BACKGROUND: The purpose of the study was to evaluate the performance and feasibility of tuberculosis diagnosis by sputum microscopy after bleach sedimentation, compared with by conventional direct smear microscopy, in a setting of high prevalence of HIV. METHODS: In a community-based study in Kenya (a population in which 50% of individuals with tuberculosis are infected with HIV), individuals with suspected pulmonary tuberculosis submitted 3 sputum specimens during 2 consecutive days, which were examined by blind evaluation. Ziehl-Neelsen-stained smears were made of fresh specimens and of specimens that were processed with 3.5% household bleach followed by overnight sedimentation. Two different cutoffs for acid-fast bacilli (AFB) per 100 high-power fields (HPF) were used to define a positive smear: >10 AFB/100 HPF and 1 AFB/100 HPF. Four smear-positive case definitions, based on 1 or 2 positive smears with the 1 AFB or 10 AFB cutoff, were used. RESULTS: Of 1879 specimens from 644 patients, 363 (19.3%) and 460 (24.5%) were positive by bleach sedimentation microscopy, compared with 301 (16.0%) and 374 (19.9%) by direct smear microscopy, with use of the 10 AFB/100 HPF (P < .001) and 1 AFB/100 HPF (P < .001) cutoffs, respectively. Regardless of the case definition used, bleach sedimentation microscopy detected significantly more positive cases than did direct smear microscopy: 26.7% (172 of 644) versus 21.7% (140 of 644), respectively, with the case definition of 1 positive smear and the 1 AFB/100 HPF cutoff (P < .001), and 21.4% (138 of 644) versus 18.6% (120 of 644), respectively, with the case definition of 1 positive smear and the 10 AFB/100 HPF cutoff (P < .001). Inter- and intrareader reproducibility were favorable, with kappa coefficients of 0.83 and 0.91, respectively. Bleach sedimentation was relatively inexpensive and was not time consuming. CONCLUSIONS: Bleach sedimentation microscopy is an effective, simple method to improve the yield of smear microscopy in a setting of high prevalence of HIV. Further evaluation of this method, under operational conditions, is urgently needed to determine its potential as a tool for tuberculosis control.