The reliability of the flagging System of the Coulter-STKS (Beckman Coulter
, Miami), an automated analyzer used worldwide to perform CBCs and differen
tial leukocyte counts, was assessed by determining the sensitivity, specifi
city, efficiency, predictive value of normal (PVN), and predictive value of
abnormal (PVA) for individual suspect-flags and regional-flags generated b
y the analyzer. Automated differentials were compared with manual different
ials, initially from the data collected prospectively in a random populatio
n of 647 blood specimens. Low prevalence of several flags in this specimen
population led us to retrospectively review hundreds of additionally flagge
d automated differentials along with the corresponding manual differentials
to gather 100 data paints per qualitative abnormality or pag, to better de
termine sensitivity and PVA. Each automated differential was assessed to be
either true normal or false normal or either true abnormal or false abnorm
al, based on predefined microscopic criteria. Of the evaluated flags, immat
ure granulocyte/band 1 (IGB1) and immature granulocyte/band 2 (IGB2), the t
wo most prevalent flags (31.6% and 10.8%, respectively), had the lowest eff
iciency (59%-68%). Specificity and PVN for all flags other than IGB1 and IG
B2 were within the desirable range, although sensitivity and PVA were at le
ss than desirable levels. Enhancement of sensitivity and PVA of many, if no
t all, flags, and specificity and PVN of IGB1 and IGB2 flags will increase
the reliability and consequently the cost-effectiveness of the STKS automat
ed flagging system.