Sensitivity, Specificity, and Likelihood Ratios: A Quick Guide

Every orthopaedic special test comes with numbers attached. Sensitivity and specificity are the familiar ones, but likelihood ratios are the better way to judge whether a test can actually rule a condition in or out. Here's a short guide, with quick reference tables at the end.

Sensitivity and specificity

Sensitivity is the percentage of people with the condition who test positive. A highly sensitive test rarely misses the condition.

Specificity is the percentage of people without the condition who test negative. A highly specific test rarely flags someone who doesn't have it.

This leads to two well-known rules of thumb:

• SnNout: with a highly sensitive test, a negative result helps rule the condition out.

• SpPin: with a highly specific test, a positive result helps rule the condition in.

Likelihood ratios: the better check

SnNout and SpPin each look at only one number. A negative result is only reassuring if it's rare in people who have the condition and common in people who don't, so both numbers matter. Likelihood ratios (LRs) combine them into one figure.

• Positive LR: how many times more likely a positive result is in someone with the condition than in someone without it. Values above 1 point toward the condition, but a test usually needs a positive LR of 5 or more to meaningfully help rule it in.

• Negative LR: how likely a negative result is in someone with the condition compared with someone without it. Values below 1 point away from the condition, but a test usually needs a negative LR of 0.2 or less to meaningfully help rule it out.

Here's why it matters. A test with 95% sensitivity looks like a strong rule-out test. If its specificity is only 20%, its negative LR is 0.05 ÷ 0.20 = 0.25, which barely moves the needle. Raise specificity to 50% and the negative LR drops to 0.10, a strong rule-out. Same sensitivity, very different value.

Limitations of likelihood ratios

• One LR covers one direction. A strong positive LR says nothing about how useful a negative result is, and vice versa.

• Precision matters. Small studies give wide confidence intervals. A negative LR of 0.1 with an interval reaching 0.6 isn't a reliable rule-out.

• The study population matters. LRs from surgical candidates may not hold for patients walking into a general clinic.

• Your starting point matters. An LR adjusts your pre-test suspicion; it doesn't replace it. A negative LR of 0.3 takes 60% suspicion down to about 31%, which rules nothing out, but takes 10% down to about 3%.

• Cutoffs are conventions. An LR of 4.8 isn't meaningfully worse than 5.1.

Quick reference

How strong is the shift? (Jaeschke et al., 1994)

References

• Jaeschke R, Guyatt GH, Sackett DL. Users' guides to the medical literature. III. How to use an article about a diagnostic test. B. What are the results and will they help me in caring for my patients? JAMA. 1994;271(9):703–707.

• McGee S. Simplifying likelihood ratios. J Gen Intern Med. 2002;17(8):646–649.