Screening Metrics - Numbers Don't Lie

-
Sensitivity & Specificity (Intrinsic to the test)
- Sensitivity: $TP / (TP + FN)$. Probability of a positive test in a diseased person.
- High sensitivity is used to Screen and Negate a disease (SNOUT).
- Specificity: $TN / (TN + FP)$. Probability of a negative test in a healthy person.
- High specificity is used to SPin and IN a diagnosis (SPIN).
- Sensitivity: $TP / (TP + FN)$. Probability of a positive test in a diseased person.
-
Predictive Values (Vary with disease prevalence)
- Positive Predictive Value (PPV): $TP / (TP + FP)$.
- Negative Predictive Value (NPV): $TN / (TN + FN)$.
-
Likelihood Ratios (LR)
- Positive LR (LR+): $Sensitivity / (1 - Specificity)$. For a positive test result.
- LR+ >10 indicates a large and often conclusive increase in the likelihood of disease.
- Negative LR (LR-): $(1 - Sensitivity) / Specificity$. For a negative test result.
- LR- <0.1 indicates a large and often conclusive decrease in the likelihood of disease.
- Positive LR (LR+): $Sensitivity / (1 - Specificity)$. For a positive test result.
⭐ Prevalence Effect: Sensitivity and Specificity are intrinsic to the test and do not change with prevalence. However, PPV is directly proportional to prevalence, while NPV is inversely proportional.
Screening Biases - Tricky Data Traps
-
Lead-Time Bias: Apparent increase in survival time just from detecting a disease earlier. The disease's natural history and outcome are not actually changed. Think of it as starting the survival clock sooner.
-
Length-Time Bias: Screening is more likely to detect slow-growing, less aggressive diseases. Fast-growing, aggressive diseases often cause symptoms between screening intervals. This makes the screening program look better than it is.
-
Selection (Volunteer) Bias: People who choose to get screened are often healthier and more health-conscious than those who don't. Their better outcomes might be due to their healthier lifestyle, not the screening.
⭐ The gold standard to overcome these biases is a Randomized Controlled Trial (RCT) demonstrating a reduction in disease-specific mortality.
Evidence Grading - The USPSTF Report Card
The USPSTF assigns a letter grade based on the strength of evidence and the balance of benefits and harms of a preventive service.
- Grade A: High certainty of substantial net benefit.
- Action: Offer or provide this service.
- Grade B: High certainty of moderate or moderate certainty of substantial net benefit.
- Action: Offer or provide this service.
- Grade C: Moderate certainty of small net benefit.
- Action: Offer selectively to individual patients.
- Grade D: Moderate/High certainty of no net benefit or that harms outweigh benefits.
- Action: Discourage use.
- I Statement: Evidence is insufficient, conflicting, or poor quality.
- Action: Clinical judgment; shared decision-making.
⭐ For the exam, services with grades A and B are generally recommended and should be offered.
High‑Yield Points - ⚡ Biggest Takeaways
- USPSTF Grades are crucial: Grade A/B means offer the service. Grade D means discourage use. Grade I indicates insufficient evidence.
- Screening studies are prone to lead-time bias (earlier detection without changing outcome) and length-time bias (detecting more indolent cases).
- Overdiagnosis is a critical harm of screening-detecting disease that would never have become clinically significant.
- The Number Needed to Screen (NNS) is a key metric for evaluating the efficiency of a screening program.
Continue reading on Oncourse
Sign up for free to access the full lesson, plus unlimited questions, flashcards, AI-powered notes, and more.
CONTINUE READING — FREEor get the app