Dear Paul
This is a familiar problem in psychiatry when doing "risk assessments" in the suicidal - clinicians often assume that this means having to predict the future and divide people into those at risk and those not at risk. (In other words risk factors often are used to "diagnose" suicidal people). The evidence (from cohort studies of diagnostic tools) is that we cannot predict who will kill themselves as the outcome is rare and the false positives swamp any meaningful connection. This doesn't mean however that there haven't been numerous risk factors identified for suicide. It seems in this instance that the purpose of risk assessments is to identify risk factors that you can do something about rather than to predict the future.
Is this an area where the literature on "clinical decision rules" is useful - a territory somewhere between prognosis and diagnosis?
Dr. Simon Hatcher
Senior Lecturer in Psychiatry
Department of Psychological Medicine
Faculty of Medical and Health Sciences
The University of Auckland
Private Bag 92019
Auckland 1
New Zealand
Telephone +64 9 373 7599 x86750
Fax +64 9 373 7493
Just say no to drug reps
http://www.nofreelunch.org/
-----Original Message-----
From: Evidence based health (EBH) [mailto:[log in to unmask]] On Behalf Of Paul Glasziou
Sent: Wednesday, 8 November 2006 7:10 a.m.
To: [log in to unmask]
Subject: Re: Risk Factors as Diagnostc Tests?
Dear Paweł
Interesting question - I don't think its a mistake and commonly use risk
factor information "diagnostically". For example, if someone comes in
with chest pain then myocardial infarction is a higher likelihood for me
if the person is 60 rather than 20 or has had a previous infarction. In
a chronic cough or hemoptysis, I'll consider lung cancer unlikely in a
non-smoker.
So intuitively that seems reasonable, but the maths also works the same.
Post-test Odds = Pre-test Odds x Likelihood Ratio
For test this is:
Post-test Odds = Pre-test Odds x Likelihood Ratio(test)
For risk factors this is:
Post-RF Odds = Pre-RF Odds x Likelihood Ratio(Risk Factor)
I'd be interested in others comments,
Best wishes
Paul Glasziou
> Dear Colleagues,
>
> I would like to ask you about common 'mistake' people made about risk
> factors (according to J. M. Last I define risk factor as exposure of
> behavioral pattern that improves a chance of developing the diseas in
> future). Namely, we frequently identify risk factors with some kind of
> diagnostic test, i. e. collecting patient medical history. Let's say there
> is 55 years old patient with smoking history in the past. Can we add the
> absolute risk of developing bladder cancer resulting from smoking up to
> other signs/symptoms, i.e. haematuria (blood in urine) to increase overall
> probability of the disease in diagnostic process?
>
> In my honest oppinion, this is incorrect. First of all, risk factor is by
> no means a part of constelation of symptoms that define and diagnose
> bladder cancer. I assume we cannot atribute past exposure to actual disease,
> because among others we didn't established the diagnosis yet and the risk
> of the disease can be drastically different that resulting from risk factor
> alone. Secondly, risk factors are not intended to be used as diagnostic
> test, because they're used for targeting preventive interventions.
>
> Do you agree with my assumptions? What is your oppinion / practice? Can you
> recommend any further reading covering this issue?
>
> Please forgive possible inaccuraces (I am not an epidemiologist).
>
> Best regards,
>
> Pawel Kanturski, M. A.
> Medical University of Wroclaw, Poland
>
>
--
Paul Glasziou
Director, Centre for Evidence-Based Medicine,
Department of Primary Health Care,
University of Oxford www.cebm.net
ph +44-1865-227055 fax +44-1865-227036
|