Why the Same Test Result Means Different Things in Different Patients

The same positive test can mean near-certainty in one patient and almost nothing in another, because a result does not stand alone. It revises an estimate you already carried into the room. How much it revises depends on the strength of the result, and where it lands depends on how likely the diagnosis was before you ever ordered the test. Read a result without that starting estimate and you are reading half the story.

Key points#

• A test result updates a prior belief. It does not replace it.
• The likelihood ratio measures how strongly a given result should shift your estimate.
• Where you end up (the post-test probability) depends on where you started (the pretest probability).
• The two situations most likely to fool you are a positive in a low-risk patient and a negative in a high-risk one.

Start with two patients, one result#

Picture two people who each hand you an identical positive result on the same test. The first has a textbook history, several risk factors, and walked into a setting where the condition is common. The second is a worried but otherwise well person with an atypical story, tested almost as an afterthought. The printout is the same. The meaning is not.

For the first patient, that positive pushes the odds close to certainty. For the second, the same positive may leave the diagnosis still more unlikely than not. Nothing about the test changed between them. What changed was the ground each patient was standing on when the result arrived. This is the single most useful idea in diagnostic reasoning, and everything below is machinery for making it precise.

A result is an update, not a light switch#

Many of us are trained, at least implicitly, to read tests as switches: positive means disease, negative means health. That habit produces two reliable mistakes, and they mirror each other. It sends clinicians chasing false alarms in low-risk people, and it lets a negative result falsely reassure in someone whose story should have kept the search alive.

The more faithful model is Bayesian. You begin with a pretest probability, an honest estimate of how likely the diagnosis is based on history, examination, and the clinical setting. The test then moves that estimate up or down, and what you are left with is the post-test probability. As Akobeng explains in a widely used primer on diagnostic tests, this is the logic clinicians already apply whether or not they name it: a result combines with a prior probability to produce a new probability, never a bare yes or no.

The likelihood ratio: one number for the size of the nudge#

Sensitivity and specificity describe a test in the abstract, but they answer the question backwards. They tell you how the test behaves given disease status, when what you actually hold in your hand is a result and a question about disease status. Likelihood ratios turn the question the right way around and compress sensitivity and specificity into a single figure for each result.

A positive likelihood ratio asks how many times more often a positive result shows up in people who have the condition than in people who do not. A negative likelihood ratio does the same for a negative result. Akobeng frames it plainly: the likelihood ratio tells you how many times more, or less, likely a particular result is in patients with disease than in those without.

The size of that number has a rough grammar, described by Jaeschke and colleagues and echoed by the Centre for Evidence-Based Medicine at Oxford:

• Above 10 (or below 0.1): a large shift, often enough to rule a condition in or out.
• 5 to 10 (or 0.1 to 0.2): a moderate, useful shift.
• 2 to 5 (or 0.2 to 0.5): a small nudge.
• 0.5 to 2: barely moves the estimate, a courteous way of saying the result answered almost nothing.
• Exactly 1: no change at all.

Notice what is missing from that list. None of these thresholds mention the patient. All of the patient lives in the pretest probability, which is precisely why the same ratio can carry very different weight from one person to the next.

The arithmetic runs through odds#

To combine the two numbers you pass briefly through odds. You convert the pretest probability into odds, multiply by the likelihood ratio, then convert back into a probability. The CEBM resource states the chain cleanly: post-test odds equal pretest odds multiplied by the likelihood ratio.

Walk it through both patients from the opening. Give each a strong positive result. The high-risk patient started with high pretest odds, so multiplying by a large ratio lands close to certainty. The low-risk patient started with tiny odds, so the same multiplication lifts the estimate to something still modest, sometimes too modest to justify acting. The test performed identically. The starting odds decided the destination.

The mirror image explains why negatives reassure unevenly. A strong negative result can nearly exclude a condition in someone with only modest risk, yet leave a meaningful residual probability in someone who was high-risk from the start. That is the classic trap: a negative test in a patient whose story keeps insisting on the diagnosis. The math says keep looking.

Skipping the algebra: the Fagan nomogram#

Few clinicians want to convert odds by hand during a full clinic. In 1975, Terry Fagan published a graphical shortcut, now called the Fagan nomogram, that reduces the whole calculation to a straightedge and three scales. Mark the pretest probability on the left scale, the likelihood ratio on the middle scale, draw a line through both, and read the post-test probability where that line crosses the right 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 since. In a commentary on its evolution, Abushouk notes that the original required awkward back-and-forth between odds and probabilities, and describes a modernized layout that sets parallel probability and odds scales side by side and spreads out the high likelihood ratios, which matters most for rare conditions with very low starting probabilities. The worked example there is memorable because it is so undramatic: a pretest probability of 18 percent combined with a positive likelihood ratio of 2.8 yields a post-test probability near 38 percent. The result was positive, and the diagnosis was still more likely absent than present. The modest starting point decided it.

Three habits that follow#

First, estimate the pretest probability before the result can anchor you, so the number does the updating rather than distorting your judgment. Second, reach for the likelihood ratio, or reconstruct it, instead of settling for the word positive or negative, because both the direction and the strength matter. Third, give surprising results the skepticism their priors have earned. A positive in a very low-risk person and a negative in a very high-risk person are the two results most likely to mislead, and they are exactly the two the arithmetic flags on its own.

None of this replaces clinical judgment. It disciplines it, and it puts words to something seasoned diagnosticians already feel: the meaning of a result is inseparable from the patient it belongs to.