EPISODE 4: BEDSIDE DIAGNOSTICS & DECISION RULES

EPISODE 4: Bedside Diagnostics — Why Prevalence Breaks Your D-Dimer & FAST Exam

Subtitle: The 2x2 Matrix, the Prevalence Paradox, Bayesian Likelihood Ratios, and Clinical Decision Rules

Episode Overview

Every shift in the Emergency Department is a high-stakes exercise in Bayesian probability. In Episode 4, we tackle diagnostic testing and Clinical Decision Rules (CDRs). We show you why memorizing a test's Positive or Negative Predictive Value from a paper can kill your patient if your ED's disease prevalence differs from the study cohort, and how to use Likelihood Ratios ($LR+$ and $LR-$) to calculate exact post-test probabilities in your head at the bedside.

High-Yield Clinical Takeaways

• The "Truth at the Top" $2\times2$ Matrix: Why Sensitivity ($\text{SnNOut}$) and Specificity ($\text{SpPIn}$) are calculated vertically down the columns (making them static properties of the test), whereas $PPV$ and $NPV$ are calculated horizontally across the rows.
• The Prevalence Paradox: How moving the exact same test from a low-risk clinic to a high-risk resuscitation bay dramatically alters $PPV$ and $NPV$.
• Likelihood Ratios ($LR+$ and $LR-$): Why Likelihood Ratios are the undefeated champions of bedside diagnostics, plus the Fagan Nomogram mental math shortcut ($LR+$ of 2, 5, and 10 shifts probability by roughly $+15%, +30%,$ and $+45%$).
• The 3 Lethal Diagnostic Biases:

1. Spectrum Bias: Cherry-picking florid ICU cases vs. healthy med students (inflates both Sensitivity and Specificity).
2. Verification (Workup) Bias: Only ordering the Gold Standard CT on patients with a positive screening test (artificially drives False Negatives toward zero and inflates Sensitivity).
3. Incorporation Bias: Including the index test inside the definition of the Gold Standard.

• The 3-Stage CDR Pathway (Stiell-McGinn): Why you must never use a Stage 1 Derived rule until it has undergone Stage 2 Prospective Broad Validation ($\kappa > 0.6$) and Stage 3 Impact Analysis.

Clinical Vignettes Covered

• The D-Dimer Prevalence Paradox (Worked Math): Proving how a D-dimer with 95% Sensitivity and 50% Specificity yields a 99.8% $NPV$ in a low-risk clinic (2% PE prevalence), but drops to an 87% $NPV$ in a high-risk cohort (60% PE prevalence)—meaning 13% of patients with a NEGATIVE D-dimer still have a PE!
• The Trauma FAST Exam (Worked Bayesian Loop): Why a negative E-FAST exam ($LR- = 0.20$) in a hypotensive blunt trauma patient with a 50% pre-test probability still leaves a 16.7% residual risk of hemoperitoneum.

The 3:00 AM Resus Bay Bottom Line: Never trust a negative screening test—even one with 95% sensitivity—when your clinical pre-test probability is sky-high. Always filter your test result through your patient's pre-test probability using Likelihood Ratios.