Diagnostic strategies used in primary care
BMJ 2009; 338 doi: https://doi.org/10.1136/bmj.b946 (Published 20 April 2009) Cite this as: BMJ 2009;338:b946
All rapid responses
Maude's response of diagnostic checklists may be of value, but a
surgical pre-op checklist is quite different from diagnosing in the real
world the presentation of uncertainty. Indeed straw-poll arguement that
diagnosis is major problem in primary care - something must always be "the
leading cause of complaints".
Diagnostic checklists can only take us only so far - eventually the
generalising processes of EBM must be applied to the one specific patient
in front of us, and with checklists never giving absolute diagnostic
certainty, so there will remain doubt as to the correct action for that
specific patient. Perhaps Maude meant to suggest a Consultation checklist
(eg of whether diagnosis given, reassurance of conditions ruled out, of
treatment options and what realistically to expect of them and of
contingency action). A checklist is not the right medium for addressing
cost-benefit of a wrong diagnosis and what is a wise/prudent clinical
approach, rather than what a well programmed robot can tick off from a
checklist.
I have on rare occasions encountered patients relieved of their chest pain by GTN spray, despite subsequent normal coronary angiogram, for it may also relieve oesophageal spasm in GORD. Similarly in patients presenting with onset of headache, neck stiffness is an unhelpful sign if they have pre-existing neck arthritis and pain.
Competing interests:
None declared
Competing interests: No competing interests
The diagnostic steps described by Henegan et al [1] are also used by
hospital doctors. The detailed history and examination model is useful to
beginners but is also useful when a doctor sees a patient with multiple
problems and no detailed past records. If the patient’s background and
other diagnoses are already well known, in hospital or general practice,
the process can be focussed more easily on a single problem.
The basic aim of diagnosis is to show that the patient is a member of
a group with findings, within which it is helpful and safe to take some
action (e.g. reassuring, admitting to hospital, prescribing some
medication or operating). A probable or definitive diagnosis becomes
final when the outcome is accepted by the patient and advising team. In
many cases the final diagnosis is ‘some unknown process that has resolved
with no active intervention’. Until the diagnosis is final, in primary or
secondary care, a differential diagnosis has to be kept in mind in order
to avoid a diagnostic error due to premature closure. It is always
important to start with the presenting complaint (a correct spot diagnosis
of eczema would be silly in a patient wanting help because of chest pain).
The appropriate model for all this is not Bayes theorem alone.
Another ‘theorem’ is also needed to explain how differential diagnoses are
formed from presenting complaints or other triggers (also called
‘diagnostic leads’ [2]) and how other information are used to
differentiate between them (also called ‘differentiators’ [2]). The
arithmetic of this process needs to be understood in order to practice
evidence-based medicine properly [2]. All this can be taught to students
and trainees so that the diagnostic process can be practiced in an
evidence-based, cost-effective way and the reasoning explained to patients
and other members of the team in writing.
[1] C Heneghan, P Glasziou, M Thompson, P Rose, J Balla, D Lasserson,
C Scott, and R Perera. Diagnostic strategies used in primary care. BMJ
2009; 338: b946
[2] H Llewelyn, H Ang, K Lewis and A Al Abdullah. The Oxford
Handbook of Clinical Diagnosis, 2nd edition. Oxford University Press,
Oxford 2009.
Competing interests:
Huw Llewelyn is an author of the Oxford handbook of Clinical Diagnosis
Competing interests: No competing interests
Heneghan et al provide a thoeretical construct for diagnostic
reasoning by GPs. However, I submit they have used a cognitive forcing
strategy themselves so missing at least one other strategy: 'pattern
failure'.
Regrettably, a search on google scholar tonight failed to identify
any published literature on this phenomenon. I will explain:
Consider a LCD monitor screen. Almost all the time there is no
problem. However, if a pixel is not working, the eye is drawn to it. It
will take time to bring to consciousness what the problem is, but one is
aware immediately.
Or think of a bobby on the beat. She passes the same houses day in
and day out for weeks, months and even years in the past. One day she is
on her beat and feels uncomfortable. She turns back and realises a front
gate which is usually closed is open. Investigation reveals a crime has
taken place.
What I am getting to, is that for some GPs the null hypothesis is the
patient before them has nothing (seriously)wrong. The decision making
process is about refuting this hypothesis by using history and examination
to attempt to identify disordered physiology/anatomy. In the absence of
such an abnormality, one can be reassured.
This is, I believe, a distinguishing feature of first contact
doctoring. It is a skill which, I believe, reduces referral to secondary
care.
General Practice consultations should, in my opinion, conclude with a
statement about what is not the problem (often allaying worrying
ideas/concerns/fears). Making a diagnosis is rare in unscheduled GP
visits; this should be acknowledged.
This is at the core of managing uncertainty in general practice and,
I believe, should be encouraged and supported. Heneghan et al describe a
process which may undermine this principle.
Competing interests:
None declared
Competing interests: No competing interests
The authors of this paper must be congratulated for this clear
analysis and
formal recognition of how the first and most important decision is made in
the patient's journey. The fact that the Healthcare Commission highlighted
that diagnosis has been the leading cause of complaints in primary care
for
the last 3 years, accounting for around 25% of all complaints, shows that
this
focus on diagnosis is long overdue. The King's Fund's recent announcement
of an 18 month inquiry into the quality of General Practice with a
particular
focus on the quality of diagnosis and referrals is also to be welcomed.
This study highlights that in about 70% of consultations the
diagnostic
strategy relies on a hypothesis, list or pattern derived from the GP's
memory.
The stakes are far too high for this vital first step, which determines
subsequent treatment and medication, to depend so heavily on one person's
memory. Medicine has got much too complex for that.
The use of surgical checklists has recently been mandated by the CMO
and
are being implemented by the NPSA as their use has been shown to
dramatically reduce morbidity and mortality. Checklists are particularly
useful
for processes that rely extensively on memory as the authors of this paper
have just shown to be also the case for diagnosis in primary care.
Technology
has now made it possible to deliver, at the point-of-care, an instant
patient
specific checklist for the cognitive process of diagnosis decision making
and
they should now be used routinely.
Competing interests:
Jason Maude is CEO and Co
Founder of Isabel Healthcare which
produces the Isabel Diagnosis
Checklist system
Competing interests: No competing interests
Dr. Heneghan and his colleagues agree that the hypothetico- deductive
method rather than the laborious inductive method is most commonly used by
clinicians during the process of making a diagnosis.1 They describe three
stages (initiation, refinement and definition) and various strategies used
in the diagnostic process. What is less clear from their article is that
the history is of prime importance in the genesis of initial hypotheses as
is what as general practitioners we know about the patient already from
personal knowledge or patient records. Furthermore we must not forget
that hypotheses are erected in physical pschological and social terms, the
so called triple diagnosis. Initial hypotheses reflect probability,
seriousness treatability and novelty. Hypotheses can be revised and re-
ranked by further history taking, focussed examination selective further
investigation and of course the use of time. Categories such as most
likely, less likely and of course rare but important then emerge which can
be tested and further refined for management.
I would like to emphasise that these ideas are not mine, nor are
they exactly new as generations of medical students who have been through
the Professorial Department of General Practice at Leicester University
will be aware. Indeed others who have been involved in post graduate
medical education and have taught clinical reasoning to GP registrars
through the years will also be very familiar with them Furthermore those
of us who have been concerned about formative and summative assesment in
the consultation will have used these ideas clinically in the form of the
Leicester Assesment Package both in this country and abroad.
May I recommend a text first published in 1987 which describes these
ideas in full.2
1. Heneghan C Diagnosis in General Practice BMJ 2009;338: b946
2. Fraser R The Diagnostic Process p36 in Clinical Method, a General
Practice Approach (ed Fraser R) Butterworth Heinemann third edition 1999.
David Brooks MD FRCGP
Competing interests:
None declared
Competing interests: No competing interests
Diagnostic strategies in primary healthcare: a simple approach to making a diagnosis
Michael Porter, Professor at Harvard Business School, says ‘The
essence of
strategy is choosing what not to do.’
Diagnosing what disease a patient has is important but observing and
making
sense of the different ways our patients minds work is one of the
keystones
of general practice. Hence (as Osler observed) we need to know what sort
of
patient has a disease more than what sort of disease a patient has and
that
whilst a good physician treats the disease; the great physician treats the
patient who has the disease.
Perhaps doctors do decide to use fourteen formal strategies in three
different
stages of diagnosis (reference 1) but perhaps it’s simpler than that, and
all we
need to do is decide what not to do.
James Reason describes a model (reference 2) of how the mind works.
We all
live in the immediate world of our ‘conscious workplace’. This is what you
are
aware is in your mind right now. Constantly and automatically our
conscious
workplace accesses our long-term memory store for additional information.
Long-term memory is searched for the missing information using one of two
processes either similarity matching or frequency gambling. Similarity
matching produces more robust results. For instance, search your mind for
a
pet that has four legs and barks and you are likely to think of a dog.
Frequency gambling produces less reliable results. Search your mind for a
furry brown animal and some furry brown answer will immediately pop into
your consciousness. What it is depends on your own experience of furry
brown animals. Frequency gambling is the process that results in us
misdiagnosing meningitis as flu in a flu epidemic. You may get a strong
but
wrong answer.
A simple approach to diagnostic strategy is to be aware if you used
similarity
matching or frequency gambling in making the diagnosis. If it is
similarity
matching you can proceed with confidence, if its frequency gambling you
must choose not rely on it and instead get more information so that
similarity
matching can provide the answer.
William Osler’s advice: ‘No human being is constituted to know the
truth, the
whole truth, and nothing but the truth; and even the best of men must be
content with fragments, with partial glimpses, never the full fruition’.
References
1. C Heneghan, P Glasziou, M Thompson, P Rose, J Balla, D Lasserson, C
Scott, and R Perera. Diagnostic strategies used in primary care. BMJ 2009;
338: b946
2. The Human Contribution: Unsafe Acts, Accidents and Heroic Recoveries
by James Reason. Ashgate Publishing Ltd 2008
Competing interests:
None declared
Competing interests: No competing interests