The challenge of overdiagnosis begins with its definition
BMJ 2015; 350 doi: https://doi.org/10.1136/bmj.h869 (Published 04 March 2015) Cite this as: BMJ 2015;350:h869- S M Carter, associate professor 1,
- W Rogers, professor 2,
- I Heath, retired general practitioner3,
- C Degeling, postdoctoral research fellow1,
- J Doust, professor4,
- A Barratt, professor5
- 1Centre for Values, Ethics and the Law in Medicine, University of Sydney, NSW 2006, Australia
- 2Philosophy Department and Australian School of Advanced Medicine, Macquarie University, NSW, Australia
- 3London, UK
- 4Faculty of Health Sciences and Medicine, Bond University, Australia
- 5Sydney School of Public Health, Sydney, Australia
- Correspondence to: S M Carter stacy.carter{at}sydney.edu.au
- Accepted 28 January 2015
The implicit social contract underpinning healthcare is that it will reduce illness and preventable death and improve quality of life. But sometimes these promises are not delivered. Sometimes health services take people who don’t need intervention, subject them to tests, label them as sick or at risk, provide unnecessary treatments, tell them to live differently, or insist on monitoring them regularly.1 These interventions don’t improve things for people; they produce complications or illness, reduce quality of life, or even cause premature death. Active health intervention is not always a good thing: it can be “too much medicine,” or produce what is often called overdiagnosis. Although the concept of overdiagnosis has been described in the literature for nearly 50 years in relation to cancer screening,2 3 it was Welch and colleagues’ 2011 book, Overdiagnosed: Making People Sick in the Pursuit of Health, that popularised the term.4
Overdiagnosis is now an acknowledged problem for patients, clinicians, researchers, and policymakers; it is discussed in journals5 6 7 and at specialist conferences8 and addressed through policy and practice initiatives.9 10 11 There is, however, no formal, agreed definition of overdiagnosis. Rather, the word has become a banner under which disparate people with similar general concerns can unite. This vagueness and breadth allows the appearance of unity but does not serve the more exacting demands of research and healthcare. Here we examine the meanings of overdiagnosis more closely and discuss related challenges for healthcare professionals, patients, and researchers. If overdiagnosis is to be understood and mitigated, the broad concept should be subdivided into different problems and its ethical dimensions better acknowledged.
Towards a definition
Overlapping concepts
Aspects of overdiagnosis overlap with existing movements in health policy and practice10 such as evidence based medicine, patient centred care,12 strategies for disinvestment,13 and quality and safety in healthcare, especially preventing iatrogenic illness and low value healthcare.14 A careful comparison with these better defined problems will allow those concerned about overdiagnosis to learn from related work, avoid redundant work, and better identify what is unique about overdiagnosis.
Social and ethical dimensions of overdiagnosis
A deeper understanding of overdiagnosis requires moral as well as technical analysis. It is tempting to seek a purely technical definition of overdiagnosis that excludes context, values, and ethics. But the much debated balance of benefit and harm central to the concept of overdiagnosis is also central to healthcare ethics. Technical definitions of overdiagnosis quickly confront moral considerations, such as what types of benefit or harm should matter 15; how different benefits and harms, or benefits and harms to different people, should be weighted; whether benefits and harms should be measured in individuals or systems and society; and who should judge which benefits and harms matter.16 For example, the harms of overdiagnosis are often side effects of treatment. But which side effects are important enough to include in any measurement of harm? Are some side effects more important than others? Who should decide—patients, clinicians, or researchers? And what if they disagree?
To understand, define, and respond to overdiagnosis, we also need to understand complex healthcare systems and the people who use and serve them. Hoffman and Kanzaria argue that overtesting and overtreatment will continue until doctors, patients, and society learn to accept the uncertainty inherent in the practice of medicine.17 Clinicians attempting to help their patients in a system loaded with incentives and penalties, citizens trying to comply with health advice, companies required to deliver profits, defensive medicolegal systems, the rise of over the counter and internet enabled self testing, and bureaucratic key performance indicators are all features in the landscape of overdiagnosis.
Multiple related concepts
The table⇓ sets out an inclusive map of concepts related to the problems of overdiagnosis or too much medicine. Each concept is contestable; we provide rough rather than authoritative definitions. As illustrated, these concepts are interrelated. For example, overdetection and overdiagnosis are drivers of overtreatment and overutilisation. Disease mongering probably leads to overdiagnosis. Overmedicalisation permits or encourages disease mongering, overtreatment, and overutilisation.
The concepts in the table range from broad to narrow. The broadest is arguably overmedicalisation—for example, defaulting to biotechnological responses rather than existential wisdom to deal with our fear of death1 or defining disruptive behaviour in children as a medical problem requiring drug treatment when social or behavioural interventions may be equally effective with a lower probability of harm.22 These are cultural problems, connected to profound questions about what constitutes a good human life. Other overdiagnosis problems are far narrower and more instrumental—for example, how mammography services can minimise the number of women experiencing both late stage breast cancer and unnecessary treatment of early stage breast cancer.23 The narrowest concepts in the table are arguably overdetection and overdiagnosis. The narrower the concern, the more concrete and individually focused it is, and the more amenable to quantitative measurement.
This leads us to our central observation. The word overdiagnosis is being used in two ways. It is used as the umbrella term for most of the concepts in the table, but it is also used to label one narrow concept in the set. Using the word overdiagnosis in both the broad and the narrow sense is imprecise for researchers and clinicians and potentially confusing for the public and decision makers. We suggest it should stop. If concerned parties continue to use the word overdiagnosis to mean several different things at different levels of generality,24 they are likely to talk past one another or waste energy on unnecessary disagreement. Instead, a new umbrella term (such as too much medicine5 or less is more medicine10 25) could be used for the broad conception summarised at the beginning of this article. Adopting a new umbrella term would spare the word overdiagnosis for the narrower, more precise, meaning.
Such a change could help resolve some disagreement over the extent and scope of the concept. Issues such as overtreatment, overutilisation, overmedicalisation, and disease mongering cannot be readily shoehorned into a narrow definition of overdiagnosis but fit easily into the broad set of too much medicine. Debate over the narrow concept of overdiagnosis could then be restricted to a few central concerns (box 1), such as whether it occurs in symptomatic, or only asymptomatic, people and the relations between overdiagnosis, risk factors, and false positive findings.
Box 1: Classic overdiagnosis, narrowly conceived, precise, and condition specific
Thyroid cancer provides a useful example of the narrow sense of overdiagnosis26 27 In the US, the rate of diagnosis of thyroid cancer has tripled over the past 30 years, from 3.6 cases/100 000 in 1973 to 11.6 cases/100 000 in 2009, with most of the extra diagnoses being of papillary cancer. This rise in thyroid cancer diagnosis has been linked to the increased use of portable ultrasound machines for screening asymptomatic people. Before ultrasonography was available lesions were identified by clinical examination, usually when patients presented with symptoms. Now lesions as small as 2 mm can be identified and biopsied. If malignant cells are found, patients are offered thyroidectomy, the rates of which in the US have increased by 60% over the past 10 years. Despite the rise in diagnoses and treatment, the death rate from thyroid cancer has remained stable.26 This suggests that the extra diagnoses and treatments are not reducing morbidity or mortality.
To tackle this example of overdiagnosis we need to understand the natural course of these very small lesions, which may grow too slowly to become symptomatic during the person’s lifetime. Are they one end of a spectrum of tumour behaviour ranging from indolent to aggressive, and if so, can we identify which will remain indolent? Or are they a separate pathological phenomenon? What are the relevant histopathological and genomic features that might answer these questions?
Precise definition for specific purposes
Broader and narrower conceptions of overdiagnosis are suited to different purposes. Different interested parties are working with different implicit understandings of the concept. They are focused on different conditions and problems (mental health, cancer screening, drug promotions, etc) and have different purposes: unpacking the logical structure of overdiagnosis, explaining how it occurs, measuring it, advocating for political change, or developing practical tools to change clinical or policy practice.
Each of these purposes is important, but they occur at different levels of generality. Advocacy, for example, can employ the broadest conception. It is here that a new umbrella term such as too much medicine may be most helpful because it is arresting, inclusive, and easy to understand. In contrast, epidemiological measurement of overdiagnosis requires greater precision and encodes multiple assumptions specific to each condition studied. Methods for measuring overdiagnosis in a particular condition, along with the assumptions underpinning these methods, implicitly define overdiagnosis in that condition. These methods and assumptions will change disease by disease, test by test.
For all conditions, measuring overdiagnosis requires good quality, large scale data collection over time, and an international effort to agree on appropriate analysis methods for that condition. This agreement has proved difficult. More work has been done on methods for estimating overdiagnosis in cancer screening than in any other condition, but still there is deep disagreement about appropriate methods, and all current methods carry considerable risk of bias.28 This process needs to be repeated for other conditions, because in each condition the drivers of overdiagnosis and potential sources of bias differ. For example, overdiagnosis of breast, prostate, thyroid, and lung cancer is driven largely by screening and early detection programmes,29 so potential biases include lead time; confounders include population trends in cancer risk factors such as use of hormone therapy, bodyweight, and smoking. In contrast, the drivers of overdiagnosis in high blood pressure, high cholesterol, and diabetes include more frequent testing and regular changes in the thresholds for what is considered abnormal (the threshold for high blood pressure, for example, has fallen from 160/100 mm Hg to 140/90 mm Hg). Rather than responding to lead time, methods for estimating overdiagnosis in these conditions must respond to threshold changes, but it is not yet clear how this should be done. For each condition the pattern repeats: different drivers, different biases, and so different methods.
Beyond advocacy and epidemiology, others are examining overdiagnosis for different purposes. Social scientists are studying how clinicians, decision makers, patients, and citizens make sense of overdiagnosis and too much medicine.16 30 For example, some of us are studying general practitioners’ understanding and management of overdiagnosis in prostate cancer in Australia and the UK and Australian decision makers’ understanding of overdiagnosis in mammography; we are mapping the diversity of what overdiagnosis means to different stakeholders and the deeply held values that support their understandings.31 These are conceptualisations based in everyday practice and will inform the policy response to the problem. At a more abstract level, scholars in the philosophy of medicine are studying the logical structure of overdiagnosis, developing precise formal definitions and typologies and sound arguments on questions such as whether overdiagnoses, false positive results, and misdiagnoses intersect or are mutually exclusive.16 24 32
Meanwhile, on the front line of medicine and public health, strategies are being developed to decrease both too much medicine and specific cases of overdiagnosis. Different groups—general practitioners, clinical specialists, policy makers, citizens, and patients—face different challenges. For those on the front line, the most important step may be accepting the unsettling general sense of too much medicine: that medicine is an uncertain practice, that healthcare may be harmful, and that attending to harms is as important as attending to benefits. Such a cultural shift will make it easier to translate the precise work done by researchers on specific cases of overdiagnosis. Such translation has begun, facilitated by professional organisations and consumer groups including Consumer Reports,33 the Academy of Royal Medical Colleges,34 the American Board of Internal Medicine Foundation,14 the National Institute for Health and Care Excellence,35 and the Royal College of General Practitioners.11
From here
We may never agree on a single definition of overdiagnosis. But we can and should be more explicit about what we mean when we use the term, including the breadth or precision, relevant conditions, assumptions regarding benefit and harm, and purpose (box 2). Clarifying these dimensions will serve our ultimate goal of getting a better grasp on the important problem of too much medicine.
Box 2: Actions to clarify overdiagnosis and too much medicine
Recognise that both overdiagnosis (narrow) and too much medicine (broad) are social and ethical problems, not just technical and scientific problems
Stop using the word overdiagnosis to refer to the broad too much medicine problem
Recognise the need to respond to overdiagnosis specialty by specialty and condition by condition
Develop a clear definition of too much medicine for use in public and political communication
Engage with others working in closely related movements, such as low value care and patient centred care, and systematically study the similarities and differences
Recognise the potentially competing values at stake in defining and tackling overdiagnosis and too much medicine and develop inclusive strategies to take full account of these
Promote public debate on the inherent uncertainty and limitations of healthcare and their implications for overdiagnosis and too much medicine
Notes
Cite this as: BMJ 2015;350:h869
Footnotes
Contributors and sources: This article was prompted by our presentations at the Preventing Overdiagnosis 2014 Conference in Oxford, UK. All authors have written or presented independently on the definition of overdiagnosis. SMC led the writing and is the guarantor. All authors contributed to conceptual development and writing.
Competing interests: We have read and understood BMJ policy on declaration of interests and declare that we have no competing interests.
Provenance and peer review: Commissioned; externally peer reviewed.