The "diff-if". Use of microcomputer analysis to triage blood specimens for microscopic examination.

Abstract

Increasingly, all automated blood counts are not accompanied by a microscopic white blood cell differential. A popular strategy is to obtain a manual differential if any part of the automated blood count and differential is outside specified limits (the "diff-if" strategy). The authors compared two sets of criteria to triage blood counts for manual differentials: previously recommended numeric values, and the analysis of a microcomputer program. In a population of subjects with a high percentage of hematologic disorders, the microcomputer program and the numeric criteria were equally specific (excluding normal blood smears); the program was more sensitive for bands, immature granulocytes, monocytes, nucleated red blood cells, reticulocytosis, teardrops, red blood cell fragments, and hypersegmented neutrophils. The numeric criteria were more sensitive for eosinophilia (less than 1.0 X 10(9)/L) and mandated fewer manual differentials. In a population of predominantly normal subjects, the program was more sensitive for increased bands and equally sensitive for eosinophilia, the only abnormalities observed on the smear. In a population of subjects with predominantly abnormal blood counts, but excluding most primary hematologic disorders, there were few blood smears with abnormalities beyond eosinophilia or increased bands. In both of these groups, the computer program mandated more manual differentials than did the numeric criteria. The authors conclude that microcomputer analysis by the program tested was more sensitive than numeric criteria to identify specimens with abnormal blood smears. Specificity depended on the patient population. The choice of a triage strategy should be based on the individual laboratory's patient population.