In message <F10DE59FA201D211989800805FD4E36B0230D212@exchange_01>,
Andrew Jull <[log in to unmask]> writes
>Dear Toby
>
>I would be very interested in reading more regarding your proposed
>methodology. 
>
>In the late 1980s, Patricia Benner adapted a phenomenological methodology
>from Dreyfus and Dreyfus's work on the development of expertise in pilots
>and published a very influential (perhaps overly so) work on expertise in
>nursing. One of the skills Benner illuminated was rapid assessment of
>impending problems, in spite of the clinical signs. Unfortunately, she did
>not go on to explore this skill in greater depth, probably because she was
>working with a group of nurses with different backgrounds. But the "real
>world" signs you elucidated have a distinct ring of truth about them.
>
>Since I got into teaching EBH to advanced practice nurses, I have been
>struck by the same difficulties in teaching the concepts behind 2x2 tables
>(and diagnostic tests) to a group who can see the possible application in
>medicine, but not necessarily to nursing. Yet in hospital-based nursing,
>there must be a point at which an expert nurse - the good nurse in doc's
>terms - decides it is time to alert a medical practitioner of the patient's
>need for further assessment; I don't believe this is just the boundary that
>delineates the nurse's scope of practice having been reached. I have thought
>in the past that this could be tested using the methodology of diagnostic
>testing, if only nursing could throw some light on what exactly it was that
>was alerting the nurse to the presence of a problem. If I may be so bold, it
>looks like you are going down a similar track yourself. 
>

The methodology is really very simple. It's based on the idea that the
"clinical expertise" side of EBP is real, important and measurable. My
interest in this goes back way before EBM was developed. When I was a
medical pre-registration house officer I had the good fortune to work
for a consultant who was an absolutely brilliant diagnostician. He was
called Tommy Studdert and worked in Carlisle, Cumbria (this was 1976).
He could make accurate diagnoses even of obscure conditions with
apparently minimal information, very soon after seeing the patient or
listening to my description. At that time house officers were expected
to do lots of investigations and I was struck by how many of them turned
out to be redundant, and how few influenced diagnosis or management.
Hampton et al had shown how most diagnoses could be made on the basis of
history alone, with clinical examination and laboratory tests
contributing relatively little (Relative contributions of history-
taking, physical examination, and laboratory investigation to diagnosis
and management of medical outpatients. BMJ 1975;2:486-9).

Twenty years later, when I became involved with teaching EBP I found 2x2
tables, sensitivity, specificity, LRs and all that very daunting at
first, but soon began to understand things I hadn't understood before -
in particular the importance of estimating pre-test probability and the
implicit false +ve and false -ve rates *before* doing a test. It was
striking that laboratory tests are often much less powerful than the
clinical history. In "Clinical epidemiology: a basic science for
clinical medicine" very high LRs are cited for histories such as a
classical history of angins (LR+ 115 if I remember correctly) to predict
coronary artery stenosis, compared to LR+ of <10 for exercise ECG). In a
paper about diagnosing lower urinary tract infection in women Wigton et
al found history of UTI, low back pain, pyuria, haematuria and
bacteruria to be independent predictors of UTI (but with LR+s <5) - what
they failed to point out or notice was that the pre-test probability or
prevalence in this population (women in emergency room with suspected
UTI) was already over 60%!

The point of all this is that experienced clinicians are able to make
diagnoses very accurately on the basis of apparently limited
information. 

What I decided to do was interview clinicians (in this case GPs) about
their experience of a particular diagnostic situation (in this case
acutely ill preschool children). The interviews were semi structured and
covered everything to do with these encounters - GPs feelings, how they
perceive and react to parental anxiety, their administrative
arrangements, their perception of how often they see "seriously" ill
children, and so on. I spent time disentangling what actually influenced
their decisions and have been reading and re-reading the transcripts to
extract themes (as I wrote in a previous email, the themes are very
strong and start with the power of first impression). They also said
that their ability to make diagnoses in this way took 2 or 3 years to
develop. It's interesting that Gray and Hampton (International Journal
of Epidemiology 1993;22:222-7) found relatively low sensitivity (55%)
and specificity (88%) for junior hospital doctors initial working
diagnosis of MI - one would like to know senior doctors' performance - I
would bet it is much higher than their juniors'.

Anyway, once the various diagnostic themes have been identified (first
impression, alertness and so forth), one can measure them against
important outcomes. The outcomes can be anything you think is relevant.
In the case of acutely ill children in primary care, I am interested in
whether they reattend, whether they are seen out of hours, whether they
are admitted to hospital. I'm discussing how to set up a template in the
electronic medical record, so that when GPs see acutely ill preschool
children they can simply tick yes/no boxes (e.g. first impression
?seriously ill - yes/no). Four weeks later I will check the record to
see what happened and gradually accumulate the necessary figures. I
could if I wish record all outcomes, and it would also be possible to
measure individual doctors' performance (although I won't be doing that
in this study).

In the long run it would be possible, by coding diagnostic findings or
impressions, to measure clinicians' performance against any important
outcome provided it is reasonably common. The essential feature of the
signs or symptoms is that they will have been identified by interviewing
clinicians, or by observing their practice in another way such as video.  

Cheers

Toby


-- 
Toby Lipman 
General practitioner, Newcastle upon Tyne
Northern and Yorkshire research training fellow

Tel 0191-2811060 (home), 0191-2437000 (surgery)


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%