Evaluation of BCDE, a microcomputer program to analyze automated blood counts and differentials

A microcomputer program (BCDE) has been developed to analyze automated blood cell counts and differentials' similarity to normal values or to 36 disease categories. In 50 normal subjects, the analytic program listed the correct diagnosis as the first diagnosis in 49 cases (the only diagnosis in 44) and second of two diagnoses in one case. In 182 subjects with known hematologic disorders, the correct diagnosis was listed first in 134 and second or third in an additional 40. Subjects with iron deficiency, heterozygous thalassemia, immune thrombocytopenia, anemia of chronic disease, reactive thrombocytosis, acute infection, and chronic leukemia had the disorder identified as the most likely one by the analytic program with both sensitivity greater than 80% and specificity greater than 98%. Subjects with acute leukemia, folate deficiency, sickle cell anemia, cytotoxic chemotherapy, and chronic liver disease had the disorder identified as most likely by the program with a sensitivity less than 80%. In a different 11 cases with known hematologic status, a panel of 37 physicians identified the disorder(s) or normality only 72% of the time, whereas the analytic program listed the correct diagnosis first in 10 of 11 (91%). The analytic program appears useful for both triage of normal from abnormal data and for the initial differential analysis of abnormal data.