TBscore II: Refining and validating a simple clinical score for treatment monitoring of patients with pulmonary tuberculosis

Abstract

Background: The TBscore, based on simple signs and symptoms, was introduced to predict unsuccessful outcome in tuberculosis patients on treatment. A recent inter-observer variation study showed profound variation in some variables. Further, some variables depend on a physician assessing them, making the score less applicable. The aim of the present study was to simplify the TBscore. Methods: Inter-observer variation assessment and exploratory factor analysis were combined to develop a simplified score, the TBscore II. To validate TBscore II we assessed the association between start score and failure (i.e. death or treatment failure), responsiveness using Cohen's effect size, and the relationship between severity class at treatment start and a decrease < 25% in score from the start until the end of the second treatment month and subsequent mortality. Results: We analyzed data from 1070 Guinean (2003–2012) and 432 Ethiopian (2007–2012) pulmonary tuberculosis patients. For the refined score, items with less than substantial agreement (κ ≤ 0.6) and/or not associated with the underlying constructs were excluded. Items kept were: cough, dyspnea, chest pain, anemia, body mass index (BMI) < 18 kg/m², BMI < 16 kg/m², mid upper arm circumference (MUAC) < 220 mm, and MUAC < 200 mm. The effect sizes for the change between the start of treatment and the 2-month follow-up were 0.51 in Guinea-Bissau and 0.68 in Ethiopia, and for the change between the start of treatment and the end of treatment were 0.68 in Guinea-Bissau and 0.74 in Ethiopia. Severity class placement at treatment start predicted failure (p < 0.001 Guinea-Bissau, p = 0.208 Ethiopia). Inability to decrease at least 25% in score was associated with a higher failure rate during the remaining 4 months of treatment (p = 0.063 Guinea-Bissau, p = 0.008 Ethiopia). Conclusion: The TBscore II could be a useful monitoring tool, aiding triage at the beginning of treatment and during treatment.

Keywords::

• Triage
• low income setting
• mortality prediction

Acknowledgements

We thank the PTB patients in Gondar and Bissau for participating in this study, and the project assistants, nurses, and physicians in Gondar and Bissau for their good and consistent work.

Declaration of interest: The authors declare that they have no competing interests. The study was funded by grants from EU/EDCTP, the A.P. Møller, Else and Mogens Wedell-Wedellborgs, Frimodt- Heineke, Aase and Ejnar Danielsens, Jacob and Olga Madsens, Julie von Müllens and the Beckett Foundation. FR received a travel grant and project support grant from the Clinical Institute of Aarhus University and the Aarhus University Research Foundation.