A machine learning-based framework for Predicting Treatment Failure in tuberculosis: A case study of six countries.

Area of Science:

• Medical research
• Data science in healthcare
• Public health

Background:

• Tuberculosis (TB) is a leading cause of death globally, with treatment failure being a significant contributor.
• The underlying reasons for TB treatment failure remain largely unknown, leading to increasing mortality rates.
• Machine learning and data analytics offer potential for identifying factors associated with disease outcomes.

Purpose of the Study:

• To identify features strongly correlated with tuberculosis treatment failure using feature selection techniques.
• To validate identified features using various classification algorithms.
• To analyze demographic-based feature associations in high-burden countries.

Main Methods:

• Utilized a real-life patient dataset from six high-burden countries: Azerbaijan, Belarus, Georgia, India, Moldova, and Romania.
• Applied feature selection techniques to identify critical attributes linked to treatment failure.
• Employed multiple classification algorithms for feature validation.

Main Results:

• Achieved an average accuracy of 78% on the combined dataset.
• Attained a higher accuracy of 92% specifically for Romania's data.
• Demonstrated that the identified features significantly influence treatment failure outcomes.

Conclusions:

• The study successfully identified key features associated with tuberculosis treatment failure.
• These findings can aid in timely identification of at-risk patients, potentially improving treatment success rates.
• The insights gained can inform public health strategies in high-burden regions to combat TB mortality.