Overdiagnosis: when too much diagnosis harms

The report arrives with a name. The image shows a nodule. The lab crosses a threshold. In minutes the patient stops being "someone who felt fine" and starts carrying a diagnosis. Sometimes that label saves. Sometimes it names a condition that would never have bothered that person in their life. When the diagnosis is true in the lab or on the slide, but useless or harmful in the patient's biography, the literature calls this overdiagnosis.

NEXO is a daily medical diagnosis game: one clinical case a day, six progressive clues, and ICD-10 proximity feedback across five levels.

This text explains the concept, separates overdiagnosis from false positive and from overtesting, shows how screening, expanded criteria and incidentalomas feed the phenomenon, summarizes what the evidence recommends for talking with the patient and training reasoning, and closes with a table, FAQ and checklist. Anyone who has read pretest probability learned to weigh the test when the hypothesis already exists; here the focus is different: the label that is born when nobody asked for a problem. Those who use clinical pearls gain recognition shortcuts; overdiagnosis calls for the opposite move: reining in the urge to name too much.

Key points

• Overdiagnosis is diagnosing a condition that, if it had not been recognized, would not cause symptoms or harm in the person's life (Welch and Black, 2010; Brodersen et al., 2018).
• It is not a false positive: the finding is "true" on the test; the problem is clinical relevance (Kale and Korenstein, 2018).
• Screening, lower diagnostic thresholds and incidentalomas on imaging exams are three classic drivers (Moynihan et al., 2012; Hoffman and Cooper, 2023).
• The harm runs through unnecessary treatment, psychological labeling, a cascade of tests and opportunity cost (Brodersen et al., 2018).
• Training means practicing the question "does this label change management in a way that is worth the risk?" before closing the case, in the same spirit as the diagnostic pause.

What overdiagnosis is

The most cited definition in cancer: diagnosing a "cancer" that would otherwise not cause symptoms or death (Welch and Black, 2010). In primary care the formulation generalizes: diagnosing a condition that, if it had not been recognized, would not cause symptoms or harm over a lifetime (Kale and Korenstein, 2018). Brodersen and colleagues stress that the phenomenon only makes sense from a population and temporal perspective: in the individual, it is almost never possible to know, on the day of the report, whether that finding would be harmless (Brodersen et al., 2018).

There are two classic prerequisites in Welch's model: a silent reservoir of disease (or of "pseudo-disease") and some activity that detects it, especially screening (Welch and Black, 2010). Without a reservoir, there is nothing to overdiagnose. Without active detection, the reservoir stays invisible and the patient goes on with life without the label.

Disease, indolent disease and label

Not every pathology progresses. Some tumors detected by screening grow so slowly that the person dies of another cause; some regress; some would never leave the subclinical state (Brodersen et al., 2014). The same reasoning appears outside cancer when thresholds for blood pressure, blood glucose or kidney function include people who would not have a relevant clinical event within their life horizon (Moynihan et al., 2012).

The label matters because medicine is not just description: it is permission to treat, to monitor, to warn the family and to change the identity of the person who receives the name. Overdiagnosis is the point where the correct description generates intervention or suffering without proportional benefit.

Why the topic matters in clinical reasoning

The student trained to "miss nothing" learns to broaden hypotheses and order tests. This is virtuous when the pretest hypothesis justifies the test, as in pretest probability and in base rate. It becomes a problem when the culture of "better to investigate" meets a reservoir of findings without relevance. The brake is not laziness: it is part of reasoning, alongside the diagnostic pause and care with obscuring (a label that hides another problem is the mirror image of the label that is left over).

What overdiagnosis is not

Confusing terms delays the clinical conversation and research. Three distinctions matter.

It is not a false positive

In a false positive the test says "disease" and the person does not have the condition according to the gold standard. In overdiagnosis the gold standard also confirms the condition; what is missing is the net benefit of knowing (Kale and Korenstein, 2018; Hoffman and Cooper, 2023). Treating the two as synonyms leads to wrong solutions: improving test specificity does not solve overdiagnosis if the "true positive" is clinically irrelevant.

It is not just overtesting

Overtesting is ordering too many tests. It can lead to overdiagnosis, but it also leads to false positives, anxiety and cost without reaching a stable label. Overdiagnosis is the outcome in which the label becomes fixed without clinical usefulness (Moynihan et al., 2012).

It is not a random naming error

Nor is it "any diagnosis you disagree with". The literature requires a logic: without the diagnosis, the natural course would be benign over the life horizon. That is why the best estimates come from screening trials, autopsy series and sudden changes in incidence without a corresponding drop in mortality (Welch and Black, 2010).

Concept	The test / the finding	Does the condition exist?	Central problem
False positive	Positive	No (by the gold standard)	Test error
Overdiagnosis	Positive (or criterion met)	Yes	No net benefit of knowing
Overtesting	Test ordered in excess	Variable	Cascade and noise
Underdiagnosis	Negative or not done	Yes and relevant	Missed opportunity to treat

Three drivers: screening, threshold and incidentaloma

Screening in people who feel well

Screening invites asymptomatic people into the health system. The implicit hypothesis is that finding things early always helps. Brodersen, Schwartz and Woloshin point out that this hypothesis fails when the pathology is indolent: the test turns a silent finding into "disease" and the person starts living as a patient (Brodersen et al., 2014). Welch and Black estimated, from trials, relevant fractions of overdiagnosis in cancers detected by mammography, chest X-ray or PSA (Welch and Black, 2010). The exact numbers vary by population and protocol; the teaching point is the direction of the effect, not memorizing a percentage.

In the conversation with the patient, this changes the script: instead of "let's just make sure", the clinician describes the trade-off between finding serious disease early and finding disease that did not need a name (Welch and Black, 2010).

Thresholds that drop and definitions that expand

When the cutoff for a blood pressure, a blood glucose or a filtration rate drops, thousands of people cross the line and receive a diagnosis. Some may benefit; some only gain a label, surveillance and treatment (Moynihan et al., 2012; Hoffman and Cooper, 2023). Moynihan and colleagues described this movement as the expansion of disease definitions, often without proportional assessment of harms (Moynihan et al., 2012).

For the student, the useful exercise is not to memorize the guideline threshold: it is to ask which outcome the threshold was designed to prevent, in whom, and with what number needed to treat. Without that, "being a little above" becomes a clinical identity with no clear plan.

Incidentalomas: the finding nobody asked for

High-resolution imaging finds nodules, cysts and calcifications outside the original clinical question. The incidentaloma forces a decision: ignore with criteria, monitor or investigate. Investigating too much pushes toward biopsies, surgeries and new labels (Hoffman and Cooper, 2023; Kale and Korenstein, 2018). Healthy reasoning ties back to the index question: what was I investigating, and does this finding change that line?

How the harm reaches the patient

Overtreatment and cascade

Once the condition is named, the default path in many clinical cultures is to treat. Surgery, radioiodine therapy, statin, antihypertensive or intensive surveillance come into play. Each step has a possible adverse effect. The diagnostic cascade (one test pulls the next) is the most common intermediate mechanism between the finding and the harm (Kale and Korenstein, 2018).

Label, anxiety and identity

Receiving a disease name changes how a person sleeps, works and sees themselves. Even without invasive treatment, the label can generate anxiety, depression and a self-image of fragility (Brodersen et al., 2018). This is not "drama": it is a measurable clinical outcome in screening studies.

Opportunity cost

Consultation time, imaging queues and the attention of the student or preceptor spent on irrelevant findings stop going to symptoms that need care. In saturated systems, overdiagnosis competes with underdiagnosis of serious diseases (Moynihan et al., 2012).

Confusion with safety

The phrase "better to prevent" sounds ethical. Without harm numerators, it hides a choice of values: preferring the risk of overtreatment to the risk of delaying a diagnosis. Evidence-based medicine asks that both risks be made explicit (Welch and Black, 2010).

What the evidence shows

Cancer and screening

The review by Welch and Black in JNCI consolidated the argument for cancer: a silent reservoir plus active detection generates measurable overdiagnosis in breast, lung and prostate, with signs also in thyroid, melanoma and kidney (Welch and Black, 2010). Davies and Welch documented the jump in thyroid cancer incidence without a proportional drop in mortality, a pattern consistent with large-scale overdiagnosis (Davies and Welch, 2014). Esserman and colleagues proposed reclassifying lesions with indolent behavior to reduce the automatic impulse to treat them as lethal cancer (Esserman et al., 2014).

Primary care and chronic diseases

Kale and Korenstein mapped overdiagnosis beyond cancer: chronic kidney disease from single GFR cutoffs, depression and ADHD with broad criteria, among others (Kale and Korenstein, 2018). Hoffman and Cooper reinforce the trio of screening, threshold and incidentaloma as entry points for harm in everyday clinical practice (Hoffman and Cooper, 2023).

Communication and shared decision

Informing the magnitude and nature of the trade-off has been part of the proposed solution since Welch and Black (Welch and Black, 2010). Brodersen et al. insisted on clarity of terms so that doctors and patients stop mixing false positive with overdiagnosis (Brodersen et al., 2018). Without a common vocabulary, the consultation becomes persuasion disguised as prevention.

Limits of what we know

In the individual, overdiagnosis is generally impossible to prove at the time of the report: only the counterfactual (what would happen without the diagnosis) would decide it. That is why the evidence works with populations, trials and incidence-mortality trends. The clinician translates this into prudence, not into absolute certainties about that nodule.

How to train the diagnostic brake

Five questions before naming

1. What was the original clinical question?
2. Does this finding answer it or is it collateral?
3. If I did not know about this finding, would the patient probably suffer harm from this condition in their lifetime?
4. Does the next step (biopsy, surgery, drug, surveillance) have clear benefit in this risk range?
5. Can I explain the trade-off in everyday language without hiding uncertainty?

These questions connect with the diagnostic pause and with the discipline of not letting a label obscure the rest of the picture.

Separate "finding" from "disease that calls for action"

Practice writing in the progress note: "incidental thyroid nodule of apparently low risk; approach X per guideline; expected benefit and harms discussed". The sentence forces the explicit leap from finding to approach. Clinical pearls speed up recognition of urgent patterns (clinical pearl); here the muscle is the opposite: taking a deliberate pause before naming.

Use probability only where it fits

When there are symptoms and a hypothesis, return to pretest probability and the base rate. When the person was well and the finding came from screening or from an exam done for another reason, the dominant question is not "what is the LR of this test?" but "is this label worth the package of consequences?". Mixing the two frames is a common teaching error.

Practice with fictional cases of realistic presentation

Build vignettes in which the report "closes" an indolent disease and ask your colleague to argue for and against naming and treating it. The goal is not cynicism: it is to calibrate the threshold for action. Then compare with active surveillance guidelines where they exist.

Quick checklist

1. Name the driver: screening, threshold or incidentaloma?
2. Confirm whether it is a false positive or potential overdiagnosis.
3. Estimate the benefit and harm of the next step.
4. Record the trade-off conversation.
5. Avoid an automatic cascade of tests.
6. Review whether the label is obscuring another hypothesis.
7. Reassess at follow-up whether the approach still makes sense.

A study scenario (fictional case with a real clinical presentation)

Imagine a middle-aged adult who had an abdominal CT for ureteric colic that has already resolved. The report describes an adrenal nodule of a few millimeters, with density suggestive of adenoma and no alarm signs. The original question (flank pain) already has an answer. Overdiagnosis comes in if the team turns this finding into a journey of hormone tests, new imaging and anxiety without proportional indication. The prudent path ties the finding to published criteria for adrenal incidentaloma, explains uncertainty and arranges surveillance only if the expected benefit outweighs the noise.

Another scenario: opportunistic thyroid screening by ultrasound requested "since you're already at the appointment". The incidence of thyroid cancer rises when you look for it; mortality does not fall in the same proportion, a classic pattern of population overdiagnosis (Davies and Welch, 2014). The student trained only for sensitivity celebrates the nodule; the student also trained for overdiagnosis asks whether that nodule changes the patient's life plan.

Relation to biases and to the culture of "missing nothing"

The fear of error by omission pushes toward more tests. That fear is understandable and, in serious hypotheses, protective. It becomes a distortion when the service culture rewards only sensitivity and never discusses the harm of the label. Training for overdiagnosis is training an ethics of calibration: the same seriousness for aggressive cancer and for the indolent finding. Without that, the calibration of confidence remains incomplete: the clinician trusts too much in the automatic value of "finding something".

The communication evidence shows that patients accept uncertainty better when it is named early, with concrete options, than when they find out later that the treatment was optional (Welch and Black, 2010; Brodersen et al., 2018). In preceptorship, ask the student to rehearse the trade-off sentence out loud before entering the room.

What to record in the chart

Vague records ("nodule to be clarified") open the door to a cascade by any on-call physician who follows. Prefer: size, context (screening versus symptom), hypothesis of indolence or not, agreed plan and reassessment deadline. This protects the patient and the collective reasoning. If the approach is observation, write return criteria (growth, new symptoms, patient preference).

How NEXO helps you train

NEXO offers fictional cases with a real clinical presentation in which the task is to decide the next step under a vignette. This includes braking when the finding does not justify a cascade: comparing hypotheses, noticing when the test is noise and calibrating confidence without celebrating an easy label. You can train the same gesture of the diagnostic pause and of critical reading of reports.

See the Economy. Those who coordinate a group or league can learn about NEXO Closer. Researchers, reviewers and creators will find NEXO Researcher, NEXO Curator and NEXO Creators.

Connection with teaching and case games

Training environments that score only the "right" diagnosis push the learner to name it quickly. Environments that also score the quality of the decision (observe, do not order a test, explain uncertainty) align the habit with the problem of overdiagnosis. Fictional cases with a real clinical presentation allow you to rehearse the brake without risk to the patient: the feedback can show that the serious hypothesis was considered and ruled out with criteria, not forgotten.

In group practice, alternate rounds: one where the task is to broaden the differential; another where the task is to defend why not to investigate a finding. Both muscles are needed. Without the second, the first becomes a label factory.

In practice, clinical reasoning improves when you separate what the evidence shows from what the service routine reinforces. Set aside a short block after the day's case to note down: decisive finding, discarded hypothesis and next step. This habit links critical reading to decision-making under uncertainty and keeps study from becoming just an accumulation of labels. Use the same notebook to record when management changes with the pre-test probability and when the test only confirms what was already clear. Over time, the list of errors stops being a mere collection of failures and becomes a map of what to train the following week, always with an open source next to the item that raised the doubt.

Frequently asked questions

What is overdiagnosis?

It is diagnosing a condition that, if it had not been recognized, would not cause symptoms or harm in the person's life (Welch and Black, 2010; Kale and Korenstein, 2018).

What is the difference from a false positive?

In a false positive the condition is not present. In overdiagnosis the test or criterion is "right", but the knowledge brings no net benefit (Hoffman and Cooper, 2023).

Does screening always overdiagnose?

No. Well-indicated screening reduces mortality in some contexts. The point is that every screening carries some risk of overdiagnosis; the decision is an informed trade-off (Brodersen et al., 2014).

Is an incidentaloma automatically overdiagnosis?

No. Some incidentalomas are clinically important. It becomes overdiagnosis when the investigation and the label do not improve the net outcome (Kale and Korenstein, 2018).

How to talk to the patient without alarming or omitting?

Describe the finding, the uncertainty about progression, the options (observe, investigate, treat) and the possible harms of each path. Avoid the empty phrase "it's just prevention" (Welch and Black, 2010).

Does this encourage underdiagnosis?

The goal is to calibrate, not to neglect. Serious hypotheses with relevant pre-test probability still call for investigation (pre-test probability).

How to train in medical school?

Use vignettes with incidental findings, discuss surveillance guidelines, practice the five questions in this text and review reports with a supervisor focusing on net benefit, not just sensitivity.

Play the case of the day on NEXO and practice deciding when the next test adds value and when it only adds a label.

References

1. Welch HG, Black WC. Overdiagnosis in cancer. J Natl Cancer Inst. 2010. PMID: [20413742](https://pubmed.ncbi.nlm.nih.gov/20413742/)
2. Brodersen J, Schwartz LM, Woloshin S. Overdiagnosis: how cancer screening can turn indolent pathology into illness. APMIS. 2014. PMID: [24862511](https://pubmed.ncbi.nlm.nih.gov/24862511/)
3. Brodersen J, Schwartz LM, Heneghan C, O'Sullivan JW, Aronson JK, Woloshin S. Overdiagnosis: what it is and what it isn't. BMJ Evid Based Med. 2018. PMID: [29317380](https://pubmed.ncbi.nlm.nih.gov/29317380/)
4. Kale MS, Korenstein D. Overdiagnosis in primary care: framing the problem and finding solutions. BMJ. 2018. PMID: [30127072](https://pubmed.ncbi.nlm.nih.gov/30127072/)
5. Moynihan R, Doust J, Henry D. Preventing overdiagnosis: how to stop harming the healthy. BMJ. 2012. PMID: [23247977](https://pubmed.ncbi.nlm.nih.gov/23247977/)
6. Hoffman JR, Cooper RJ. Beware of overdiagnosis harms from screening, lower diagnostic thresholds, and incidentalomas. Can Fam Physician. 2023. PMID: [36787976](https://pubmed.ncbi.nlm.nih.gov/36787976/)
7. Davies L, Welch HG. Current thyroid cancer trends in the United States. JAMA Otolaryngol Head Neck Surg. 2014. PMID: [24762812](https://pubmed.ncbi.nlm.nih.gov/24762812/)
8. Esserman LJ, Thompson IM, Reid B, et al. Addressing overdiagnosis and overtreatment in cancer: a prescription for change. Lancet Oncol. 2014. PMID: [24727304](https://pubmed.ncbi.nlm.nih.gov/24727304/)
9. Carter SM, Rogers W, Heath I, Degeling C, Doust J, Barratt A. The challenge of overdiagnosis begins with its definition. BMJ. 2015. PMID: [25769312](https://pubmed.ncbi.nlm.nih.gov/25769312/)
10. Rogers WA, Mintzker Y. Getting clearer on overdiagnosis. J Eval Clin Pract. 2016. PMID: [27150158](https://pubmed.ncbi.nlm.nih.gov/27150158/)
11. Bulliard JL, Chiolero A. Screening and overdiagnosis: public health implications. Public Health Rev. 2015. PMID: [29450039](https://pubmed.ncbi.nlm.nih.gov/29450039/)
12. Harris RP, Sheridan SL, Lewis CL, et al. The harms of screening: a proposed taxonomy and application to lung cancer screening. JAMA Intern Med. 2014. PMID: [24322781](https://pubmed.ncbi.nlm.nih.gov/24322781/)