Decision Instrument for the Isolation of Pneumonia Patients With Suspected Pulmonary Tuberculosis Admitted Through US Emergency Departments

Abstract

Many patients with pneumonia are admitted to respiratory isolation for possible tuberculosis (TB), but most do not have active TB. We created a decision instrument to predict which pneumonia patients do not need admission to a TB isolation bed. The design was a prospective case series conducted in 11 university-affiliated, urban, US emergency departments (EDs) (EMERGEncy ID NET). Participants were patients admitted to the hospital through the ED with a diagnosis of pneumonia or suspected TB. The main outcome measure was derivation and validation of a sensitive decision instrument to identify patients not having TB (and not requiring isolation) according to clinical data and chest radiographs. Of 5,079 pneumonia patients, 224 (4.4%) had pulmonary TB according to sputum cultures or tissue staining. The instrument derived to predict which patients did not have pulmonary TB included no TB history or previous positive tuberculin skin test result, nonimmigrant, not homeless, not recently incarcerated, no recent weight loss, and no apical infiltrate or cavitary lesion on plain chest radiograph. When tested on the validation subgroup, the decision instrument exhibited a negative predictive value of 99.7% (95% confidence interval [CI] 99.1% to 99.9%), and a sensitivity of 96.4% (95% CI 91.1% to 99.0%). A decision instrument can accurately predict which patients with pneumonia do not require admission to TB isolation rooms.