Scenario analysis and optimization of tuberculosis management based on a multi-compartment model

Abstract

Background. Management of the multidrug-resistant tuberculosis epidemic requires resource optimization under conditions of high dimensionality and model nonlinearity. The purpose of the study is to develop a methodology for analyzing optimal control scenarios for a multi-compartmental tuberculosis epidemiology model to minimize the overall disease burden. Materials and methods. The TB-SIR/Q± model accounts for the cocirculation of drug-sensitive and drug-resistant strains, stratification by bacterial excretion status, and patient isolation. Validation confirmed high accuracy (R2 = 0.94, MAPE < 4.2 %), supported by comparison with alternative models (ΔAIC > 10) and residual analysis. Parameter identification was performed using the Sobol method, and the optimal control problem was solved using sequential linearization and quadratic programming methods. Results. The analysis revealed the dominant influence of drug-sensitive strain parameters: transmission rate βS (34%, 95% CI: 0.31–0.37) and isolation intensity qS (18%, 95% CI: 0.16–0.20). The resource-optimal scenario demonstrates a 42.5% reduction in disease burden compared to the baseline forecast with a 15% cost increase, achieving 74% efficiency of intensive intervention. Conclusions. The comprehensive methodology combining verified multi-compartmental modeling, global sensitivity analysis, and constraint optimization proved effective for managing epidemiological systems under limited healthcare resources.