Decision making under uncertainty

Diagnosis - The Initial Hunch

• Pattern Recognition (System 1 Thinking): Rapid, intuitive matching of a patient's presentation to learned "illness scripts." This forms the initial diagnostic hypothesis.
• Illness Scripts: Mental frameworks for diseases, including key features: epidemiology, pathophysiology, symptoms, and expected course.

• Generating a Differential (DDx): A crucial step to avoid premature closure. Use a systematic approach.
  • 📌 VINDICATE Mnemonic:
    • Vascular, Infectious, Neoplastic, Degenerative, Iatrogenic, Congenital, Autoimmune, Traumatic, Endocrine.

⭐ Availability Heuristic: A cognitive bias where clinicians overestimate the likelihood of diseases that are more memorable or recently seen, potentially skewing the initial hunch.

Clinical Reasoning - Mind's Tricky Shortcuts

Cognitive biases are systematic errors in thinking that affect clinical judgment. Awareness is the first step to mitigation.

• Availability Heuristic: Overestimating likelihood of diagnoses that are recent, dramatic, or memorable.
  • Example: Seeing a rare case and then diagnosing it more frequently.
• Anchoring Bias: Over-relying on initial information, failing to adjust for later findings.
  • Example: Sticking to an ER diagnosis despite new, conflicting lab results.
• Confirmation Bias: Seeking and favoring data that confirms your initial hypothesis.
  • Example: Ordering tests to prove a suspected diagnosis, not to rule out others.
• Premature Closure: Accepting a diagnosis too early, failing to consider other reasonable alternatives.

⭐ Metacognition-stepping back to reflect on one's own thinking process ("thinking about thinking")-is the most effective strategy to reduce cognitive errors.

Decision Making - Playing The Odds

• Threshold Model: Clinical decisions hinge on probability.

  • Treatment Threshold: Probability above which you treat without further testing.
  • Testing Threshold: Probability below which you dismiss the diagnosis without testing.
  • The zone between these thresholds is the "test zone."

• Bayesian Reasoning: Update pre-test probability (PTP) to post-test probability using test results.

  • Likelihood Ratios (LRs): Quantify the power of a test.
    • LR+ > 10 is strong evidence to rule IN a disease.
    • LR- < 0.1 is strong evidence to rule OUT a disease.

⭐ A test with a high likelihood ratio positive (LR+) is most useful for confirming a diagnosis when your clinical suspicion (pre-test probability) is already moderate to high.

Reasoning Models - Two Brains, One Doc

• Dual Process Theory: A model of clinical reasoning involving two distinct but complementary cognitive systems.
• System 1 (Intuitive):
  • Fast, automatic, and based on pattern recognition and heuristics.
  • Efficient for experienced clinicians in common scenarios.
  • Vulnerable to cognitive biases (e.g., anchoring, availability).
• System 2 (Analytical):
  • Slow, deliberate, and systematic.
  • Involves hypothesis testing and weighing evidence.
  • Reduces bias but is resource-intensive; essential for complex or novel cases.

⭐ Most diagnostic errors are linked to System 1 heuristics operating unchecked. Effective reasoning involves toggling between systems, using System 2 to verify System 1's initial impressions.

High‑Yield Points - ⚡ Biggest Takeaways

• Bayesian inference is crucial: pre-test probability is not static and directly impacts post-test probability after a diagnostic result.
• Cognitive biases like anchoring, availability, and confirmation bias are common pitfalls that lead to diagnostic errors.
• Likelihood Ratios (LRs) are powerful: LR+ >10 significantly increases disease probability; LR- <0.1 significantly decreases it.
• The threshold model dictates action: don't test below the testing threshold; treat empirically above the treatment threshold.
• Leverage both System 1 (intuitive) and System 2 (analytical) thinking, but be wary of System 1's potential for error.