Posts made by Zoe Brooks
Yesterday I challenged you to: Try this: Select a QC chart and ask a selection of staff or colleagues to tell you the What, Why, Where, When Who and How of the values on the chart and QC rules applied.
Today’s exercise is to: Select a monthly summary QC report, from your Peer QC provider or your own internal spreadsheets. Now ask a colleague (or look in the mirror) and comment below. Look at the Bias, Total Error, Sigma Value, SDI and CVI. Do you really know what they mean? If you asked 10 people, how many gave the same answer?
Before you begin to assess a new method or examine QC data, you should be able to state for each statistical indicator: If it is below 'x' I will think or do 'this' and "if it is above 'y' I will think or do 'that'."
Laboratory professionals currently look at the same number and words, but reach vastly different conclusions and take different action. Let's face that truth - and change it!
Join the Scientific Advisory Board http://rmw.awesomenumbers.org/scientificadvisoryboard
There are many statistical processes to estimate bias or calculate sigma, but there is only ONE truth in the patient population. What do you think? What do you do? What is the best way to measure the truth?
It makes sense that if PATIENT samples produce the same results on four different instruments, then those instruments have the same bias, the same sigma value, and require the same quality control process to detect failure of acceptable risk criteria.
It does NOT make sense that each laboratory professional is free to choose different QC samples with no obligation to prove that they reflect patients, and then calculate bias as variation between the current measured mean and
[A] the peer group mean,
[B] the package insert mean, or
[C] the lab’s own historical mean or
[D] none of the above: don’t calculate bias at all - estimate it from any proficiency program using any samples at any time.
Aug. 14, 2018. Zoe Brooks
Take poll below and please discuss.
As I skimmed posts the past few morning and commented on a few, it occurred to me that all this valuable information is being lost. If you would you like to do something about that, then like this post and I will create the process. I have this naive notion that people can work together for the greater good, if we can create the right environment.
"AWEsome-U" is a full University level learning platform (Moodle) with the ability to create discussion forums, databases of literature and data, groups, projects, courses, etc. (Unfortunately, the operator (myself) is not as qualified, so please be patient.)
1. to validate the ability of Statistical QC / Risk Management software to improve patient lives and minimize healthcare costs.
2. to create a process to educate, implement and competency-certify staff to effectively use software to guide ISO/CLSI best practice for risk management.
3. to enable public transparency of risk and cost.
See the full article here https://www.incitedecisiontech.com/tutorials/Bayes_Male_or_Female_1.nb.html#
This will become very relevant as we get into creating probabilities in mixed populations. Robert also works mainly with non-Gaussian data, as most patient populations are when examined.
For now, we have this information when needed.
Question: What is the best way to access the information we need quickly? We can create a database with keywords and comments, where we break articles down by paragraph. Let's use this as a planning model.