Why No Test Is Perfect

A positive result is not automatically a diagnosis, and a negative result does not always exclude disease. Test accuracy has several parts. Understanding sensitivity, specificity, predictive values and pre test probability explains why the same result can mean different things in different situations.

A test answers a defined question A diagnostic test is used to assess a target condition, such as a disease, infection, risk state or treatment response. The test being evaluated is sometimes called the index test. To study accuracy, the index test is compared with the best available method for deciding whether the target condition is present. This is called the reference standard. A reference standard may combine tests, expert assessment and follow up. It is the best available comparison, not necessarily a perfect source of truth. Four possible classifications A true positive means the test is positive and the target condition is present. A true negative means the test is negative and the condition is absent. A false positive means the test is positive but the target condition is absent. A false negative means the test is negative but the condition is present. These terms describe classification against the target condition and reference standard. They are not the same as an out of range blood result, an incidental scan finding or a damaged specimen. Sensitivity and specificity ask group level questions Sensitivity is the proportion of people with the target condition who test positive. High sensitivity means fewer false negatives among people who have the condition. Specificity is the proportion of people without the target condition who test negative. High specificity means fewer false positives among people who do not have it. Sensitivity does not directly answer, "If my result is negative, what is the chance I am disease free?" Specificity does not directly answer, "If my result is positive, what is the chance I have the condition?" Those individual facing questions require predictive values or an updated post test probability. A threshold can create a trade off Some tests produce a number or score and use a threshold to classify the result as positive or negative. Changing that threshold can change sensitivity and specificity. Lowering a threshold often identifies more people with the condition, increasing sensitivity. It may also classify more people without the condition as positive, reducing specificity. Raising the threshold often has the opposite effect. This trade off applies when the threshold of the same test is changed. It is not a rule that every improvement in sensitivity must worsen specificity. A better test or better method may improve both. Not every result is simply positive or negative. Tests can be borderline, indeterminate, technically inadequate or reported as a continuous value. Pre test probability changes what a result means Pre test probability is the estimated chance that the target condition is present before the result is known. It reflects the clinical setting, prevalence, symptoms, risk factors and previous findings. Positive predictive value is the proportion of positive results that are true positives. Negative predictive value is the proportion of negative results that are true negatives. Predictive values depend strongly on how common the condition is in the tested population. When a condition is uncommon, false positives from the much larger unaffected group can outnumber true positives. Consider 1,000 people in a setting where 1% have the condition. A test with 90% sensitivity and 95% specificity would identify about 9 true positives and miss 1 case. Among the 990 people without the condition, it would produce about 50 false positives. Around 59 people would therefore test positive, but only about 9 would truly have the condition. This example is illustrative, but it shows why a positive result can need confirmation even when sensitivity and specificity sound high. Results update probability rather than creating certainty A test result changes the estimated probability of a condition. The new estimate is called the post test probability. Likelihood ratios summarise how much a positive or negative result tends to shift the odds. A useful result can produce a large change, but most results do not move probability exactly to zero or 100%. The simple rules that a sensitive test rules out and a specific test rules in are only memory aids. They are safe only when the test, threshold, population, timing and clinical pathway support that conclusion. Accuracy is not one fixed property Published sensitivity and specificity are estimates, usually reported with confidence intervals. They apply to the population, test version, threshold and reference standard that were studied. Performance can change with disease stage or severity, specimen type, collection quality, timing, treatment, operator skill, equipment and interpretation. The same named test may therefore perform differently in another setting. A test validated in specialist clinics may not produce the same predictive value when used for low risk population screening. A result obtained too early or from a poor specimen may also be less reliable. Repeat and confirmatory testing have different purposes Repeating the same test can help when random variation, timing or specimen quality may have affected the result. It does not guarantee that the second result is correct. A different confirmatory test may provide more independent evidence or assess the target condition through another method. Some diagnostic pathways use several results together rather than relying on one test. The follow up test is not always simply more specific. The correct sequence depends on the condition, consequences of error and available reference standard. Testing should improve a decision False positives can cause anxiety, further procedures and unnecessary treatment. False negatives can delay diagnosis or create false reassurance. Borderline results can also prolong uncertainty. A test is most useful when the result can change a clinical decision. When the result would not alter the plan, testing may add burden without enough benefit. If a healthcare service tells you that a result needs urgent action, follow its instructions immediately. Do not wait for a test result if you have severe breathing difficulty, tight or spreading chest pain, collapse with abnormal responsiveness, or sudden face weakness, arm weakness or speech difficulty. Call 999. This lesson explains general diagnostic accuracy. It cannot calculate the meaning of an individual result or replace condition specific clinical assessment.

Sensitivity and specificity describe how a test classifies groups. The meaning of one person's result also depends on pre test probability, predictive value, threshold, timing, specimen quality and the clinical pathway.