[Evaluation of 5 selection criteria for sputum as a valid sample for culture]

We compare five selection criteria to discriminate between good quality expectorated sputum and specimens contaminated with upper respiratory tract secretions. We employed two microscopic methods (Bartlett, Heineman) which rely upon the number of squamo-epithelial cells and polymorphonuclear leucocytes; and a macroscopic method (the Medical Research Council) which evaluates the gross appearance of sputum, with three different criteria for admission established by our own. Three hundred and thirty two sputum were included. We evaluated sensitivity, specificity and predictive values as compared with isolation or not of a potential pathogen. We obtained 72% and 45% sensitivity, and 81% and 86% specificity for Bartlett and Heineman methods respectively. Most restrictive criterion for macroscopic method provided 57% sensitivity and 82% specificity; least restrictive showed 97% sensitivity and 17% specificity. Macroscopic method is a quick, valuable screening for discarding sputum contaminated with upper respiratory tract secretions.