A new diagnostic test for tuberculosis has a sensitivity of 95% and specificity of 90%. In a population with a TB prevalence of 5%, which test characteristic BEST distinguishes this test's ability to rule IN disease from its ability to rule OUT disease?

Explanation

## Understanding Rule-In vs Rule-Out: Likelihood Ratios **Key Point:** Likelihood ratios (LR+ and LR−) are the **best** measures to distinguish a test's ability to **rule IN** disease from its ability to **rule OUT** disease, because they are prevalence-independent and directly quantify diagnostic utility in each direction. ### Why Option D is Correct The question asks which characteristic **BEST distinguishes** the test's ability to rule IN disease **from** its ability to rule OUT disease. This is a direct comparison of two directional diagnostic properties. **Likelihood Ratio for a Positive Test (LR+):** $$LR+ = \frac{Sensitivity}{1 - Specificity} = \frac{0.95}{1 - 0.90} = \frac{0.95}{0.10} = 9.5$$ **Likelihood Ratio for a Negative Test (LR−):** $$LR- = \frac{1 - Sensitivity}{Specificity} = \frac{1 - 0.95}{0.90} = \frac{0.05}{0.90} \approx 0.056$$ - **LR+ = 9.5** → A positive result increases the odds of disease ~9.5-fold → good for **ruling IN** disease - **LR− ≈ 0.056** → A negative result reduces the odds of disease to ~1/18th → excellent for **ruling OUT** disease The contrast between LR+ (9.5) and LR− (0.056) **directly and simultaneously** captures both directional abilities in a single comparison, independent of prevalence. Stating that LR+ > LR− is the single statement that best encapsulates this distinction. ### Why Option B is Incorrect While PPV (~33%) is indeed much lower than NPV (~99.7%) in this low-prevalence setting, this difference is **driven by prevalence**, not by the intrinsic test characteristics. PPV and NPV change with every population; they do not "best distinguish" the test's inherent rule-in vs rule-out ability. The question asks about a test characteristic — LRs are fixed properties of the test, PPV/NPV are not. ### Why Options A and C are Incorrect - **Option A:** Sensitivity being higher than specificity (95% vs 90%) is a minor numerical difference and does not directly address rule-in vs rule-out ability. - **Option C:** This reverses the correct relationship — sensitivity determines NPV (via true negatives) and specificity determines PPV (via true positives), making this statement factually incorrect. **High-Yield:** Per Sackett and evidence-based medicine principles (Harrison's Principles of Internal Medicine), LR+ > 10 strongly rules IN disease, and LR− < 0.1 strongly rules OUT disease. LRs are the gold standard for comparing directional diagnostic utility. ### Mnemonic: SPIN & SNOUT via LRs - **High LR+** → rules IN (like high SPecificity → SPIN) - **Low LR−** → rules OUT (like high SeNsitivity → SNOUT) **Clinical Pearl:** Unlike PPV and NPV, likelihood ratios can be applied across different prevalence settings using Bayes' theorem (Fagan nomogram), making them the preferred tool for comparing a test's rule-in vs rule-out performance in clinical practice (Harrison's Principles of Internal Medicine, 21st ed.).