Validating flagging criteria for automated cell counters - an institutional study

Context: Recently, automated hematology analysers have become available for haematological abnormalities in modern hospitals and clinics. With the advancement to flowcytometric counters, the analysis has improved by leaps and bounds with regards to white blood sub-population identification and detection of abnormal cells by indicating various flags. The appearance of flag just suggests a higher likelihood of an abnormality that can only be dismissed or established through slide examination. However, the factory thresholds may not be optimally set for the specific region and laboratory setting and need to modify to increase the yield and decrease the manual smear review. Methods: An observational study was carried out between the periods of May 2021 to October 2022 in a tertiary care hospital, karamsad. A 750 samples were reviewed for five WBC flags-Blast, immature granulocyte, left shift, atypical lymphocyte and abnormal lymphocyte/Lymphoblast on smear examination at factory default in an automated cell analyser. Results: We have calculated parameters of diagnostic accuracy stating Sensitivity, Specificity, Positive predictive value, Negative predictive value and Accuracy for each of the five flags and observed low specificity (Increased number of false positive) for blasts, atypical lymphocytes, and abnormal lymphocyte/lymphoblast. Conclusion: The flags with their low sensitivity and specificity does require modification in factory default threshold in order to decrease manual smear review rate and increase positive predictive value for each flag and to develop a systematic method of optimizing the flagging thresholds of automated cell counts, by combining statistical analysis and clinical judgment.