Predictors of Initial Smear-Negative Active Pulmonary Tuberculosis with Acute Early Stage Lung Injury by High-Resolution Computed Tomography and Clinical Manifestations: An Auxiliary Model in Critical Patients.

This study evaluated the diagnostic use of high-resolution computed tomography (HRCT), chest X-ray (CXR), and clinical manifestations (CM) to identify initial smear-negative (iSN) active pulmonary tuberculosis (aPTB) [iSN-aPTB] in patients with iSN-pulmonary diseases (PD) and acute lung injury (ALI). In the derivation cohort, the [iSN-PD] with ALI patients were divided into the [iSN-aPTB] (G1, n = 26) and [non-aPTB-PD] (G2, n = 233) groups. Lung morphology, number, and lobar (segmental) distribution were evaluated using CXR and HRCT. A multivariate analysis was performed to identify independent variables associated with G1, which were used to generate predictive score models for G1. The predictive model was validated in a separate population of patients (n = 372) with [iSN-PD] and (ALI). The validated model for [HRCT (CXR + Hypoalbuminemia)] had 93.5% (25.8%) sensitivity, 99.5% (89.4%) specificity, and a negative predictive value of 99.5% (93.0%). For [iSN-aPTB], the post-test probability in the derivation cohort (prevalence = 10%), validation cohort (prevalence = 8.3%), and the given prevalence (prevalence = 1%) was 88.7%, 94.4%, and 41.5%, respectively. The HRCT model effectively identified the [iSN-aPTB] subjects among the [iSN-PD] with ALI, regardless of CM. The [non-aPTB-PD] were also correctly classified by the HRCT and [CXR + Hypoalbuminemia] models.