Distinction of microcytic disorders: comparison of expert, numerical-discriminant, and microcomputer analysis

Presumptive distinction between iron deficiency and heterozygous thalassemia by analysis of the automated blood count and differential continues to be a challenge. We compared two proposed numerical discriminants (MCV2 x MCH, and MCV2 x RDW/100 x Hb) with an analytic microcomputer program (BCDE2 Lea & Febiger). In 7114 subjects, the numerical discriminants and the BCDE2 program correctly identified greater than 90% of thalassemia. In subjects with iron deficiency, the BCDE program was greater than 90% sensitive and specific for positive identification, while the numerical discriminants were less than 75% sensitive and specific at inferential identification. The numerical discriminants, BCDE2, and 17 experts in blood counting were asked to interpret the blood-count data in 7 fully-defined actual cases. The mean experts' score was 5.65 cases correct out of 7. The BCDE program was correct in all cases. The numerical discriminants could not analyze all cases, and both were incorrect in at least one case. We conclude that for the task of analyzing blood counts for microcytic disorders, microcomputer analysis by BCDE outperforms both numerical discriminant functions and analysis by expert hematologists.