OBJECTIVE: To assess the diagnostic value of provider-initiated symptom screening for tuberculosis (TB) and how HIV status affects it. METHODS: We performed a secondary analysis of randomly selected participants in a community-based TB-HIV prevalence survey in Harare, Zimbabwe. All completed a five-symptom questionnaire and underwent sputum TB culture and HIV testing. We calculated the sensitivity, specificity, and positive and negative predictive values of various symptoms and used regression analysis to investigate the relationship between symptoms and TB disease. FINDINGS: We found one or more symptoms of TB in 21.2% of 1858 HIV-positive (HIV+) and 9.9% of 7121 HIV-negative (HIV-) participants (P < 0.001). TB was subsequently diagnosed in 48 HIV+ and 31 HIV- participants. TB was asymptomatic in 18 culture-positive individuals, 8 of whom (4 in each HIV status group) had positive sputum smears. Cough of any duration, weight loss and, for HIV+ participants only, drenching night sweats were independent predictors of TB. In HIV+ participants, cough of > or = 2 weeks' duration, any symptom and a positive sputum culture had sensitivities of 48%, 81% and 65%, respectively; in HIV- participants, the sensitivities were 45%, 71% and 74%, respectively. Symptoms had a similar sensitivity and specificity in HIV+ and HIV- participants, but in HIV+ participants they had a higher positive and a lower negative predictive value. CONCLUSION: Even smear-positive TB may be missed by provider-initiated symptom screening, especially in HIV+ individuals. Symptom screening is useful for ruling out TB, but better TB diagnostics are urgently needed for resource-poor settings.