Validity of a decision tree for predicting active pulmonary tuberculosis

Area of Science:

• Infectious Diseases
• Medical Diagnostics
• Public Health

Background:

• Outbreaks of multidrug-resistant tuberculosis (MDR-TB) in healthcare settings raise concerns about transmission.
• Current recommendations for isolating all suspected tuberculosis patients are costly and may not be feasible.

Purpose of the Study:

• To develop and validate a classification tree to identify patients with suspected tuberculosis who are unlikely to require isolation.
• To reduce the number of unnecessary isolation episodes without compromising patient safety.

Main Methods:

• A classification tree was developed using binary recursive partitioning on clinical and radiographic data from 277 isolation episodes.
• Predictor variables included upper zone disease on chest radiograph, fever, weight loss, and CD4 count.
• The tree was validated on a separate cohort of 286 isolation episodes.

Main Results:

• The classification tree achieved 100% sensitivity and negative predictive value in predicting patients unlikely to require isolation.
• In the validation cohort, the tree demonstrated 100% sensitivity and 100% negative predictive value.
• The model had a specificity of 48.1%.

Conclusions:

• The developed classification tree can reliably identify patients with suspected tuberculosis who do not require isolation.
• Implementing this tool could decrease isolation requirements by over 40%, reducing healthcare costs and resource utilization.
• The approach ensures no increase in the risk of cross-infection.