Regarding the number of significant figures to which to report results,
I agree with Roger Ekins that an analytical result SHOULD be reported with
measurement uncertainty attached. I gather this has been tried many years
ago in a Sydney laboratory, and I am sure in other locations, and was very
poorly received by the clinicians. The increased complexity of the reports
compared to the small amount of useful additional data for clinical
decision-making provided by the added expression of uncertainty lead to its
rapid abandonment.
In general I would agree with Jonathan Middle's response about
understanding the numbers, ie 123.4 indicates a number with implied
analytical accuracy such that the result is likely to be between 123.35 and
123.45.
The two problems I have with this are firstly the implied imprecision
changes as the numbers change, for example a result of 1.1 implies a CV of
(1.15 - 1.05)/(4 x 1.1) = 2.2%; whereas the same number of significant
figures at a different numerical value, eg 1.9, implies a different CV
(1.95 - 1.85)/(4 x 1.9) = 1.3%.
The second issue is the expression of significant figures with large
numbers, eg does 12,300 mean 12,300 +/- 0.5; +/- 5 or +/- 50. There are
ways to make this unambiguous (eg 1.23 x 10^4) but this again would be
difficult for most laboratory computer systems as well as the receiving
doctors.
The paper referred to by Leslie Burnett is of interest in trying to address
this issue. The authors (from the faded yellow pages on my top shelf)
recognise the difficulty in obtaining the data for making the calculation.
Some brief modelling shows that the number of decimal places (significant
figures) changes within a decade. eg Modelling urea with an analyical CV of
2% and a within-person CV of 11%, values up to 6 mmol/L require 1 decimal
place, but above 7 require no decimal places (model tests from 1 to 10
mmol/L). This would be very difficult for most lab computer systems.
Additionally I have a problem with setting 95% certainty levels
arbitrarily, a clinician may be happy with 80% certainty for a clinical
decision, and even the difference between 1 and 2 tailed probability (of
there is an expected direction of change or not) will change the calculation.
Also measurement of Lab CV should consider between-instrument CV if a
patient may move between laboratories.
So in the absence of a decent theoretical construct we need to rely on that
lowly form of decision making: expert opinion. My two bobs worth would be
as follows:
Creatinine: report to the nearest 10 umol/L (0.01 mmol/L) for all results.
Urea: report to the nearest 0.1 mmol/L up to 10 mmol/L then to the nearest
1 mmol/L.
If your LIS can change the number of significant figures other than changes
in decades other possibilities may be considered.
Note that extra significant figures may be very useful for QC purposes.
As this is my "expert opinion" I am sure there are many others. I look
forward to hearing other ways of approaching this issue.
Regards,
Graham
Dr Graham Jones
Staff Specialist in Chemical Pathology
St Vincent's Hospital, Sydney
Ph: (02) 8382-9160
Fax: (02) 8382-2489
**********************************************************************
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom they
are addressed. If you have received this email in error please notify
the system manager.
This footnote also confirms that this email message has been
virus scanned and although no viruses were detected by the system,
St Vincent's Hospital accepts no liability for any consequential
damage resulting from email containing any computer viruses.
**********************************************************************
------ACB discussion List Information--------
This is an open discussion list for the academic and clinical
community working in clinical biochemistry.
Please note, archived messages are public and can be viewed
via the internet. Views expressed are those of the individual and
they are responsible for all message content.
ACB Web Site
http://www.acb.org.uk
List Archives
http://www.jiscmail.ac.uk/lists/ACB-CLIN-CHEM-GEN.html
List Instructions (How to leave etc.)
http://www.jiscmail.ac.uk/
|