Solving the Diagnostic Challenge: A Patient-Centered Approach

I would like to thank Jef van den Ende for his comments on my paper titled 'Solving the Diagnostic Challenge: A Patient-Centered Approach'. His comments highlight some interesting aspects pertaining to making a diagnosis in primary care as well as other medical settings.

When proposing their threshold model, Pauker and Kassirer (1) presumably did not have the primary care situation in mind. I agree with Jef van den Ende that they were rather focusing on the question of ordering "a costly or dangerous test" after an appropriate history, clinical examination and less invasive tests had been performed. This occasion arises more frequently in hospital specialist than in primary care. However, since their model has interesting implications regarding the termination of the diagnostic process, I find it a useful tool to understand problems and dilemmata in primary care practice.

In primary care the diagnostic possibilities are only rarely exhausted. Quite often, the case is "solved" on the basis of a brief history alone, sometimes a focused physical examination is needed, occasionally additional tests. In other words, clinicians are constantly checking whether sufficient data have been gathered to terminate the diagnostic process and make a decision, such as prescribing treatment (which may consist of suggesting OTC drugs and home remedies), or giving up a particular hypothesis, pursuing an alternative or resort to watchful waiting. This checking is mostly tacit, but the Pauker and Kassirer theory provides us with a tool to make our steps in deciding when to stop explicit. I agree with Jef van den Ende that this is quite a step from Pauker and Kassirers paper, but one of considerable relevance to a reflection on diagnostic processes. Given the explosion of diagnostic technologies we are facing, I wonder whether this continuous process of checking, whether you have crossed a threshold or not, would also be appropriate for secondary or tertiary care. We have become aware of the risks inherent in tests, which look innocent at first glance, but carry with them the potential for incidental distractions, overdiagnosis and cascade effects.

I agree with Jef van den Ende that simple questions can bring down the probability for relevant serious disease. My focus in the paper, however, is on how we enter into considering these in the first place. Here one would better start the discussion with concrete patients with all their idiosyncratic symptoms and findings, not diseases. It then becomes obvious, that the selection of hypotheses to be evaluated in an individual patient is critical. Given the large problem space of primary care practice and the potential to evaluate a large number of possible explanations, this first step deserves more attention. We have proposed the term 'inductive foraging' to make this accessible to research and relevant for clinical teaching. As Jef van den Ende rightly points out, findings with good 'confirming power' (positive likelihood ratio; high specificity) are especially valuable at this stage.

I do not agree with Jef van den Ende in that Pauker and Kassirer consider only the accuracy of diagnostic tests. Benefits and risks of treatment as well as risks of a test are an explicit part of their determination of decision thresholds. They only require these components to be quantified on the same utility scale. I would agree with them that it this not a weakness of their model but a strength, because underlying utilities (values) associated with relevant outcomes have to be made explicit. (1) That decision thresholds are greatly influenced by values is underlined by Tsalatsanis et al. (2) They require regrets of commission and omission to be quantified on a visual analogue scale for their procedure of determining decision thresholds (see also H. Sox, Ch 9, for more detailed formalizations (3) ) - The threshold models discussed here are a major step forward in our understanding of diagnostic processes since they make clear that diagnostic recommendations and actions are necessarily based upon values.

Jef van den Ende rightly points to an error. In the section titled THE ECOLOGY OF THE CLINICAL PROBLEM SPACE, 2nd paragraph, it should read "In low-prevalence settings, however, the likelihood of disease being present after testing, that is, the negative predictive value for disease, is invariably HIGH".

Another area of agreement between us is the value of the history and physical examination. Nowadays, doctors are overwhelmed by a plethora of laboratory tests and imaging procedures, which make them loose sight of wood for all the trees. While Jef van den Ende stresses the value of the history and the clinical examination in excluding disease, I would like to emphasize the value of the first minute of the encounter with the patient to set the agenda, i.e. to decide which hypotheses should be considered at all.

Literature Cited
1. Pauker SG, Kassirer JP. The threshold approach to clinical decision making. N Engl J Med 1980; 302(20):1109-17. Available from: URL: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=7366635.
2. Tsalatsanis A, Hozo I, Vickers A, Djulbegovic B. A regret theory approach to decision curve analysis: a novel method for eliciting decision makers' preferences and decision-making. Bmc Med Inform Decis 2010; 10:51.
3. Sox HC. Medical decision making. Philadelphia: American College of Physicians; 2007.

Competing interests: None declared

To the editor:

We agree with the author that carefully listening to the patient is crucial in diagnosis. As Osler said: "listen to the patient, he will give you the diagnosis on a silver plate". But we have some concerns.

Colleague Donner-Banzhoff does not interpret the 1980 article by Kassirer and Pauker correctly. These authors discuss the decision to order a dangerous or costly test, after a thorough workup with anamnesis, physical exam and common diagnostic means.

It is not correct that when you cross the test-treatment threshold (this is NOT the therapeutical threshold, they discuss in their 1975 publication), you are allowed to treat. When you cross it, you further work out the diagnostic problem of your patient, correctly and completely. Only faced with the question of ordering such a last costly or dangerous test, you apply this threshold model.

On the down side, under the test threshold we follow the same rules: in the beginning of your contact, it might be that you arrive under the test threshold for a certain hypothesis. But if findings with a good confirming power (positive likelihood ratio) remain, you have to play them out.

Pauker and Kassirer also do not state these thresholds are definite: in their calculation, they only considered diagnostic accuracy of the test, not the risk, cost or preferences. But they suggest clinicians should at least consider the maximum potential of the test: if the case is already high over the EUT (1975, 'therapeutical') threshold, it makes no sense if the test, if negative, would not bring the probability under this threshold. Mutatis mutandis for the confirming power and the test threshold.

We agree the diagnostic space in family medicine is quite large. But as the author states further, you may increase the probability of coronary artery disease substantially by asking if the chest pain is linked to walking up the stairs, stays for a few minutes and subsides after a few minutes. You go from 1% up to greater than 30% if the patient answers yes. And you are over the test threshold for treadmill testing.

Pulmonary embolism: ask for recent travel, recent operation, painful or/and swollen leg, dyspnea. If these are negative, you are at a quite reassuring very low probability.

I train students and lecturers worldwide in diagnostic reasoning. The problem for most clinicians nowadays is that they do not belief any more in the value of anamnesis and physical examination. Both are often of incredible high value, especially for excluding diseases.

"In low-prevalence settings, however, the likelihood of disease being present after testing, that is, the negative predictive value for disease, is invariably small. Even highly sensitive tests do not usefully modify this low probability."

Negative predictive value is the probability NOT having the disease, if a finding is negative.

Moreover, for dangerous and treatable diseases, we should bring the probability to such a low level, that no single other finding or test could cross again the test threshold.

An enormous problem is the representation of disease probability on a linear scale. This does not represent the 'space' of family medicine. The fact to bring the probability down from 1% to 0.1%, has got as much value as bringing it down from 50% to 10%. The same test generates both decreases, following Bayes theorem. Therefore, it is necessary to work on a logarithmic scale, preferably a logodds scale, which shows both extremities of probability, high and low.

"In other words, with low prevalence, sensitive tests often are seldom informative.15"

Notice that the reference you use is 33 years old... In the mean time, science has progressed.

"This situation conflicts with the threshold model's implicit assumption that the probability of disease is above the diagnostic threshold and below the therapeutic threshold at the beginning of the diagnostic process."

This is alas not true at all. I do not know where you found this. It is all about the question what the 'test field' of P and K means. It is about a last test.

"Obviously, criteria with high specificity are particularly helpful at this stage: if they are met, they indicate the presence of disease with high probability."

Alas not true: for a very rare disease, e.g., less than 0.01%, a highly specific finding can bring your probability not higher than 1%. See reverend Bayes.

Van den Ende J, Bisoffi Z, Van Puymbroeck H, Van der Stuyft P, Van Gompel A, Derese A, Lynen L, Moreira J, Janssen PAJ. Bridging the gap between clinical practice and diagnostic clinical epidemiology: pilot experiences with a didactic model based on a logarithmic scale. J Eval Clin Pract 2007; 13(3): 374-380.

Competing interests: None declared

Herbert, Stewart and Reeve have each welcomed the message that 'patient-centred care pays' in different ways - and I certainly share this reaction.

But I'm wondering 'so what?' and 'where now?' Here are three possible approaches.

The first is the obvious question - does it really work? Is inductive foraging as good as Donner-Banzhof implies? He makes a provocative comment in the first paragraph of p 356 "As long as disease prevalence is low, physicians are therefore perfectly justified in searching for suggestive (positive) findings." It sounds almost too simple, so I do hope this paper stimulates more research - it needs replicating in other settings and deepening and extending in any setting. We know diagnosis can be very challenging in General or Family Practice; can we be confident that, as well as improving care for the large group of patients with ill-defined conditions, inductive foraging with triggered routines will also help us diagnose rare but dangerous conditions? Intuitively I would judge yes, but it needs more testing. Perhaps the real issue here remains how best to combine medical expertise with patients' wants and needs.

This leads on to my second thought - how much can collaboratively searching the problem space with patients help our colleagues in other specialties? Donner-Banzhoff is clear that ER physicians have much to gain and I would like to go further and see the 'method' tried out in a variety of specialist settings. Surely any setting where unselected patients present will benefit, but I think we can also look at truly specialist settings. All specialists meet patients with symptoms not clearly based on structural disease. Years ago I gained much from a study of patients with headache referred to neurologists in London (1) Patients referred by GPs were interviewed about their expectations before they saw the specialist and then reviewed afterwards. The outcomes were interesting. Around one third of patients expressed dissatisfaction with their specialist encounter and this was particularly likely when the patient had ideas about the cause of their headaches that were not discussed or addressed. It is always worth finding out the patient's concerns and expectations!

So this leads me to a third question - how best may we teach inductive foraging? Can we incorporate it right from the earliest clinical experience? I hope the concept may enlighten the 'black box' of diagnostic inspiration for students at all stages of a medical career. A final caution: while I believe collaboratively searching the problem space is a 'must do' for high quality clinical practice, some patients can combine a forceful personality with abundant but perhaps irrelevant information. Physicians must be careful to combine empathy with a professional perspective. Patients sometimes need protection from themselves!

Reference

1 Fitzpatrick R, Hopkins A. Referrals to neurologists for headaches not due to structural disease. Journal of Neurology, Neurosurgery, and Psychiatry 1981;44:1061-1067.

Competing interests: None declared

I would like to thank Moira Stewart, Carol P Herbert and Joanne Reeve for their thoughtful comments on my paper. Its main idea was to explore the cognitive aspects of the consultation. How do physicians meet the challenge of making a diagnosis under the conditions of primary care, i.e. low (or very low) prevalence, uncharacteristic presentations and with little advanced technology?

In answering this question, the paper as well as the research it is based upon has a cognitive focus. As such it contrasts with current formulations of the aims, scope and methods of primary care, not least those provided by Moira, Carol and Joanne themselves. I am grateful to see Joanne welcoming new work on the intellectual task of clinical reasoning and decision making in our discipline. My intention is to enrich rather than to replace previous statements on what makes family medicine unique and different from other medical disciplines. Moreover, I think that placing conceptions of our own work in the context of existing theories helps us clarify our own thoughts and justify our approach to others.

This, however, is a question we should ask ourselves: are the strategies and concepts presented on the basis of a primary care survey applicable to primary care only? From an empirical point of view, the answer is yes. 'Inductive foraging' is in fact occurring in almost every primary care consultation. A comparison with emergency room (ER) physicians shows inductive foraging to be used much less frequently in the hospital setting (publication submitted).

But what should the normative consequences be? Should we accept that our colleagues in different settings have their own specific style and leave it at that? Which would give us even a certain feeling of moral superiority (WE are patient-centered, THEY are not). Or should inductive foraging as an efficient AND patient-centered strategy become a norm outside primary care? In my view, dozens of routine questions thrown at baffled patients in the ER are ineffective and confusing. ER physicians might improve their practice by giving their patients a chance for an 'inductive foraging' opening of their consultations. We should keep in mind that Sir William Osler, rightly mentioned by Carol, was not a primary care physician. In other words: seen from a normative point of view, inductive foraging and related strategies might be an appropriate any physician or setting. The more generalist their practice, the more this would apply.

In this I assume a much more restricted notion of generalism than my co-discussants in their writings. My understanding of 'generalism' applies to any setting where unselected and/or unreferred patients are seen. Primary care, which in most health care systems is practiced with ambulatory patients only, and hospital ERs are the most important settings in this regard. This notion of generalism is focused on the epidemiology of reasons for encounter, diagnostic accuracy and related cognitive processes. Holistic, or whole person care would not be part of this definition generalism (although essential for good [primary] care).

We still tend to define our discipline and our particular ways of working by negative contrast to other physicians or disciplines . 'Hospital specialists' are the most notorious (and useful) in this regard. Perhaps it is time to actively point out the value of traditions worked out in primary care to other disciplines, and to do so in a more active fashion than in the past.

Competing interests: None declared

This important study offers valuable insights into the work, practice and scholarship (11) of medical generalists/family physicians/GPs. And so contributes to new thinking on primary are redesign (1).

We have long known/understood the goal of generalist medical care to be the task of person-centred decision making. We have a rich description of the consultation/contextual processes that we believe to be important to enable this to happen - consultation models, the biopsychosocial lens, and the importance of relationships, communication, continuity, empathy. But we have focused less on the intellectual task of clinical reasoning and decision making that happens within these processes and contexts.

The current dominant assumption with regard to clinical reasoning is that described within accounts of Evidence Based Medicine (EBM) - a model of professional practice developed from within specialist medicine. In 2010, I used a metaphor of "mining" to describe this consultation approach: the systematic identification of data, analysed using a hypothetic-deductive approach to describe the (statistical) certainty of the presence/absence of a given condition (2).

In that paper, I argued that generalist, whole-person-focused, medicine uses a different approach. Where the goal is to work with the individual patient to construct a person-centred interpretation of an illness experience - " a social interaction with potential consequences for the individual beyond the identification and selection of an evidence-based intervention". I used a contrasting metaphor of "travelling" and of "exploration" over mining (2).

My account drew on the analysis of consultation data (3). Yet it is perhaps surprising how little empirical research (then or since) takes a critical look at the clinical reasoning process. Gabbay and le May's observations of UK General Practice - highlighting the importance of professional capital, the collective professional action that generates knowledge-in-practice-in-context(or 'mindlines' ) (4). McKelvie's observation of clinical reasoning in the ambulatory care setting is due to report in 2019 (5) to name but two. Donner-Banzhoff's work is an important addition to this small dataset (6)

Some of my own data suggests that clinicians may be losing their skills and/or confidence in "inductive foraging" (7,8) with implications for the future of generalist medicine?

If we are to respond to international calls for a strengthening of generalist medicine, then further work is needed to better understand what clinicians - and patients - do with their foraged data to generate new personalised knowledge/accounts of individual illness, and crucially how we recognise best practice in this area (9). Also to consider how we teach "inductive foraging" to students and clinicians; as well as enable patients to be full active partners. Along with new conversations on how generalist and specialist medicine work together - providing distinct but synergistic insights into health needs.

I welcome this paper which provides further empirical data to open up our professional discussions on clinical reasoning and the intellectual task of professional practice (10,11).

1. Reeve J. Primary care redesign for person-centred care: delivering an international generalist revolution. Australian Journal of Primary Health 2018 24(4) 330-336 https://doi.org/10.1071/PY18019
2. Reeve J. 2010. Interpretive Medicine: supporting generalism in a changing primary care world. London: Royal College of General Practitioners Occasional Paper Series, 88.
3. Karasz A, Dowrick C, Byng R, Buszewicz M, Ferri L, Olde Hartmann T, van Dulmen S, van Weel-Baumgarten E, Reeve J. 2012. What we talk about what we talk about depression: doctor-patient conversations and treatment decision outcomes. British Journal General Practice; 62: 30-31.
4. Gabbay J, Le May. 2010. Practice based evidence for health care - clinical mindlines. Routledge, Oxon.
5. www.clahrc-oxford.nihr.ac.uk/research/clinical-decision-making-in-ambulatory-emergency-care-settings
6. Donner-Banzhoff N. Solving the diagnostic challenge: a patient-centred approach. Annals of Family Medicine 2018, 16(4), 353-358
7. Reeve J, Dowrick C, Freeman G, Gunn J, Mair F, May, C, Mercer S, Palmer V, Howe A, Irving G, Shiner A, Watson J. 2013. Examining the practice of generalist expertise: a qualitative study identifying constraints and solutions. Journal of the Royal Society of Medicine Short Reports 4: 2042533313510155.
8. Reeve J, Fleming J, Britten N, Byng R, Krska J, Heaton J. 2018. Identifying enablers and barriers to individually tailored (expert generalist) prescribing: a survey of English health care professionals. BMC Family Practice 19(1): 17.
9. Reeve et al. Revitalizing generalist practice: the Montreal Statement. Annals of Family Medicine 2018; 16(4): 371-373
10. SAPC. GP Scholarship. https://sapc.ac.uk/article/gp-scholarship
11. Reeve J, Firth A. 2017 Revitalising General Practice: unleashing our Inner Scholar. British Journal of General Practice; 67: 266

Competing interests: I am actively engaged in research in the fields of generalist medicine and clinical reasoning. I chair the NAPCRG SIG on Advancing Generalist Expertise

The paper by Donner-Banzhoff is important because of its focus on diagnosis, which for a variety of reasons gains little attention in primary care/family practice. Shining a light on diagnosis, forces us to recognize, yet again, another difference between family medicine and other medical disciplines. Medical education is dominated by the ethos of the central importance of diagnosis to the medical task, but in family medicine, many patient problems do not fit recognised diagnostic categories. Therefore, because of the dominance of the ethos, attempts to practice the threshold approach or the hypothetico-deductive approach are common but lead to mistakes in primary care/family medicine. Hurrah for the paper that described and named an appropriate alternative that should be taught, learned and researched from now on: the inductive foraging approach followed by a triggered routine.

Putting Diagnosis in its Place: Diagnosis is an important task in medicine but it is not the only task. Furthermore, in family practice, 15.3% of encounters do not culminate in any diagnosis (1). Also the workload of family physicians is as follows: 19.2% of encounters for follow-up of already diagnosed chronic diseases alone; 13.8% caring for symptoms alone that have no diagnostic label; and only 11.4% acute illnesses alone which have recently been diagnosed; and the remainder of encounters which had combinations of the above(1). As well, for 1790 patients seen, there are 290 diagnostic labels given (1), reinforcing the point made by the author Donner-Banzhoff that the "clinical space" is quite vast. An allergist whom I was consulting said to me: "My job is easy. A patient like you may have one or two of only 40 diagnoses that I care for. I know how to distinguish the 40 and skin tests confirm the diagnosis". No wonder the conventional diagnostic discovery method works in other medical disciplines but does not work in family medicine.

Diagnosis of the Person: Stretching a little beyond the point of Donner-Banzhoff's paper, I want to suggest that also important in family medicine is the deeper diagnosis of the person, language used by Michael Balint (2) to focus clinicians' attention on the person having the disease rather than finding the disease. Not only is inductive foraging valuable for conventional diagnosis, it likely makes a deeper assessment and understanding of the situation possible. In those myriad of encounters when disease diagnosis is not central (such as for patients with multimorbidity (3) and with unexplained symptoms (4)), a diagnosis of the person is still warranted for reasons given in my final paragraph.

Case Examples: A highly accessible way of making the theoretical discourse more understandable to clinician-learners, is using case examples. Here are some additions to the very apt ones presented in the paper by Donner-Banzhoff. Dr. Fred Platt's famous case (5) points out that the cue to diagnosis need not necessarily come from the clinical domain. A patient in her 30s had a serious cough lasting more than a week, waking at night, and night sweats. One clue was clinical, blood on the pillow after coughing at night. The second clue came from the patient's context; that she worked in an aid center for new immigrants - tuberculosis. A 60 ish female patient experienced weight loss, nausea and extreme fatigue for 4 months before visiting the family physician; after 3 visits and appropriate investigation, the family physician asked, in desperation "What is different in your life in the last few months compared to before?" Only one thing: scraping paint from an old rural building - lead poisoning. The Value of Patient-Centered Inductive Foraging: Research has shown the positive self-rated health outcomes of patient-centered approaches: with family practice patients presenting new symptoms (6); with cancer patients (7); and with patients who have multimorbidity (8). Patient-centeredness has been shown to be a force for equity, creating positive outcomes for patients living in socio-economically deprived areas (9). As well, patient -centered care costs less and therefore creating efficiencies at the health system level (10). To all these benefits, we can now add better diagnosis, because of this seminal work by Donner-Banzhoff.

References 1. Maddocks H, Stewart M, Terry A. Symptoms and Diagnoses at Encounters in the Deliver Primary Healthcare Information (DELPHI) Database. Centre for Studies in Family Medicine, The University of Western Ontario, London, Canada. Report, July 26, 2018. http://www.schulich.uwo.ca/familymedicine/research/csfm/research/current_projects/delphi.html

2. Balint M. The Doctor, His Patient and the Illness. 3rd ed. Philadelphia, PA: Churchill Livingstone; 2000.

3. Fortin M, Stewart M, Poitras M, Almirall J, Maddocks H. A Systematic Review of Prevalence Studies on Multimorbidity: Toward a More Uniform Methodology. Ann Fam Med. 2012 Apr;10:142-151.

4. Rosendal M, Olde Hartman T, Aanland A, et al. Medically unexplained symptoms and symptom disorders in primary care: prognosis- based recognition and classification. BMC Family Practice (2017) 18:18 doi 10.1186/s 12875-017-0592-6.

5. Platt FW, Director, The Patient's Narrative, The Medical Interview Videotape, Colorado University School of Medicine, Foundations of Doctoring Curriculum, July 2000.

6. Stewart M, Brown JB, Donner A, McWhinney IR, Oates J, Weston WW, Jordan J. The Impact of Patient-Centered Care on Outcomes. Journal of Family Practice. 2000;49(9):796-804.

7. Stewart M, Brown JB, Hammerton J, Donner A, Gavin A, Holliday RL, Whelan T, Leslie K, Cohen I, Weston W, Freeman T. Improving Communication Between Doctors and Breast Cancer Patients. Ann Fam Med. 2007 Sep- Oct;5(5):387-94.

8. Mercer SW, Zhou Y, Humphris GM, McConnachie A, Bakhshi A, Bikker A, Higgins M, Little P, Fitzpatrick B, Watt GCM. Multimorbidity and Socioeconomic Deprivation in Primary Care Consultations. Ann Fam Med. 2018 Mar;16(2):127-131. doi: 10.1370/afm.2202.

9. Jani B, Bikker AP, Higgins M, Fitzpatrick B, Little P, Watt GC, Mercer SW. Patient centredness and the outcome of primary care consultations with patients with depression in areas of high and low socioeconomic deprivation. Br J Gen Pract. 2012 Aug;62(601):e576-81. doi: 10.3399/bjgp12X653633.

10. Stewart M, Ryan BL, Bodea C. Is Patient-Centred Care Associated with Lower Diagnostic Costs? Healthc Policy. 2011 May;6(4):27-31

Competing interests: None

Competing interests: None declared