Dear Dan,
There seems to be a slightly different question (or answer) lurking
here, though. Before we consider the confidence intervals, etc. around
the population estimate, we might want to consider 'how much' the
patient belongs to the population studied. At least in theory, this
would be amenable to a mathematical approach (although I suspect you
would end up making a fair number of judgement calls).
This seems (to me) to be slightly different from asking about how we
individualise the answer - it seems to be asking an earlier question,
which is about applicability. They're linked, but I'm not sure they're
the same.
Matt
Paul Glasziou wrote:
> Dear Dan,
> An excellent question! There has been quite a bit of interest in this in
> the past decade. An early paper was:
> Glasziou PP, Irwig LM. An evidence based approach to individualising
> treatment. BMJ. 1995 Nov 18;311(7016):1356-9.
> which suggested a 4-step process: you need to work out the individual
> prognosis and then apply the relative risk of the study (and then think
> about harms!).
> A more recent and thorough exploration is the book edited by Peter
> Rothwell: Treating Individuals From randomised trials to personalised
> medicine, Elsevier · Published July 2007.
> Best Wishes
> Paul Glasziou
>
>
> Olive Goddard wrote:
>> Dear Colleagues,
>>
>> If anyone can help Dan Hughes I should be grateful if you would
>> respond to him directly.
>>
>> All good wishes,
>>
>> Olive
>>
>>
>>>>> "Dan Hughes" <[log in to unmask]> 04/01/2008 13:53 >>>
>>>>>
>> Olive, I wonder if anyone in the group might be aware of a paper or a
>> good discussion on the issue of applying evidence to a "single" patient
>> in the clinic or at the "coal face".
>> The issue is how to apply the "evidence" that we find in our searches
>> to a single patient. I am assuming that we have already added in
>> physician expertise and patient values. Kathryn Montgomery calls this
>> generalization and particularization or lumping and splitting in her
>> book.."How doctors think..." Someone else refers to the issue as
>> "misplaced concreteness". Stephen Jay Gould touched on the subject in
>> his essay.."The median isn't the message". He says too much emphasis is
>> placed on the treatment effect value rather than the variation. As we
>> know he lived more than 20 years after being diagnosed with an abdominal
>> mesiothelioma which, at the time, had a median survival of 8 months. He
>> obviously was in the tail end of a right skewed curve. Standard
>> deviations tell us the variability within a specific sample
>> and standard errors the variability of means with many samples.
>> Confidence intervals give us an idea of where our study results fit in
>> the "real" population, but where does our single patient fit?
>> Is there an "interval" where we can say our one patient might fit? How
>> do we translate the results from a particular study into something that
>> a patient can understand? Is this something the Knowledge Translation
>> folks are dealing with?
>> I await with great anticipation. Thanks , Dan.
>> Dr. Dan Hughes
>> IWK Health Centre
>> Children's Site-1st floor
>> Box 9700
>> 5850/5980 University Ave.
>> Halifax,N.S. B3K 6R8
>> Ph.(902)470-8218
>> Fax.(902)470-7223
>> E-mail: [log in to unmask]
>>
>>
>
--
http://acl.icnet.uk/~mw
http://adhominem.blogsome.com/
+44 (0)7834 899570
|