The aim of this study was to develop a weighted clinical prediction model for isolating patients with suspected pulmonary tuberculosis (PTB). We conducted a retrospective case control study in which all patients placed on respiratory isolation for suspected PTB from January 1997 to December 2006 were identified. During this 10-year period, 6957 patients were placed in respiratory isolation, 132 of which were diagnosed with PTB. A representative sample was randomly chosen from the group of patients placed in respiratory isolation but with negative Mycobacterium tuberculosis (MTB) cultures (PTB-nonPTB, 1:4 ratio). These 528 patients along with the 132 patients with positive culture for MTB were included in this study. In multivariate analysis, African American race, history of active tuberculosis (TB), history of positive tuberculin skin testing, TB exposure, weight loss, malaise, night sweats, purulent sputum, positive tuberculin skin testing during hospitalization, white blood cell between 4 and 10,000, upper lobe disease on chest x-rays (CXR), and thoracic computed tomography scans were statistically significantly associated with PTB. Age, human immunodeficiency virus-positive status, and white blood cell count greater than 10,000 were negative predictors. A scoring system was developed using these variables. A patient's score of 4 or higher indicated a need for respiratory isolation, predicting a culture positive for MTB (99% sensitivity; 33.7% specificity). If the prediction model was used for determining isolation for TB at the time of admission, 2734 episodes of isolation could have been avoided during the 10-year study period.
From the *Division of Infectious Diseases, Henry Ford Hospital; and †Biostatistics and Research Epidemiology, Henry Ford Health System, Detroit, MI.
Reprints: Javier Aguilar, MD, Division of Infectious Diseases, Henry Ford Hospital, Detroit, MI 48202. E-mail: firstname.lastname@example.org.