, so the same test performs differently in a high-risk versus low-risk population. Likelihood ratios translate pretest probability into post-test probability. Know that no test is perfect: a negative result from a moderately sensitive test does not exclude disease when pretest probability is high.
Traps that appear in items
- Anchoring on the first plausible diagnosis and ignoring a competing hypothesis that better fits one finding.
- Confusing sensitivity with specificity, or treating a highly specific test as a good screening tool.
- Assuming a normal result rules out disease regardless of pretest probability.
- Selecting a diagnosis whose required criteria are absent from the vignette.
- Mistaking statistical significance for clinical importance.
Rule of thumb
SnNout and SpPin: a highly Sensitive test that is Negative helps rule out; a highly Specific test that is Positive helps rule in.
Worked example
A 34-year-old with unilateral calf pain and swelling has a low pretest probability by a validated clinical prediction rule. The appropriate next step is a D-dimer; a negative result in a low-probability patient makes venous thromboembolism unlikely, while a positive result leads to ultrasound. Ordering ultrasound first in a low-risk patient ignores pretest probability.