Background: This study was carried out in Guinea-Bissau's capital Bissau among inpatients and outpatients attending for tuberculosis (TB) treatment within the study area of the Bandim Health Project, a Health and Demographic Surveillance Site. Our aim was to assess the variability between 2 physicians in performing the Bandim tuberculosis score (TBscore), a clinical severity score for pulmonary TB (PTB), and to compare it to the Karnofsky performance score (KPS).
Method: From December 2008 to July 2009 we assessed the TBscore and the KPS of 100 PTB patients at inclusion in the TB cohort and/or at 1 or more follow-up visits; 61 baseline and 130 follow-up double assessments were obtained.
Results: The inter-observer variability of the TBscore (5 symptoms and 6 clinical findings) varied from slight to almost perfect agreement. For the TBscore, all 3 severity classes (SC I-III) were observed, while the KPS only yielded 2 of its 3 possible classes. The grading of PTB patients into severity classes showed moderate agreement for both the TBscore (κ(w) = 0.52, 95% confidence interval 0.46-0.56) and the KPS (κ(w) = 0.49, 95% confidence interval 0.33-0.65). The intra-class correlation coefficient (ICC) was larger for the TBscore than for the KPS (0.822 vs 0.632).
Conclusions: The Bandim TBscore had an acceptable inter-observer variability, seemed to be more disease-related, and performed better than the KPS.