Background: The burden of childhood tuberculosis (TB) remains significant especially in areas of high HIV prevalence. Clinical diagnosis predominates, despite advances in molecular and microbiological diagnostics. The aim of this study is to identify clinical features associated with culture-confirmed pulmonary TB (PTB) in children. Methods: Children admitted to hospital were enrolled in a study of novel diagnostics for PTB in South Africa. Standardized clinical, radiological and microbiological data were collected. Definite TB was defined by culture of Mycobacterium tuberculosis from a respiratory specimen. Adjusted odds ratios for definite TB were calculated using a multivariate logistic regression model. Results: Adjusted odds ratio (AOR) for definite TB increased with a history of fever for more than 1 week [AOR: 8.54, 95% confidence interval (CI): 2.37-30.74], with a chest radiograph (CXR) suggestive of PTB (AOR: 10.0, 95% CI: 3.22-31.2) and with a positive tuberculin skin test (TST; AOR: 64.4, 95% CI: 14.3-290.5). The likelihood ratio of having definite TB if 2 of these factors (CXR and TST) were present compared with having none of them was 17.7. Cough, household contact with TB, HIV status and wheezing were not significantly associated with definite TB. Conclusions: Prolonged fever, CXR suggestive of TB or a positive TST were predictive of definite TB and should be considered in composite scoring systems for TB diagnosis in high HIV prevalence settings. Other commonly associated symptoms were not associated with definite TB.