Can tuberculosis case finding among health-care seeking adults be improved? Observations from Bissau

SETTING: The Bandim Health Project study area in Bissau, Guinea-Bissau. OBJECTIVE: To assess the potential usefulness of predictors (elsewhere applied) and clinical scores (TBscore and TBscore II) based on signs and symptoms typical of tuberculosis (TB) in case finding. DESIGN: Observational prospective cohort study of patients with signs and symptoms suggestive of pulmonary TB (PTB) from 2010 to 2012. RESULTS: We included 1089 PTB suspects with a mean age of 34 years (95%CI 33-35); human immunodeficiency virus (HIV) prevalence was 15.1%. PTB was diagnosed in 107 suspects (76.4% sputum smear-positive, 25.2% HIV-infected). Cough > 2 weeks had the highest diagnostic ability (area under the receiver operating characteristic curve [AUC] 0.66, 95%CI 0.62-0.71), while TBscore < 3 best excluded PTB (negative likelihood ratio [LR-] 0.3) when HIV status was not known. TBscore II >= 3 had the highest diagnostic ability in HIV-infected PTB suspects (AUC 0.62, 95%CI 0.53-0.72), while the absence of self-reported weight loss best excluded PTB (LR - 0.2). Cough > 2 weeks as a trigger for smear microscopy missed 32.1% of smear-positive PTB cases. CONCLUSION: Case finding could be improved by screening symptomatic adults for cough and/or weight loss using TBscore II as the trigger for smear microscopy. To suspect PTB only in patients with cough > 2 weeks (non-HIV-infected) or with current cough, fever, weight loss or night sweats (HIV-infected) was not effective in patients whose HIV status was unknown at first visit.