Objectives: To develop a diagnostic predictive model for the identification of patients with presumptive pulmonary tuberculosis (PTB) at high risk for active disease and those requiring nucleic acid amplification (NAAT) testing and/or preventive respiratory isolation in low-incidence, high-income countries.
Design: A 1:1 case-control study was conducted in consecutive immunocompetent patients with presumed PTB hospitalised between 2009 and 2012 in Paris, France. Cases were defined as individuals with culture-confirmed PTB, regardless of smear result. Those with presumed PTB and three smear- and culture-negative samples were selected as controls. A score was derived using conditional logistic regression. Internal validity of the score was assessed using the bootstrap method.
Results: A total of 354 patients were included in the analysis (177 cases, 177 controls). Among the 177 cases, 74 (42%) were smear-negative but culture-positive. Factors independently associated with PTB were age <50 years (adjusted OR [aOR] 4.7, 95%CI 1.8-12), diabetes (aOR 3.2, 95%CI 1.1-9.8), absence of cough with or without sputum (aOR 3.7, 95%CI 1.7-8.3), fever >15 days (aOR 3.5, 95%CI 1.3-9.5), apical infiltration without cavity (aOR 3.4, 95%CI 1.4-8.5) and cavitation or miliary pattern (aOR 19.7, 95%CI 7.6-51.1). Score C-index was 0.84 (95%CI 0.79-0.88). Calibration for the overall population (P = 0.770) and in smear-negative patients (P = 0.980) was appropriate. A score of 3.3 had 90% sensitivity, 50% specificity and 79% (IQR 28-95) median probability of PTB.
Conclusions: This score could be used to build an algorithm to determine the need for respiratory isolation and/or NAAT use in PTB disease.