Diagnostic testing—decision utility depends on pretest probability: a test is valuable only when its result can change a clinical decision more effectively than examination, observation, treatment, or no testing. Pretest probability is the estimated chance that a patient has a condition before testing, based on symptoms, history, examination, risk factors, and disease prevalence. Bayes’ theorem shows why the same result can mean different things in different patients: a highly sensitive and specific test may produce mostly false positives when disease is unlikely, while a negative result may be insufficiently reassuring when disease is very likely. The World Health Organization, the U.S. Preventive Services Task Force, and the Choosing Wisely campaign all support matching testing to clinical questions, expected benefits, potential harms, and the actions that follow.

Determining Diagnostic Testing—Decision Utility Through Pretest Probability

Diagnostic testing–decision utility is the relationship between the probability of a condition before testing, the information produced by a test, and the consequences of acting on the result. In the classic threshold model developed by Howard Pauker and Jerome Kassirer, testing is useful when the pretest probability lies between two thresholds: below the testing threshold, the condition is unlikely enough that testing or treatment offers little benefit; above the treatment threshold, the condition is likely enough that treatment or further action may be justified without waiting for a test.

The central characteristic is not whether a test is available or technically accurate. It is whether the result will cross a decision threshold. A test that leaves management unchanged has low practical value, even when its analytic performance is excellent. Conversely, a modest test can be useful if it separates patients who should receive treatment, additional evaluation, or reassurance.

Pretest probability as the starting estimate

Pretest probability is an estimate, not a guess made after seeing the test result. Clinicians derive it from the patient’s presentation and from epidemiologic data. A compatible symptom pattern, a known exposure, a family history, or an abnormal physical finding can raise it; an alternative diagnosis, a low-risk setting, or an absent hallmark feature can lower it. Structured clinical prediction rules, such as the Wells score for pulmonary embolism or the Centor and McIsaac scores for streptococcal pharyngitis, formalize some of these inputs.

Disease prevalence is often a useful starting point, but prevalence is not the entire pretest probability. The prevalence of pulmonary embolism in an emergency department population differs from that in primary care, and the prevalence of a genetic disorder in a selected family differs from that in the general population. Applying a test without accounting for this context transfers a population estimate into the wrong patient group.

Likelihood ratios and post-test probability

Likelihood ratios translate a test result into a change in probability. A positive likelihood ratio indicates how much more likely a positive result is among people with the condition than among those without it; a negative likelihood ratio describes the corresponding effect of a negative result. The result is most informative when the likelihood ratio moves the patient across a treatment or exclusion threshold.

A simple example illustrates the problem with reflexive testing. Suppose a test has 90 percent sensitivity and 95 percent specificity. If the pretest probability is 1 percent, testing 10,000 people would identify about 90 true positives but produce about 495 false positives, giving a positive predictive value of approximately 15 percent. If the pretest probability is 50 percent, the same test would have a positive predictive value of approximately 95 percent. Test accuracy did not change; the meaning of the result did.

Figure 1 could display two identical test-performance curves with different baseline prevalences, showing how positive predictive value rises as pretest probability increases. A probability-based graph is often more informative for decision-making than a sensitivity-and-specificity table alone.

Choosing Diagnostic Testing—Decision Utility by Clinical Purpose

Once pretest probability has been estimated, the next question is what kind of decision the test is intended to support. The major hyponyms include screening tests, diagnostic tests, rule-out tests, rule-in tests, confirmatory tests, and monitoring tests. These categories overlap, but each has a different acceptable balance of false positives, false negatives, cost, delay, and downstream intervention.

Screening tests for people without symptoms

Screening tests are applied to people who do not have a clinical diagnosis and may have a relatively low pretest probability. Because low prevalence increases the proportion of false positives, screening programs need a clearly defined target population, an effective confirmatory pathway, and evidence that earlier detection improves outcomes.

The U.S. Preventive Services Task Force illustrates this principle in its 2024 breast cancer screening recommendation. Its modeling estimated that biennial mammography from ages 40 through 74 could prevent approximately 1.3 breast cancer deaths per 1,000 women screened from ages 40 to 49 and approximately 1.8 deaths per 1,000 women screened from ages 50 to 59, while also producing false-positive results and additional imaging. The relevant decision is therefore not simply whether mammography can detect cancer; it is whether the expected benefit outweighs the harms for a defined age and risk group.

Rule-out and rule-in tests

A rule-out test is designed to reduce probability below a threshold at which further investigation or treatment is unnecessary. Such a test generally requires a sufficiently low negative likelihood ratio and should be used only after the patient begins in a low or intermediate probability range. A negative D-dimer result, for example, can help exclude pulmonary embolism in appropriately selected patients with low or intermediate clinical probability, but it is not a safe substitute for imaging in a patient with high clinical probability.

A rule-in test is intended to raise probability above a treatment or confirmatory threshold. It generally requires a high positive likelihood ratio. A positive result from a low-specificity test may be inadequate for ruling in disease, especially when the starting probability is low. The American College of Physicians and the American College of Emergency Physicians emphasize that clinical assessment should determine which patients are suitable for these pathways.

Confirmatory and monitoring tests

Confirmatory tests are used when an initial result, symptom pattern, or examination finding has raised probability but does not yet justify a consequential diagnosis or treatment. Their value comes from resolving uncertainty before an irreversible action. Examples include tissue sampling after suspicious imaging or a second, more specific assay after an initial positive infectious-disease screen.

Monitoring tests answer a different question: how has a known condition or treatment response changed over time? Repeating a test at a fixed interval can be useful when results are expected to change management, but serial testing can also amplify random variation and create apparent abnormalities. The American Society of Clinical Oncology’s Choosing Wisely recommendations caution against routine surveillance imaging and laboratory testing when symptoms and examination do not suggest recurrence and when earlier detection has not been shown to improve outcomes.

Avoiding Reflexive Diagnostic Testing—Decision Utility and Unintended Harm

False positives, incidental findings, and cascades

A false-positive result can trigger repeat testing, specialist referrals, invasive procedures, anxiety, unnecessary treatment, and financial costs. Imaging may reveal incidental findings that are unrelated to the presenting problem but lead to months or years of surveillance. These downstream effects are part of the test’s real risk and should be included in the pretest discussion.

The American College of Radiology’s Incidental Findings Committee has developed management recommendations precisely because incidental findings are common and vary in clinical importance. Finding an abnormality is not equivalent to finding a disease that benefits from intervention. The more sensitive the test and the lower the pretest probability, the more important this distinction becomes.

False negatives, overconfidence, and missed disease

Avoiding unnecessary tests does not mean ignoring serious disease. A negative result can be falsely reassuring when the pretest probability is high, when the disease is early, or when specimen quality and test timing are poor. The appropriate response may be repeat testing, a different test, observation with explicit safety-net instructions, or immediate treatment based on clinical probability.

The National Institute for Health and Care Excellence recommends using clinical assessment alongside test results rather than treating a laboratory value as an isolated verdict. A sound plan states what result would change management, what result would not, and what will happen if symptoms progress despite a negative test.

The five-question pretest checklist

Before ordering a diagnostic test, clinicians and patients can ask:

• What condition is the test intended to evaluate?
• What is the estimated pretest probability in this patient?
• What result would meaningfully change treatment, referral, observation, or reassurance?
• What are the likely harms of false-positive, false-negative, incidental, or borderline results?
• Is there a safer or more informative alternative, including time, examination, or watchful waiting?

This checklist converts “Can we order the test?” into the more useful question, “What decision will the test improve?” It also supports shared decision-making because patients can weigh the expected benefit against inconvenience, anxiety, cost, radiation, invasive follow-up, and treatment effects.

Applying Diagnostic Testing—Decision Utility in Practice

Consider a patient with a low-risk presentation for pulmonary embolism. If clinical assessment places the patient below the testing threshold, ordering a D-dimer may create a false-positive result and lead to unnecessary computed tomography. If the patient is in an intermediate range, a negative D-dimer may safely avoid imaging. If the patient is at high probability, the same D-dimer is unlikely to settle the decision, and imaging or treatment may be more appropriate. The test is not intrinsically good or bad; its value changes with the starting probability and the decision threshold.

A similar logic applies to respiratory viral testing, thyroid function testing, cardiac biomarkers, genetic panels, and full-body imaging. Broad panels increase the number of opportunities for abnormal results unrelated to the patient’s problem. Ordering a focused test after defining the clinical question usually produces more actionable information than ordering many tests “just to be safe.”

Health systems can reinforce this behavior through electronic-ordering prompts, reflexive-test restrictions, evidence-based pathways, audit and feedback, and clear follow-up ownership. Such interventions should be evaluated for both underuse and overuse: reducing low-value testing is beneficial only if patients with meaningful risk are still identified promptly.

Conclusion: Diagnostic Testing—Decision Utility Requires Probability-Based Action

Diagnostic testing—decision utility is governed by pretest probability, likelihood ratios, decision thresholds, and the consequences of acting on results. Screening, rule-out, rule-in, confirmatory, and monitoring tests each require a different interpretation. The same test can be useful in one patient and harmful in another because predictive value changes with baseline risk.

The practical call to action is simple: estimate probability before ordering, define the action that each possible result would trigger, and include downstream harms in the decision. Clinicians, patients, laboratories, and health systems should use evidence-based pathways and shared decision-making to replace reflexive testing with purposeful testing. Further reading from the U.S. Preventive Services Task Force, Choosing Wisely, the World Health Organization, and evidence-based clinical guideline organizations can help translate this framework into specialty-specific practice.

Sources: Pauker SG and Kassirer JP, “The Threshold Approach to Clinical Decision Making,” New England Journal of Medicine, https://www.nejm.org/doi/full/10.1056/NEJM197509252931307; U.S. Preventive Services Task Force, “Breast Cancer: Screening,” https://www.uspreventiveservicestaskforce.org/uspstf/recommendation/breast-cancer-screening; Choosing Wisely, American Society of Clinical Oncology Recommendations, https://www.choosingwisely.org/societies/american-society-of-clinical-oncology/; World Health Organization, Guide to Cancer Early Diagnosis, https://www.who.int/publications/i/item/9789241511940; American College of Physicians, “Diagnostic Testing for Suspected Pulmonary Embolism,” https://www.acpjournals.org/doi/10.7326/M20-2999; National Institute for Health and Care Excellence, “Suspected Sepsis: Recognition, Diagnosis and Early Management,” https://www.nice.org.uk/guidance/ng51; American College of Radiology, Incidental Findings, https://www.acr.org/Clinical-Resources/Incidental-Findings