Prediction of positive culture for Mycobacterium tuberculosis in inpatients

Purpose: Respiratory isolation has been recommended for all patients suspected to have tuberculosis to avoid transmission to other inpatients and health care personnel. The implementation of these guidelines has generated the isolation of patients with as well as without tuberculosis, significantly increasing hospital costs. The objective of this study is to derive a clinical rule to predict the need for respiratory isolation of inpatients suspected to have tuberculosis. Methods: To identify potential predictors of the need for isolation, 56 inpatients sputum positive for tuberculosis were retrospectively compared with 56 control subjects isolated on admission to the hospital based on the clinical suspicion of the disease, but whose sputum tested negative for tuberculosis. Variables analyzed were all available at the time of admission to the hospital and included risk factors for the disease, clinical symptoms, findings on physical examination, and chest roentgenogram results. Results: Multivariate analysis revealed that the presence of risk factors or chronic symptoms were significantly associated with a positive culture for tuberculosis (odds ratio 7.9; 95% confidence interval [CI]: 24.2-4.4), as were a history of a positive PPD (odds ratio 13.2; 95% CI: 40.7-4.4), high temperature (odds ratio 2.8; 95% CI: 8.3-1.0), and upper lobe disease on chest radiograph (odds ratio 14.6; 95% CI: 57.5-3.7). Shortness of breath (odds ratio 0.2; 95% CI: 0.12-0.53 ) and crackles on physical exam (odds ratio 0.29; 95% CI: 0.15-0.57) were negative predictors of tuberculosis. A point scoring system was developed using these variables. The total score for an individual patient ranged from -6 to 21, with higher values associated with a higher risk of tuberculosis. A score of 1 or higher indicated the need for respiratory isolation, accurately predicting a positive culture for tuberculosis (sensitivity 98%; 95% CI: 95-100%, specificity 46%; 95% CI: 33-59%). Conclusions: Among inpatients with suspected pulmonary tuberculosis, a prediction rule based on clinical findings and chest roentgenogram results accurately discriminates patients requiring respiratory isolation. Clinical Implications: The use of this rule may reduce the number of inpatients requiring respiratory isolation without increasing the risk of nosocomial transmission of tuberculosis.