Perhaps I did not make myself clear about your attitude to statistics.
Whilst I agree that I cannot experience your personal attitude towards
statistics (which obviously is an animate process unique to you as an
individual) I can comment that your 'attitude' in terms of how statistics
are applied to everyday clinical situations and perhaps more importantly how
you interpret statistical occurrence is not commensurate with the commonly
accepted and widely practiced model of biostatistics that you may find in
any standard medical or statistical textbook or indeed that you may question
any competent bio/statistician on.
Therefore the question is not whether I believe you are telling the truth or
not but instead to challenge your assumptions and interpretations of
statistics. This includes how you do (or perhaps don't) apply statistical
conventions in your clinical practice and even more importantly the fact
that you do not seem to be aware of the potential implications of your
approach. Unfortunately you are not alone as the misuse, abuse or even
omission of the important role of statistics in practice is endemic.
Clinicians are 'clinicians' but they also require a sound understanding of
these issues to deliver the best and most appropriate treatment to their
patients. For a very good introductory overview to this issue read Chapter 1
in Cambpell MJ, Machin D, Medical Statistics- A commonsense approach, John
Wiley & Sons, 3rd ed in 1999.
I fear that even having a research background is not a guarantee to
understanding and appropriately applying statistical knowledge. If you do
not agree with this I would ask you to seek any editor, statistical reviewer
or internal/external referee of a peer reviewed biomedical journal and ask
them how many papers they reject on the basis of statistical errors alone. I
know for a fact that a sizeable proportion of papers submitted to the Lancet
and the BMJ are rejected outright before they even reach the editorial team
as the statistical referee gets first look to spot all the non-correctable
statistical errors that render the paper invalid. This problem has also been
given exposure in the literature in the various solicitations from those
desperate for better research and better clinical care; see any of Trish
Greenhalgh's work or Doug Altman's classic editorial in the BMJ
(1994;308:283-284 - available on bmj.com).
Therefore I do argue that your approach to statistics is harmful by the very
nature of your approach which in postings to this list reveals several
fundamental misunderstandings or mis-assumptions about statistical inference
and interventions on both an individual and population level. Of particular
concern is your continued belief that statistics does not concern itself
with the individual- this reveals the extent of your unfamiliarity!). I
continue to argue these points as they apply to real patients in real time
and although may be conceptualised (just like states of health and disease
which is why we are in clinical practice) they are most definitely not
limited to just theoretical or philosophical debate.
I work within a major University department of the UK NHS R&D programme and
I am a strong advocate, like all my colleagues are, of improving the quality
of i) health services research and ultimately through this ii) clinical
practice (which is the whole point of doing research!). Inherent in this
quality improvement process is an appropriate, an enlightened and a reasoned
approach to at least understanding the clinical decision making process from
a perspective other that of 'clinical blindness' or 'clinical tunnel
vision'. Your approach to statistics does not appear to endorse this.
Unfortunately this discussion list is not an appropriate medium for
illustrating these important issues in further detail (and I fear our
quibbles must be growing tiresome for our colleagues on the list) however
perhaps I could suggest you investigate the excellent UK Health Technology
Assessment Methodology Monographs (available online at
http://www.hta.nhsweb.nhs.uk) and also the high quality Cochrane
Collaboration Methodology Register which is now widely available (although
you will need a password for the online version) which contain a series of
papers that explain in detail but with clarity these points.
Whilst I suspect that you may never embrace the statistical concepts that I
have attempted to outline here I do urge you to consider more carefully your
somewhat reckless approach to clinical decision making. Like it or not or be
aware of it or not your decisions and the process by which you get to those
decisions have implications on your patients. I think Gina Radford, one of
the first directors of the UK National Institute for Clinical Excellence
(NICE) summed it up when she said "nobody in healthcare goes to work to do a
bad job". Of course how you interpret whether somebody is actually doing a
good job or not is the fundamental premise of what we are debating…
Alistair Grant
Institute of Public Health
University of Cambridge
>From: "Stanislav A. Korobov" <[log in to unmask]>
>Reply-To: PHYSIO - for physiotherapists in education and practice
> <[log in to unmask]>
>To: [log in to unmask]
>Subject: Re: The Myth of Homeostasis (former; currently rather:
> "Statistics in Physiotherapy")
>Date: Tue, 23 Jan 2001 12:11:39 +0000
>
>You wrote: <Unfortunately I cannot agree with your claim that you
>“have a respectful attitude towards statistics including statistics
>in
>biomedicine”.>
>
>Your claim about my claim seems to me strange. You can agree or cannot
>agree
>but my attitude is mine. If I have an attitude to a thing, then I have it.
>And if I say that my attitude is respectful, it means that I feel and deem
>and assess it exactly so. Respectfulness, as far as I know English words,
>is
>a sense. Excuse me, do you know my senses better than I do? Your inability
>to agree with my claim is meaning for me only one thing -- you do not
>believe that I told the truth about my sense. Interestingly...
>
>I am forced to repeat: my attitude to statistics in biomedicine is
>respectful. Even very respectful. The main reason of this is probably my
>researcher's background. I had many statistical calculations when dealt
>with
>the results of my investigations. And I saw, of course, that statistics is
>an useful tool, and in order to make a generalized conclusion as to a group
>of objects we cannot dispense with statistics' assistance. And that
>statistics often helps us to see such things that are hidden if each object
>is considered separately.
>
>However I think that YOUR misunderstanding the situation is that you
>attempt, as far as I can coclude from your texts, to look at any separate
>phenomenon as a particular case of an appropriate distribution curve. I
>think it is a mistake. Any curve obligatorily has the restrictions due to,
>at least, the initial mathematical assumptions and the limited quantity of
>observations taken into consideration. In principle, you can make any your
>statistical conclusion only in regard to those particular points which do
>form this curve! Any additional (new) points need, first of all, to be
>proved as those pertinent to this curve. Are you sure that any new object
>is
>condemned to fall within the curve (no matter -- at tails or at the
>middle)?
>If yes, you risk to make the fatal for this object mistake. Of course, a
>likelihood of such event may be negligible but it always exists!
>
>Your mistake is also a philosophical one. Indeed, dialectics does teach us
>that there are unity and struggle of contraries. One thing is
>simultaneously
>general and particular, big and small, bad and good. As to the subject of
>our discussion, each our separate patient or his/her separate index is
>simultaneously belonging to the whole population and strictly individual! I
>think you will agree that EACH given organism is definitely unique and its
>combination of components is non-replicated. If so, have you a right to
>look
>at this organism as one of known-to-you points of the curve (even of those
>of its tail)? I suppose you have not such a right. You may only ASSUME that
>this object MAY belong to the group of those known points. To determine
>definitely whether this is true you need to make ALL accessible particular
>measurements and calculations, and only AFTER this you have grounds to say
>PRESUMABLY about such belonging. As well as grounds to make a clinical
>decision (keeping permanently in mind that you are about a failure of this
>making!). Consequently, when you are at the beginning of dealing with a
>patient and his/her data you are only AROUND the curve but not oligatorily
>at it (surely a likelihood of all appropriate events may be more or less;
>it
>is already a quantitative aspect of the problem; I am interested more in a
>qualitative analysis of the situation). So you should be very careful with
>regard to your conclusions based on your knowledge of appropriate
>statistical curve. Otherwise, you risk to make an unpredictable clinical
>error.
>
>You write about contradictions in my line of argument. The drama is that
>the
>life is contradictive in all its aspects! In ALL. It seems, our world was
>initially made contradictive. If you deem that you have not contradictions
>in your past, present or future considerations, a disillusion does wait for
>you.
>
><Your attitude to separation of statistical interpretation in clinical
>trials and the clinical decision making process is harmful...>. I really
>separate these things (although they have close relations, of course). You
>may see above -- why. And I think that uniting these processes is more
>harmful because it gives a temptation to classify easily (but a priori) a
>clinical case as such that conforms to the results of above-mentioned
>trials. This is a huge mistake. In a given case, we can only assume that
>such conforming is possible with a certain likelihood.
>
>Statistics really does not take an interest in a separate individual. If
>you
>think otherwise, please provide me with an example of statistics when n=1.
>Statistics needs in many (or, at least, several) individuals or individual
>indices to make its conclusions. Isn't? Hereof is my statement that
>statistics cannot live without groups of digits of the same kind (I meant
>groups of results of uniform measurements; sorry if my English is
>incorrect).
>
><John Smith as a ‘sample’ of the wider population must lie
>somewhere within a distribution curve...>. Why MUST?? About 15 years ago I
>had a patient who MUST rise his body temperature when I treated him with
>large applications of warm peloid (mud). At least, the most of people did
>exactly so. But he was not wishing to rise it. Absolutely. And I do not
>know
>why up to today (methodical failures are rather excluded since we had
>measurements repeated using different investigators and different patients
>to compare). Maybe he was without of all known laws of physics? Hardly.
>Then
>why? Maybe because he was, using your words, from the tail of the curve?
>Maybe. However I have not a relief from such an 'explanation'. And should
>I
>look at him as a point of that curve (even -- from its tail, or the end of
>the tail) I would get his heat shock definitely.
>
>Perhaps your key misunderstanding is best reflected by the following your
>words: <this is not a ‘philosophical’ situation but a real life
>situation>. Until you have misunderstood that EACH our 'real life
>situation'
>is exclusively a philosophical one, you will make a mistake of too
>schematic
>(mathematical, formalizing; at any event -- simplifying) look at a
>phenomenon. Unfortunately (fortunately?) the life is wider and deeper and
>--
>it is also obvious -- less understood than ANY its statistical description.
>Even -- generated by the best unit in Europe or in the world.
>
>I think we, clinicians, should indubitably use all available statistical
>data and approaches relevant to our separate clinical cases. But we MUST
>make our crucial clinical decisions only using all our opinions, senses,
>experiences and intuitions.
>
>Stanislav A. Korobov, MD, PhD
>Physician-Physiotherapist
>'Mountain Air' Study,
>'Lermontovskii' Clinical Sanatorium.
>PO Box 7, Odessa, 65089, Ukraine
>[log in to unmask]
_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
|