Guide systematic construction of differential diagnoses that are comprehensive, appropriately prioritized, and resistant to common cognitive errors. A well-constructed DDx is the foundation of diagnostic accuracy.
Core DDx Frameworks
VINDICATE Mnemonic (Etiologic)
The classic categorical approach ensures broad etiologic coverage:
V — Vascular (thrombosis, embolism, hemorrhage, vasculitis, aneurysm)
I — Infectious (bacterial, viral, fungal, parasitic, prion)
N — Neoplastic (primary, metastatic, paraneoplastic, hematologic malignancy)
D — Degenerative / Deficiency (osteoarthritis, nutritional deficiency, neurodegeneration)
C — Congenital (structural anomalies, inborn errors of metabolism, genetic syndromes)
A — Autoimmune / Allergic (systemic autoimmunity, organ-specific, hypersensitivity)
関連 Skill
T — Traumatic (blunt, penetrating, thermal, radiation, barotrauma)
E — Endocrine / Metabolic (hormonal excess or deficiency, electrolyte disorders, acid-base)
Anatomical Framework
Systematically walk through structures in the affected region:
Start at the skin and work inward (or vice versa)
Consider each organ system traversed
Include referred pain sources (e.g., diaphragmatic irritation presenting as shoulder pain)
Consider vascular supply and drainage pathways
Map lymphatic and neural distributions
Physiological / Pathophysiologic Framework
Organize by mechanism of disease:
Obstruction (mechanical vs functional)
Perfusion (ischemia, infarction, congestion)
Inflammation (infectious vs sterile)
Metabolic derangement
Neoplastic proliferation
Immune dysregulation
Structural failure
Worst-First (Threat-Based) Approach
Prioritize by potential lethality and time-sensitivity:
Immediately life-threatening — Must rule out now (e.g., PE, MI, aortic dissection, ectopic pregnancy, meningitis)
Serious but not immediately lethal — Require urgent workup (e.g., malignancy, deep abscess, unstable fracture)
Common and likely — High prior probability diagnoses
Must-not-miss — Low probability but catastrophic if missed (e.g., subarachnoid hemorrhage in headache)
Treatable conditions — Conditions where specific therapy changes outcomes
Structured DDx Construction Process
Step 1: Define the Problem Representation
Construct a one-sentence semantic summary before generating differentials. Example: "A 65-year-old male smoker with acute-onset pleuritic chest pain and hypoxemia" immediately narrows the search space.
Step 2: Generate the Initial List
Apply at least two frameworks (e.g., VINDICATE + worst-first)
Aim for 5-10 diagnoses initially
Include at least one diagnosis from each relevant category
Force consideration of "cannot-miss" diagnoses regardless of perceived probability
Step 3: Prioritize
Rank by a combination of:
Prior probability — Epidemiologic likelihood given demographics, risk factors, setting
Clinical fit — How well the presentation matches the illness script
Severity/urgency — Consequence of missing the diagnosis
Treatability — Whether specific therapy exists
Step 4: Test and Refine
Identify discriminating features (findings that distinguish between top diagnoses)
Select targeted investigations based on likelihood ratios
Apply Bayesian updating as results return
Actively seek disconfirming evidence for the leading diagnosis
Cognitive Pitfalls in DDx Generation
Anchoring Bias
Fixating on an initial diagnosis despite disconfirming evidence. The first piece of information disproportionately influences reasoning.
Mitigation: Explicitly revisit the DDx when new data arrives. Ask: "What if my leading diagnosis is wrong?"
Premature Closure
Accepting a diagnosis before adequate verification. The most common cognitive error in diagnostic failure.
Mitigation: Apply the "rule of three" — always have at least three active diagnoses until one is confirmed. Use the diagnostic timeout: pause and reconsider before finalizing.
Availability Bias
Overweighting diagnoses that come easily to mind (recent cases, dramatic presentations, board-review conditions).
Mitigation: Use systematic frameworks rather than free recall. Ask: "Am I thinking of this because it's likely, or because I saw it recently?"
Search Satisfying
Stopping the search once one abnormality is found (e.g., finding a UTI and missing urosepsis, or finding a rib fracture and missing a pneumothorax).
Mitigation: Complete the full evaluation even after finding an initial diagnosis.
Framing Effect
Being influenced by how information is presented (e.g., referral diagnosis, triage label).
Mitigation: Obtain your own history. Reframe the presentation from primary data.
Base Rate Neglect
Ignoring disease prevalence when estimating probability. A positive D-dimer in a low-risk patient is more likely false-positive than true-positive.
Mitigation: Explicitly estimate pre-test probability before ordering tests. Apply validated clinical decision rules.
Representativeness
Expecting diseases to present classically. Atypical presentations are common, especially in elderly, immunocompromised, and pediatric populations.
Mitigation: Consider atypical presentations explicitly. Ask: "How might this disease present differently in this patient?"
Quality Markers of a Strong DDx
Includes at least one "cannot-miss" diagnosis
Covers multiple etiologic categories
Accounts for the patient's specific demographics and risk factors
Explains all major findings (or acknowledges unexplained features)
Contains diagnoses that are testable and distinguishable
Is dynamically updated as new information emerges
Applying This Skill
When asked about differential diagnosis:
Always ask for or establish the clinical context (age, sex, acuity, setting)
Present diagnoses in a structured, prioritized format
Flag "cannot-miss" diagnoses explicitly
Identify the key discriminating features and recommended next investigations
Note which cognitive biases are most relevant to the specific scenario