Background: Even though most differential leukocyte counts are performed by automated hematology platforms, turn-around time is often prolonged as flagging of test results trigger additional confirmatory manual procedures. However, frequently only the absolute neutrophil count (ANC) is needed. We therefore examined if an algorithm could be developed to identify samples in which the automated ANC is valid despite flagged test results.
Methods: During a 3-wk period, a training set consisting of 1448 consecutive flagged test-results from the Sysmex XE-2100 system and associated manual differential counts was collected. The training set was used to determine which alarms were associated with valid ANCs. The algorithm was then tested on a new set of 1371 test results collected during a later 3-wk period.
Results: Analysis of the training set data revealed that the ANC from test results flagged with the 'WBC abnormal scattergram' alarm was often invalid. After removal of these test results, the correlation between automated flagged and manual results was excellent (r = 0.95-0.99, P < 0.0001). The validity of this algorithm was confirmed in a subsequent analysis using an independent test data set.
Conclusion: We have developed a simple algorithm that identifies samples from which a valid ANC can be extracted from Sysmex XE-2100 flagged differential counts. In patients where only ANC is required for treatment purposes, this procedure can greatly shorten turn around time and facilitate rapid decision making.