Differential identification of mycobacteria in smears of sputum

As a test of their ability to differentiate mycobacteria in Ziehl-Neelsen stained smears of sputum, 4 readers examined an assortment of 200 smears, of which half were known to contain Mycobacterium tuberculosis and half were known to contain atypical mycobacteria. In 2 separate trials, readers made a prediction of either 'typical' or 'atypical' for each smear. Reader consistency between trials ranged from 69.5% to 86.5%, indicating that readers were basing their predictions on definite criteria. Overall accuracy ranged from 62.5% to 80%, with experienced readers being the most accurate. Reliability of readers' predictions was influenced by smear positivity, but certain smears exhibited uncharacteristic features which led to incorrect predictions. When considered along with other available data, smear-based presumptive diagnosis can assist the clinician in his management of the patient with suspected tuberculosis in the period before culture results are known.