Why Pretest Probability Decides What a Test Result Means

A test result is not a verdict. It is an update to what you already believed. A likelihood ratio tells you how strongly a given result should move your estimate of disease, but the place you end up depends on the place you started. That starting point is the pretest probability, and it is the reason a single positive result can mean near-certainty in one patient and almost nothing in another. Reading a test well means holding both numbers at once: the strength of the result and the risk of the person in front of you.

The result updates a belief you already hold

Most people are taught to read a test as a light switch. Positive means disease, negative means health. Diagnostic reasoning does not work that way, and treating it as though it does produces two predictable errors: chasing false alarms in low-risk people and falsely reassuring high-risk people who test negative.

The better model is Bayesian. You begin with an estimate of how likely disease is before testing, based on the patient's history, exam, and the setting. The test then revises that estimate up or down. The revised number is the post-test probability. As Akobeng describes in a widely used primer on diagnostic tests, this is exactly the logic clinicians apply, whether or not they name it: a test combines with a prior probability to yield a new probability, not a yes-or-no answer.

The tool that makes this concrete is the likelihood ratio.

Likelihood ratios: one number for how much a result moves you

Sensitivity and specificity describe a test in the abstract, but they are awkward at the bedside because they answer a backwards question: given disease status, how does the test behave. What a clinician actually faces is the reverse: given this result, how does disease status change. Likelihood ratios bridge that gap by folding sensitivity and specificity into a single figure per result.

A positive likelihood ratio asks how many times more often a positive result occurs in people with the disease than in people without it. A negative likelihood ratio does the same for a negative result. Akobeng frames the likelihood ratio as the measure that tells you how many times more, or less, likely a particular result is in diseased versus non-diseased patients.

The magnitude has a rough grammar, drawn from work by Jaeschke and colleagues and echoed by the Centre for Evidence-Based Medicine at Oxford. A likelihood ratio above 10, or below 0.1, shifts probability a large amount and can effectively rule a condition in or out. Values of 5 to 10, or 0.1 to 0.2, produce moderate shifts. Values of 2 to 5, or 0.2 to 0.5, nudge only a little. Anything between 0.5 and 2 barely moves the needle, a polite way of saying the result was nearly useless for that question. A ratio of 1 changes nothing.

Notice what these thresholds do not contain: any mention of the patient. That information lives entirely in the pretest probability, which is why the same ratio can carry very different weight.

The same ratio, two different patients

The arithmetic runs through odds. You convert pretest probability to odds, multiply by the likelihood ratio, then convert back to a probability. The CEBM resource lays out this chain plainly: post-test odds equal pretest odds times the likelihood ratio, and you translate the result back into a probability at the end.

The consequence is that identical results diverge. Take a positive test with a strong positive likelihood ratio. In a patient whose pretest probability is high, say a classic presentation in a high-prevalence setting, that positive pushes the post-test probability close to certainty. In a patient whose pretest probability is low, perhaps a screening context or an atypical story, the same positive lands at a far more modest number, sometimes low enough that acting on it would be premature. The test did its job identically in both. The patients were standing in different places when the result arrived.

This is also why a negative result reassures unevenly. A strong negative likelihood ratio can nearly exclude disease in someone with modest pretest risk, yet leave meaningful residual probability in someone who was high-risk to begin with. The famous failure mode here is the negative test in a patient whose story screams disease. Bayes says: do not stop looking.

The Fagan nomogram: doing this without algebra

Few clinicians want to run odds conversions during a busy clinic. In 1975, Terry Fagan published a graphical shortcut, now called the Fagan nomogram, that turns the calculation into a straightedge and three scales. You mark the pretest probability on the left, the likelihood ratio in the middle, draw a line through both, and read the post-test probability where it crosses the right-hand scale. Akobeng recommends it as a convenient way to combine a likelihood ratio with a patient's pretest probability at the point of care.

The tool has been refined over time. In a commentary on its evolution, Abushouk notes that the original nomogram required awkward conversions between odds and probabilities, and describes a modernized version that places parallel probability and odds scales side by side and better spreads out high likelihood ratio values, which matters most for rare conditions with very low pretest probabilities. The worked example in that commentary is instructive precisely because it is undramatic: a pretest probability of 18 percent with a positive likelihood ratio of 2.8 yields a post-test probability of about 38 percent. A positive result, and still more likely than not that the patient does not have the disease. The number that decided this was the modest starting point.

What this changes about reading a report

Three habits follow. First, estimate pretest probability before you look at the result, or at least before you let it anchor you, so the number does the updating rather than the anchoring. Second, ask for the likelihood ratio, or reconstruct it, rather than settling for positive versus negative; the direction and the strength both matter. Third, treat surprising results with the skepticism their priors deserve. A positive in a very low-risk person and a negative in a very high-risk person are the two situations most likely to mislead, and they are the two the math flags automatically.

None of this replaces clinical judgment. It disciplines it. The value of the Bayesian frame is that it makes explicit what experienced diagnosticians already sense: the meaning of a result is inseparable from the patient it belongs to.

This article is educational and not medical advice.