FROM MY PT SCHOOL NOTES · EVIDENCE-BASED PHYSICAL THERAPY · CHAPTER 8
What Do Sensitivity, Specificity and Likelihood Ratios Mean?
Diagnostic accuracy matters most when students can move from a test result to a reasoned change in probability.
Start with a 2-by-2 table
The lecture distinguishes diagnostic classification from clinical measures used to quantify impairments or function. Examples include special tests, imaging or laboratory results, ROM measured by goniometry, and functional measures. Whether a test is useful depends on its question, population, practicality and potential effect on care.
Diagnostic test studies compare an index test with a suitable reference standard. The course asks whether participants represent the patients we see, whether the reference standard is appropriate and consistently applied, and whether investigators know other test results.
Understand probability and likelihood ratios
A 2×2 diagnostic table defines true positives (TP), false positives (FP), false negatives (FN) and true negatives (TN). Sensitivity = TP/(TP+FN); specificity = TN/(TN+FP). Positive predictive value = TP/(TP+FP); negative predictive value = TN/(TN+FN). The original class cheat sheet presents these same calculations.
Likelihood ratios combine test sensitivity and specificity: LR+ = sensitivity/(1−specificity); LR− = (1−sensitivity)/specificity. They can adjust an estimated pretest probability toward a post-test probability. The lecture also introduces receiver operating characteristic (ROC) curves and diagnostic/test thresholds.
Worked example: 100 people, including 10 with the target condition. The test correctly identifies 8 of those 10, but incorrectly flags 9 of the 90 without the condition.
| Test result | Condition present | Condition absent |
|---|---|---|
| Positive | 8 true positives | 9 false positives |
| Negative | 2 false negatives | 81 true negatives |
Interpretation: Sensitivity = 8/10 = 80%; specificity = 81/90 = 90%. But positive predictive value = 8/17 ≈ 47% in this sample. A positive result is not an 80% chance of disease. Pretest probability and the population matter.
Apply a result cautiously
Class definitions from the scanned diagnostic glossary also distinguish pretest probability, posttest probability, reference standard, index test, and test/treatment thresholds.
Check your reasoning: Is a positive predictive value fixed for a test across every population?
No. Predictive values depend on prevalence or pretest probability in the population being considered.
Updated-standard note
The older terminology sheet uses ‘odds (probability)’ as if they were interchangeable. They are different: odds = probability/(1−probability). Updating with a likelihood ratio uses pretest odds × LR = post-test odds; convert odds back to probability for a probability estimate. This is an interpretation note—not a separate diagnosis rule.
Put it into practice
Scenario: In a hypothetical sample of 100 people, 10 have the target disease. A test identifies eight of those ten and incorrectly tests positive in nine of the ninety without disease.
What should you conclude?
Reasoning: Sensitivity is 8/10 = 80%; specificity is 81/90 = 90%; and positive predictive value is 8/17, roughly 47% in this particular sample. Pretest probability affects how a positive result should be understood.
Common trap: Confusing a test’s sensitivity with the probability of disease after a positive result. Remember: Always identify the population and build the 2×2 table.
Course basis: my Evidence-Based PT school notes, source lecture Chapter 11. New questions and headings are teaching aids, not official exam items. For full definitions and examples, use the class notes and assigned readings.
