Decreasing insensitivity deals only with false negatives. There is always a
risk that there will be an accompanying reduction in specificity which could
lead to an increase in the number of false positives exceeding the decrease
in false negatives. Net change in clinical value depends upon the benefit
from reduction in false negatives and harm occasioned from any increase in
the false positive group.
The general principle of measuring change in harm is, I agree, meritorious.
It may not, however, always lead to successful argument for method
improvement. For example, if we change from a second to third generation
method the cost per benefit is the overall cost difference divided not by
the total number of correct assignments but by the change in incorrect
assignments. Thus suppose the proportion of case positives among those
tested were 50% and the new assay improved the pickup from 96% to 98%
without any change in specificity the net cost per benefit would be
100*(difference in assay cost). Convincing administrators that this is
worthwhile may not be so easy.
Trevor Tickner,
Norwich
> -----Original Message-----
> From: Sten Öhman [SMTP:[log in to unmask]]
> Sent: 30 November 2001 08:23
> To: [log in to unmask]
> Subject: Accuracy etc
>
> >Dear All,
>
> There are a lot of words used either to describe the technical performance
>
> of a laboratory method or its ability to confirm/exclude certain diseases.
>
> Such word are:
>
> Sensitivity, specificity, accurracy, precicion etc.
>
> When first I began to evaluate methods I had knowledge in analytical
> chemistry and statistics but I was not well-informed in the nomenclature
> of
> clinical chemistry.
>
> By intuition, therefore, I first suggested to use the opposite of all
> these
> words, i.e. insensitivity, inspecificity, inaccurracy and imprecision, but
>
> soon I was aware that this was not in accordance with current practice.
>
> Logically these words are complementary. For example sensitivity
> constitutes a probability level p, where insensitivity is the complement:
> q
> = 1 - p.
>
> Increasing the sensitivity from 96 to 98 percent does not seem to be
> important, but this means that the insensitivity decreases from 4 to 2
> percent which reduces the number of faulty answers with 50 percent.
>
> Although, strictly logically, the same information, reporting a decrease
> of
> the number of faults should better than reporting an increase the number
> of
> successes indicate the value of any method improvement and hence better
> convince administrators etc to choose the better method.
>
> Best wishes
>
> Mr Sten Öhman, PhD
>
>
---------------------------------------------------------------------------
The information contained in this e-mail is confidential and is intended
only for the named recipient(s). If you are not the intended recipient you
must not copy, distribute, or take any action or reliance on it. If you have
received this e-mail in error, please notify the sender. Any unauthorised
disclosure of the information contained in this e-mail is strictly
prohibited.
|