Generations of physicians around the globe have been and still are taught the Jones criteria to diagnose rheumatic fever [1]. These criteria were developed by Dr. Jones in 1944 by a strictly ad hoc and eminence based approach and we are afraid this legacy continued in several later updates. The last attempt to better these criteria was a consensus conference in 1992 [2]. In this last conference the issue of potential geographical differences in the utility of these criteria were brought up, as if this was something unique to rheumatic fever. As we will bring up once more below, the utility of any diagnostic criteria is strictly dependent in the setting in which the criteria are applied. It is no surprise then that more recent formal surveys keep on showing that the sensitivity of the Jones criteria for diagnosing rheumatic fever is only around 30 % in endemic areas like India [3].

In 1974 Dr. Desmond O’Duffy proposed his Behcet’s Disease criteria [4]. The more senior author of this chapter (HY) was in the audience as a young fellow when this set of criteria was presented in a rheumatology meeting. At the end of the presentation he got up and had the courage to ask the presenter “ How successfully do your criteria tell Behcet’s from ingrown toe nails, particularly since I saw no attempts to prospectively test these criteria in a real setting nor a control group in your exercise?” There was little discussion and few heated exchanges, but this outburst was probably the initial stimulus for the latter work related to the formulation of the International Study Group Criteria for Behcet’s Disease (ISGC) [5] currently in use.

Perhaps a new era began in criteria making when American College of Rheumatology (ACR) began publishing criteria for many of the vasculitides [6]. These criteria were no longer ad hoc. A survey was conducted seeking formal sensitivity and specificity for many of the common primary vasculitides. Granted they were based on retrospective analyses and lacked prospective testing, they were an important step away from sheer eminence.

Then came the realization that these ACR vasculitis criteria were not useful for diagnostic purposes [7]. In a formal sensitivity and specificity study it was shown that these criteria had limited (17 % to 29 %) positive predictive values when applied to 198 patients with various vasculitides and connective tissue diseases. Two years later, a further study showed that Chapel Hill Consensus Conference (CHC) criteria, another widely recognized vasculitis criteria set mainly based on the size of the vessel involved, correctly identified only 8 of 27 patients with Wegener’s granulomatosis and 4 of 12 patients with microscopic polyangiitis [8]. The response to this issue was that these two sets of criteria were not intended for diagnostic use but were strictly classification criteria for research and educational purposes [9]. This contention sounded very reasonable when first heard and over the years it became the standard to call all disease criteria classification criteria. This was followed by a new desire and the promise to prepare diagnostic criteria in addition to classification criteria for our diseases, an exercise, which we are afraid, might be likened to constructing a perpetual motion machine.

As with other major and hotly debated issues the “why” of an exercise is often a neglected caveat. Apart from preparing for boards and other such evils there are some very good reasons for having disease criteria. These include:

The presence of separate reasons, at first sight, might be taken to indicate that we might actually need separate classification criteria for research and diagnostic criteria for our patients but perhaps other sets of criteria for still other purposes and even perhaps one for Brussels and another for Washington, as well. We, however, propose that one set of criteria should be good enough for diagnosis, research and public awareness as long we explain, first to ourselves, then to our patients and rest of the public what we intend to with these criteria openly, frankly admitting we cannot diagnose every ill we see. This explanation should obviously continue with the explanation that we can manage some of these ills rather effectively even when we do not know what the exact diagnosis is.

As we emphasized in the previous chapter, many rheumatologic diseases do not have specific clinical, histologic, laboratory or radiologic features. Hence we have to come up with constructs to specify what we mean by a “disease”. For example if we have a shoulder which is swollen in the shape of a shoulder pad and when we biopsy it we find amyloidosis, we do not have to come up with a construct to tell us and the patient that he/she has amyloidosis. The same is true for a painful, hot and swollen knee from which you isolate staphylococci. On the other hand, in a patient with chronic mouth ulcers, attacks of diarrhea and episodes of uveitis you have to build up a construct to identify Behçet’s syndrome and another to identify Crohn’s. Still yet, you have to build up a construct to tell one from the other. Why do you have to resort to constructs? Simply because neither Behçet’s nor Crohn’s can be identified by a specific appearance, histology or a laboratory finding. So you need to build up a concept composed of specific features, in other words a construct. Surely the need for such constructs in rheumatology is not as extensive as in psychiatry with their voluminous standard reference manual, DSM (Diagnostic and Statistical Manual of Mental Disorders) of the American Psychiatric Association which includes definitions of over 400 different mental disorders [10] but we still need them. A set of criteria in turn is nothing more or less than the declaration of the components of a construct with some hierarchy (more commonly called weighing) of these components. It is to be underlined that elements which we decide to exclude from this construct make up the exclusions of our criteria. We propose that the first step in understanding disease criteria is to realize that they are constructs put together for specific purposes to explain and convey departures from the normal. In addition such constructs are needed not only for diseases of unknown origin. Sometimes we resort to constructs in handling diseases we know in depth the etiology and/or the pathogenesis of. For example in a tuberculosis endemic area we can justifiably begin treating a patient with a cough for so many weeks and a chest radiograph according to a well built up construct for diagnosis of tuberculosis or admit a patient with a chest pain for a suspected myocardial infarction if he/she fulfills the Cook County criteria for chest pain [11]. The main message then is that in the science and practice of medicine a diagnosis is needed mainly after we consider what we do with it.

We have emphasized that we need disease criteria especially when our disease is a construct. The 3 basic elements of criteria making all have to do with concepts in probability. They are sensitivity, specificity and the pretest probability.

The second reason we propose for what makes specificity more difficult to grasp than sensitivity is the way we verbalize either concept. When we say “Among 100 patients with SLE 95 patients were ANA positive. Therefore the sensitivity of having a positive ANA test is 95 % sensitive for SLE.”, three positive bits of information follow each other. On the other hand, when we declare “Among 100 patients without SLE, 70 were negative for ANA. Therefore the specificity of having a positive ANA is 70 % specific for SLE.” we again verbalize three consecutive bits of fact however, now, the first two of these are negative while the 3^rd is a positive bit of information. We propose that this mental incongruity is the second reason why specificity is a relatively more difficult concept to remember.