Overdiagnosis: Would we better better off not looking?

There is little doubt that the use of testing has increased dramatically in emergency medicine during my career. Between 2001 and 2010 the use of CT in emergency departments increased 3-fold (and the use of MRI increased 9-fold, but for some reason it is still almost impossible for me to get one done.) (Carpenter 2015) The question is whether all this testing is actually helping patients. Modern imaging is amazing, and I am glad to be working in an era where CT is readily available, but it seems pretty clear that at least some portion of the medical testing being done each day isn’t helping patients. For example, despite massive increases in the use of CT pulmonary angiogram, and an increasing rate of diagnosis of PE, mortality from PE is completely unchanged. (Carpenter 2015) Thus, we are testing, labeling patients as diseased, putting them on long term anticoagulation, but with no ultimate benefits. Part of the problem is that doctors have not embraced the importance of pretest probability, and medicine is awash with false positives. The other issue is overdiagnosis.

“Overdiagnosis occurs when a test finds an abnormality that is technically “true positive,” in that the individual has the pathology that is diagnosed, but that in this particular case, would never have caused actual illness even if undiscovered and untreated.” (Hoffman 2017)

The problem with overtesting and overdiagnosis is that, although we can easily identify these patients in aggregate statistical data, they really can’t be identified clinically at the point of care. Unlike false positives, which can potentially be identified, overdiagnosis occurs in the presence of real pathology, and so additional testing can never put the genie back in the bottle. Thus, overdiagnosis necessarily leads to overtreatment, and therefore patients are not just exposed to the harms of the tests, but also the harms of our (unnecessary) treatments.

Although it might sound too philosophical for practicing emergency physicians, we really must consider the definition of disease. Language has the power to shape reality. In applying a diagnosis to a patient, we essentially make it so. But physiologically, many conditions exist in a spectrum, and although it makes sense to refer to some of that spectrum as disease, it certainly doesn’t make sense to label everyone as diseased.

For example, there is certainly a level at which high blood pressure causes harm, and at which the benefits of intervention outweigh the risks. However, the exact threshold is hotly debated. Should we really be labeling someone with a systolic blood pressure of 142 as diseased? Is that resulting in aggregate good?

More relevant to emergency medicine, we know that small blood clots are tremendously common in the human body. There is an ongoing balance between clotting and lysis that is inherently normal (not diseased). If a subsegmental PE is incidentally found on a trauma CT, should that patient be labeled as diseased? When we find incidental nodules on a trauma CT, at what threshold should we label these as abnormalities? The more tests we order, the most pressing these questions become. 

The explosion of diagnostic information available to us emphasizes the “importance of distinguishing between data (a collection of isolated facts), information (recognition of the pattern that such data implies), knowledge (an understanding of what that information means), and wisdom (knowing how to apply knowledge in a way that improves outcomes).” (Hoffman 2017)

Evidence of overdiagnosis

Most of the evidence for overdiagnosis comes from the realm of primary care and screening. Although not the purview of emergency medicine, these examples help elucidate the problems with testing in low risk populations.

Thyroid cancer is a classic example. The incidence of thyroid cancer was relatively stable for decades, until screening started to become popular in the 1990s. The incidence then tripled between 1990 and 2009, but the mortality remained completely unchanged. “We found that there is an ongoing epidemic of thyroid cancer in the United States. It does not seem to be an epidemic of disease, however. Instead, it seems to be substantially an epidemic of diagnosis… Our findings demonstrate that the problem is due to the overdiagnosis of papillary thyroid cancer, an abnormality often present in people who never develop symptoms from it.” (Davies 2014) The numbers are even worse in South Korea, where a government funded screening program led to a 15 fold increase in thyroid cancer diagnosis with no evidence of improved health outcomes for patients. (Lee 2014; Park 2016)

We see similar evidence for overdiagnosis in breast cancer, prostate cancer, and lung cancer. (Draisma 2009; Welch 2010; Patz 2014)

Emergency medicine overdiagnosis

In a prior post on overdiagnosis, I discussed a paper looking at examples of overdiagnosis in the emergency department. (Vigna 2022) That paper discusses examples of overdiagnosis in pulmonary embolism, coronary artery disease, subarachnoid hemorrhage, and anaphylaxis. In other words, overdiagnosis is probably occurring in the most common ailments we assess every single shift.

I think most people understand how overdiagnosis might occur with pulmonary embolism. However, the more I read about this topic, the more I realize that overdiagnosis is a potential problem for every single patient we encounter.  

In response to a database study suggesting that we might be missing cervical fractures in the geriatric population, Hoffman and colleagues present data to suggest that many spinal fractures in the elderly are asymptomatic, and so “not only would routine intervention not lead to benefit, it would almost certainly produce substantial (and avoidable) harm for many of these people.”  (Hoffman 2017) In other words, although finding a cervical fracture always seems important, they may represent yet another example of overdiagnosis.

Is it possible for a cervical spine fracture to be “overdiagnosis”? At first glance, that seems like an odd claim. Fractures of the cervical spine are important to find, aren’t they? However, I think Hoffman is clearly correct. For the past decade, c-spine x-rays have fallen completely out of favour, being replaced almost entirely by CT. We were told that this was necessary, as the sensitivity of x-ray just wasn’t good enough. That never felt right to me. Obviously, CT is going to find a lot more than x-ray, but was this really a problem we needed to solve? How many patients were sent home from the hospital in the 1990s after a normal x-ray only to become paralyzed because of a missed c-spine fracture? If CT is truly better, we should be able to point to true patient harm from the era before CT, but can we? Or are all the extra injuries that CT picks up meaningless to our patients, with no necessary change in management? (I don’t order imaging for nasal fractures, because it isn’t going to change practice. Perhaps the CT c-spine is similar?)

If overdiagnosis is possible in c-spine fractures, it seems like it is possible for every diagnosis we make. 

Are we really helping our patients by identifying and casting an avulsion fracture? Is an admission for a small troponin bump good or bad? Did we need to find that positive urine culture? For almost every test we order, there are probably examples of patients who would be better off not knowing the results.

How do we improve?

The causes of overtesting are complex and varied, including things like malpractice risk, financial incentives, the preferences of consultants, lack of available follow-up, patient expectations, the increasing complexity of emergency medicine, and the prevailing culture of perfection in medicine.(Carpenter 2015) Carpenter and colleagues also suggest that lack of knowledge of decision rules might result in overtesting, but personally I think the exact opposite is probably true; decision rules with imperfect evidence, and an emphasis on perfect sensitivity, but poor specificity, are probably resulting in overtesting.

We need better research. Many of the arguments around testing are opinion rather than data driven. For example, many emergency physicians feel like ordering tests takes less time than shared decision making, but good data on this topic might prove them entirely wrong. (Ordering tests necessitates at least one extra patient encounter, not to mention time interpreting the test, documenting the results, and the inefficiencies often involved in finding patients in the waiting room after tests have been completed.)

We also need to encourage RCTs of tests. Too often, tests are introduced based only on test characteristics, but without information about patient outcomes. (Or worse, RCTs are performed showing no benefit, but we ignore the RCTs because the sensitivity of the test is pretty good. Think BNP or high sensitivity troponins.) Without appropriate research, it is impossible for clinicians to know when testing is appropriate, and when it is inappropriate.

Unfortunately, even with properly funded research, we are likely to be left with important gaps in our knowledge. “Randomized controlled trials (RCTs), appropriately considered the criterion standard for evaluating the potential benefit of an intervention, are notoriously poor at evaluating potential harm. RCTs are almost always underpowered to look for harm, they rarely look for (and thus fail to identify) harms that were not expected before the study was conducted, they almost never last long enough to evaluate harms that occur over time, and (as geriatricians so well know) they typically exclude precisely those individuals who are most at risk.” (Hoffman 2017)

We need financial systems that reward physicians for good care, and not just efficiency or speed. We need more reasonable legal systems, with the ability to recognize the long term benefits of avoiding testing, rather than just punishing doctors when tests aren’t ordered. 

We need a medical culture that acknowledges test thresholds and the appropriate miss rate, rather than shaming and blaming through traditional M&M rounds. Good education can’t just emphasize a ‘worst first’ mentality, or misdiagnosis, without also emphasizing the many harms of additional testing.

In order for the emergency department to cut back on testing, we need functioning healthcare systems, with reliable follow-up, so we are not left as the only physicians that people can access. 

We need to recognize that tests are valuable when we choose the correct test. We need access to appropriate tests in the emergency department. I can’t access MRI in a timely fashion, so my patients with possible cauda equina often get CT scans first, followed by MRI, increasing the potential harm from testing with no benefit at all. 

Although not really relevant to emergency medicine, the entire logic behind screening and early detection needs to be reconsidered, and screening programs need strong evidence of improvement in real patient oriented outcomes (not ridiculous, artificial outcomes like disease-specific mortality).

We need to recognize that all tests have the potential for harm (through false positives and overdiagnosis), and no test should be done ‘routinely’. (Yes, I am calling out my good friends, the ED ultrasound aficionados.) 

If decision rules are going to be used, they need to be designed with rational goals (the test threshold) rather than simply aiming at 100% sensitivity, and they need to be thoroughly tested through implementation studies that actually demonstrate patient benefit, rather than just stopping when we see a decent sensitivity (while also ignoring the low specificity that will drive over-testing). 

We need better education. The problems of overdiagnosis are not well known in medicine, and it makes sense to start there, but we will not make any headway on this issue until patients also realize that tests can cause harm. 

Emergency physicians recognize wasteful testing, and accept that shared decision making is necessary and realistic. We all want what is best for our patients. However, solutions to these issues will not be found at the individual level. Education is a necessary, but insufficient approach. Current diagnostic testing decisions are largely driven by societal or cultural pressures. Fixing these problems will require societal or cultural change.

For now, all you can do as a practicing doctor is to remember to ask: might my patient be better off if we just didn’t look?

Other FOAMed

Overdiagnosis in the emergency department

Books to read

Overdiagnosed: Making People Sick in the Pursuit of Health

Overtreated: Why Too Much Medicine Is Making Us Sicker and Poorer

References

Carpenter CR, Raja AS, Brown MD. Overtesting and the Downstream Consequences of Overtreatment: Implications of “Preventing Overdiagnosis” for Emergency Medicine. Acad Emerg Med. 2015 Dec;22(12):1484-92. doi: 10.1111/acem.12820. Epub 2015 Nov 14. PMID: 26568269

Davies L, Welch HG. Current thyroid cancer trends in the United States. JAMA Otolaryngol Head Neck Surg. 2014 Apr;140(4):317-22. doi: 10.1001/jamaoto.2014.1. PMID: 24557566

Draisma G, Etzioni R, Tsodikov A, Mariotto A, Wever E, Gulati R, Feuer E, de Koning H. Lead time and overdiagnosis in prostate-specific antigen screening: importance of methods and context. J Natl Cancer Inst. 2009 Mar 18;101(6):374-83. doi: 10.1093/jnci/djp001. Epub 2009 Mar 10. PMID: 19276453

Hoffman JR, Carpenter CR. Guarding Against Overtesting, Overdiagnosis, and Overtreatment of Older Adults: Thinking Beyond Imaging and Injuries to Weigh Harms and Benefits. J Am Geriatr Soc. 2017 May;65(5):903-905. doi: 10.1111/jgs.14737. Epub 2017 Feb 7. PMID: 28170085

Hoffman JR, Cooper RJ. Overdiagnosis of disease: a modern epidemic. Arch Intern Med. 2012 Aug 13;172(15):1123-4. doi: 10.1001/archinternmed.2012.3319. PMID: 22733387

Lee JH, Shin SW. Overdiagnosis and screening for thyroid cancer in Korea. Lancet. 2014 Nov 22;384(9957):1848. doi: 10.1016/S0140-6736(14)62242-X. Epub 2014 Nov 21. PMID: 25457916

Park S, Oh CM, Cho H, Lee JY, Jung KW, Jun JK, Won YJ, Kong HJ, Choi KS, Lee YJ, Lee JS. Association between screening and the thyroid cancer “epidemic” in South Korea: evidence from a nationwide study. BMJ. 2016 Nov 30;355:i5745. doi: 10.1136/bmj.i5745. PMID: 27903497

Patz EF Jr, Pinsky P, Gatsonis C, Sicks JD, Kramer BS, Tammemägi MC, Chiles C, Black WC, Aberle DR; NLST Overdiagnosis Manuscript Writing Team. Overdiagnosis in low-dose computed tomography screening for lung cancer. JAMA Intern Med. 2014 Feb 1;174(2):269-74. doi: 10.1001/jamainternmed.2013.12738. Erratum in: JAMA Intern Med. 2014 May;174(5):828. PMID: 24322569

Vigna M, Vigna C, Lang ES. Overdiagnosis in the emergency department: a sharper focus. Intern Emerg Med. 2022 Mar 5. doi: 10.1007/s11739-022-02952-8. Epub ahead of print. PMID: 35249191

Welch HG, Black WC. Overdiagnosis in cancer. J Natl Cancer Inst. 2010 May 5;102(9):605-13. doi: 10.1093/jnci/djq099. Epub 2010 Apr 22. PMID: 20413742