A Multi‐Strain Tuberculosis Transmission Model With Treatment Stratification and Neural Networks

Abstract Tuberculosis (TB) remains a major global health threat, made worse by the emergence of multidrug‐resistant TB (MDR‐TB). This study presents a multi‐strain transmission model for TB that incorporates drug‐susceptible (DS‐TB) and multidrug‐resistant (MDR‐TB) strains. The model introduces key innovations: waning vaccine immunity, separate treatment pathways for latent and active TB, and endogenous development of drug resistance. The system's dynamics are analyzed and control strategies are identified. The methodology combined a deterministic compartmental model with an artificial neural network (ANN) trained via the Levenberg‐Marquardt algorithm (LMB) to enhance computational efficiency. The basic reproduction numbers are calculated as for DS‐TB and for MDR‐TB, confirming the endemic potential of both strains. A global sensitivity analysis using Latin Hypercube Sampling and Partial Rank Correlation Coefficients (LHS‐PRCC) quantified key drivers: a increase in the DS‐TB transmission rate raised by approximately . In contrast, a increase in the treatment initiation rate for active DS‐TB reduced it by , with similar effects for MDR‐TB. These results provide a quantitative tool for policymakers, demonstrating that prioritizing rapid diagnosis and treatment of active cases offers the most effective strategy for immediate TB reduction.