2.2 The doctor as watchmaker - making a diagnosis

The clinician makes a diagnosis from data provided by:

• the patient's account of his symptoms
• signs of disease observed by the patient or clinician
• tests which produce numbers
• tests which produce images

The process is, however, more complicated than hunt-the-thimble because not all diseases are as clearly defined as a thimble, and there are numerous factors that the clinician has to bear in mind when making a diagnosis.

2.2.1 Two types of disease

Diseases are names given to conditions of the mind or body which medical experts agree to label as a disease, which are then included within the glossaries of diseases that they produce, for example the Medical Subject Headings of the National Library of Medicine. However, experts may not agree on whether or not a condition is a disease.

Some diseases are unequivocally different from the normal condition, for example, fractured femur, tuberculosis and lung cancer. In other instances, people who are said to have a 'disease' simply have an exaggerated version of the normal condition, for example:

• if we say that someone has 'high blood pressure' it implies that they are qualitatively different from the general population, but
• everyone has a blood pressure, just as everyone has a height; however
• some people's blood pressure is higher than others and
• in a proportion of the population, the level of blood pressure is sufficiently high to justify the person being offered treatment to reduce it, thus lessening the risk of stroke; such a person is said to 'have high blood pressure'

Both types of disease have to be diagnosed using symptoms, signs or tests, and the accuracy of a symptom, sign or test is measured by its sensitivity and specificity.

2.2.2 The clinician must balance sensitivity and specificity when choosing a test

The two key characteristics of a test are sensitivity and specificity.

• The sensitivity of a symptom, sign, or test is measured by the proportion of people who have the disease, among whom the test is positive
• The specificity is measured by the proportion of people without the disease, among whom the test is negative

The ideal test should be 100% sensitive and 100% specific, but none are. Because of this, tests have false positives and false negatives.

However, the sensitivity and specificity of a test do not entirely define its usefulness. It is also important to assess the predictive value of a test.

2.2.3 When interpreting a test result, the predictive values of a test vary between hospital and general practice

The predictive value of a test is determined in part by its sensitivity, and in part by the prevalence of the disease in the population.

• The positive predictive value of a test is measured by the probability that someone with a positive test has the disease
• The negative predictive value of a test is measured by the probability that someone with a negative test does not have the disease

In general practice, for example, where many diseases are relatively uncommon, tests with a low sensitivity are often negative. Among patients referred to hospital, however, the predictive value of the test is much higher, because it is measured against a different group of people. This can lead to conflict between clinicians in primary and secondary care. In Jerome Groopman's account in Second Opinions of 'A Routine Case of Asthma', he recounts how he has to feed back to the primary care physician the fact that a diagnosis of leukaemia had been overlooked because the primary care physician had not confirmed the diagnosis of bronchitis, based on the patient's symptoms, with a laboratory test or a chest x-ray. As Dr Groopman said,

'We usually do that when symptoms don't respond to empiric therapy.'

'Dr Groopman, come down from your ivory tower and try working here in the trenches.' His voice spat acid and he recoiled in stunned silence. Rarely do physicians confront each other so directly.

'How many patients do you see in a week - six or maybe seven? - with residents and Fellows to do your scut work. Spend a week here with me; I have ten Isabella Monteras a day in the waiting room complaining in broken English that they need time off work because they are tired or can't breathe.'

'We have proven guidelines for what tests to order and what treatment to give. It is not cost-effective to do more than I did for a routine case of asthma. She wasn't bringing up sputum, her chest x-ray and blood count are outside our clinical algorithm for these cases. How many turn out to be a rare manifestation of leukaemia and leukostasis? For every thousand it's asthma in 999 plus times, so don't interrogate me.'

Differences in predictive value between hospital and primary care populations lead to:

• hospital clinicians accusing primary care clinicians of under-diagnosis, and
• primary care clinicians accusing hospital physicians of over-investigation.

Within both primary and secondary care there is a significant problem when diagnosis is based on human observation because observers vary in their detection rate of significant signs.

2.2.4 Inter-observer variability

Some images obviously indicate disease while others indicate absence of disease, but the problem in medicine often lies in grey zones and there are many images, for example x-rays or a pathology specimen, in which a number of different clinicians classify the same image differently. To some the image has a positive appearance whereas others classify it as a negative result. As a result, measurements of inter-observer variability are increasingly included in healthcare evaluation.

2.2.5 MUPS (medically unexplained physical symptoms)

Notwithstanding the weaknesses of diagnostic tests (which have only recently been recognised, but have always existed), it is now possible to diagnose accurately a much higher proportion of disease than in the past because of the power of modern diagnostic tools such as Magnetic Resonance Imaging (MRI). Many patients are therefore often reassured when no diagnosis is made. However, some patients are not reassured, and it is now understood that their symptoms can arise from a condition called MUPS, (medically unexplained physical symptoms), a recently recognised disorder, that also needs to be actively managed.

2.2.6 Bringing pathology face to face with patients

Clinical decisions can be classified as those which are made during the consultation - face to face decisions - and those which, at present, are made by a clinician remote from the patient. These latter, faceless, decisions are usually made by radiologists or pathologists who look at the image of a part of the patient - for example, a chest x-ray or a mammogram in the case of radiologists, or a cervical smear or lymph node removed during an operation in the case of pathologists. Alternatively, the decision about the patient may be made on the basis of some numbers calculated from the analysis of a patient's blood sample. This is what happens when a clinician sends a blood sample to the biochemistry laboratory. Haematologists making decisions about blood disease may make their faceless decisions on the basis of either an image, for example a slide on which the patient's blood cells can be examined under the microscope, or on the basis of numbers.

The reports produced by pathologists would be meaningless to most patients. There is, however, a movement among clinicians who are having to interpret the reports of pathologists and radiologists to force them to make their reports more comprehensible for both clinicians and patients. The clinicians who have to interpret these reports have pointed out the peculiar language used by pathologists, for example in a letter to The Lancet that was at the one time both humorous and serious, Dr Kay pointed out the incomprehensibility of systems for describing a certain type of lymphoma. (1)

Sir - The announcement in 'The Lancet' of two more classifications of non-Hodgkin's lymphomas encourages me to put forward my classification of these classifications.

This system makes no claim to be comprehensive or even comprehensible, so there may well be scope for other classifications of classifications and ultimately, one hopes, a classification of classifications of classifications. At that point we shall need a conference in the Caribbean.
Yours sincerely,
(Dr) H.E.M. Kay

One solution to this has been to encourage those who make faceless decisions to be more straightforward in their systems of classification and the reports they write for clinical colleagues.

A second, more radical and more sensible, approach has been to suggest that pathologists start thinking about writing reports for patients. In a powerful Editorial in The Lancet, one distinguished clinician suggested that pathologists had 'run amok' in describing carcinoma in situ of the breast. Elliott Foucar, an American pathologist, pointed out that the term 'carcinoma in situ' is often assumed by the patient to mean carcinoma, whereas it is more appropriate to consider the changes as representing a degree of risk rather than a clinical disease state. (2) His conclusion was that although pathology can be very difficult, 'it does not require specialised training in pathology to recognise that the patient's diagnosis should not be an anachronism sustained by anecdotes, conjecture, and tradition'. He called for a more honest description of the pathologist's observations and it would obviously be sensible to try to write these pathology reports both for the patient and for the clinician.

In biochemistry numbers are given out and expressed as lying within or outside a "normal range". The normal range, however, for a chemical in the bloodstream which has defined upper and lower limits is simply a description of the frequency of results. If, for example, the normal range is from 50 to 60, the person whose blood result is below 50 or above 60 may indeed have a disease but cannot be assumed to have a disease solely on the basis of their blood result, any more than the man who is below 5'5" or above 6'5" can be assumed to have a disease although they lie outside what could be called the normal range of height.

What is needed is for pathologists to summarise their findings so that they can be understood by patients. Admittedly that is a daunting task when one considers the wide range of educational levels of patients, but even if pathologists were to write their reports for no more than the 10 or 20 percent of patients who had a high school or university education, the discipline of doing so would not only force them to think of the consequences of the decision and bring them, if not face to face with the patient, at least directly in contact with the patient and the decision. Furthermore, from the complaints of clinicians about pathology results, if results were written to be comprehensible to the 10 or 20 percent of patients with high school or university education, they would almost certainly be more comprehensible for the general practitioners, physicians and surgeons who all too often find pathology reports difficult to interpret.

2.2.7 From diagnosis to action

Having made a diagnosis, the clinician has next to appraise options for action.

Top