Professional Documents
Culture Documents
Diagnostic Models For Computing Measurement Uncertainty in Blood Bank Screening Tests
Diagnostic Models For Computing Measurement Uncertainty in Blood Bank Screening Tests
Diagnostic Models For Computing Measurement Uncertainty in Blood Bank Screening Tests
0 2 4 6 8 10 12 14 16 17 18 19 20
0 2 4 6 8 10 12 14 16 17 18 19 20
False-positive
Positive True-positive
results (FP) TP + FP
(Y=1) results (TP)
α-error
False-negative
Negative True-negative
results (FN) FN + TN
(Y=0) results (TN)
β-error
Total TP + FN FP + TN N
Positive b
a a+b
(Y=1) α-error
Negative c
d c+d
(Y=0) β-error
Disease
D1
Non disease Disease Non disease Disease Non disease
D0 D1 D0 D1 D0
1 1 1
ROCt
ROCt
Sensitivity
ROCt
The results show the relevance of alternative to GUM concepts for the evaluation of
post-transfusion risks components. They answer to chance questions, as well as to a
measurement trueness question unanswered by GUM models which are
inapplicable to ordinal quantities/binary results not dealing with probabilities. The
seroconversion window period, such as the 95% CIs represent chances of uncertain
results/false results arising from human biological variation as well as from the
analytical sensitivity of tests (biased).
© Paulo Pereira 17/20
Practicability and relevance of GUM approach for computing
measurement uncertainty in blood bank screening tests
Measurement uncertainty
determination
Measurement of
trueness or Probability
Measurement of probabilitty?
trueness Are patients and
Yes healthy subjects
No
sample available?
Diagnostic accuracy
Seroconversion
models when the Agreement of results
window period
diagnosis is known
Diagnostic accuracy
when cutoff varies?
Overall agreement,
Window period No positive agreement &
Yes Sensitivty & negative agreement
Receiver operating specificity
characteristic curve &
area under curve
Sensitivity, specificity & area under curve
Reported window period & probability of results on
patients and healthy subjects samples and populations © Paulo Pereira 18/20
Practicability and relevance of GUM approach for computing
measurement uncertainty in blood bank screening tests
Conclusions
The measurement uncertainty of screening immunoassays results can be
determined by alternative (to GUM) models. The presented models allow the
laboratory to claim the performance requirements for the measurement uncertainty
and these determinations could be regularly reviewed.