Objective: To propose a methodology for characterizing proficiency in sputum smear microscopy for acid-fast bacilli (AFB) in the diagnosis of tuberculosis and to determine the number of microscopies necessary to establish this proficiency, as well as the quality of the transcription of results, the causes of the discrepancies in the readings (rater or microscope used), and the criterion for classification of microscopy results that poses the most difficulty in characterizing proficiency. Methods: Four hundred sputum smear microscopies for the diagnosis of tuberculosis were analyzed through double-blind readings by six professionals who usually read/supervise microscopies performed in public health care facilities. The sample was stratified to obtain, at least, a reliability of 90% in the double-blind readings, an α error of 5%, and a precision of 3%. The results were analyzed using observed reliability and the kappa index. Results: Thirteen errors (0.27%) were found in the transcription of results. Reliability increased when the three distinct categories of positive results (AFB+, AFB++, and AFB+++) were grouped or when inconclusive results were excluded from the analysis. The quantification of the bacterial load was the classification criterion that posed the most difficulty in establishing proficiency. Using higher quality microscopes increased reliability. Reliability values stabilized only from the reading of 75 microscopies onward. Conclusions: Double-blind sputum smear microscopy readings using a panel containing 75 slides (36 negative, 4 inconclusive, and 35 positive) proved to be appropriate for characterizing proficiency in sputum smear microscopy for the diagnosis of tuberculosis when such proficiency is intended to reproduce laboratory routine.

Keywords: Tuberculosis; Microscopy; Reproducibility of results.