Modelo predictivo clínico-radiológico para diagnosticar tuberculosis pulmonar activa

Abstract

Introduction: The objective of this study was to develop a predictive model on the presence of active pulmonary tuberculosis using clinical-epidemiological data and findings of chest radiography and thoracic computed tomography (CT). Material and methods: An observational, retrospective, descriptive and analytical study was conducted, which collected 22 clinical and epidemiological variables, 11 radiological findings on chest x-ray and 23 on CT that were performed in patients with clinical suspicion of pulmonary tuberculosis during a period of 10 years. A multivariate logistic regression model was applied to the potential predictors of positive culture, obtaining a predictive model. Results: We collected 1,540 patients with clinical suspicion of tuberculosis who underwent radiography and thoracic CT. The culture was positive in 101 cases. A backward elimination process was used to obtain the best set of predictive variables. We obtained 24 variables that were significant (6 clinical, 5 of chest plain films and 13 of CT) and were assigned a score. The sum of these scores was subtracted from the age in years and multiplied by 0.03. The model suggests the diagnosis of active pulmonary tuberculosis in patients with a score higher than 1.845. The model obtained a sensitivity of 85.1%, specificity of 83.6%, positive predictive value of 26.6%, and negative predictive value of 98.7%. The area under the ROC curve was 0.9163. Conclusion: This scoring system based on clinical-epidemiological criteria and radiological findings can help rule out active pulmonary tuberculosis in cases of diagnostic suspicion.