The validity of medical history, symptoms and chest X-ray in predicting pulmonary tuberculosis. Derivation of a predictive model.

Introduction. - As numerous nosocomial outbreaks of pulmonary tuberculosis have been reported during the last two decades, prompt identification and e ffective isolation of contagious patients should be made a priority in tube rculosis control policies. There is a need to develop a predictive model wh ich would allow prompt recognition and isolation of smear-positive patients . Current knowledge and key points. - Various authors have attempted to impro ve the respiratory isolation policies for patients suspected of having pulm onary tuberculosis. A French multicenter prospective study of 211 patients suspected of having pulmonary tuberculosis established that: I) the current respiratory isolation policy of suspected pulmonary tuberculosis needs imp rovement (sensitivity = 71.4%; i.e., 28.6% of smear-positive patients are a dmitted without isolation) and 2) better interpretation of clinical and rad iological data available on patient admission could improve the adequacy of respiratory isolation. Univariate analysis showed that predictive factors of pulmonary tuberculosis were chest X-rays (P < 0.00001), symptoms (P = 0. 0004), age (mean: 40.8 years for TB vs. 47.5 for non-TB, P = 0.04), HIV inf ection (10.6% vs. 28.7%, P = 0.01), immigrant (72% vs. 55%, P = 0.03) and B CG status (P = 0.025), while multivariate analysis demonstrated that chest X-ray pattern (P < 0.00001), HIV infection (P = 0.002) and symptoms (P = 0. 009) were independent predictive factors. Future prospects and projects. - From these data, a model was proposed and evaluated in the derivation cohort using the receiver operating characteris tics (ROC) curve. We retrospectively studied the predictive model in two po pulations different from the one from which if was derived. The model would have improved sensitivity of the respiratory isolation policy from 71.4% ( current respiratory isolation policy) to 82.4% and 91.1%, respectively. Pro spective, multicenter studies are requested to establish the value of such a predictive model in improving the respiratory isolation policy for patien ts suspected of having pulmonary tuberculosis. (C) 2000 Editions scientifiq ues et medicales Elsevier SAS.