Feature importance of socio-economic parameters in Tuberculosis modeling

TL;DR

This study models tuberculosis and HIV co-infection dynamics across Russian regions using a modified SIR model, incorporating socio-economic factors and Shapley values to identify key parameters influencing disease spread.

Contribution

Introduces a unified modeling approach that integrates socio-economic parameters with disease dynamics, utilizing Shapley values for parameter importance estimation.

Findings

Identifies key socio-economic factors affecting TB-HIV co-infection rates.

Demonstrates the effectiveness of the modified SIR model in regional epidemic modeling.

Shows the utility of Shapley values in reducing model complexity.

Abstract

This paper considers the problem of modeling epidemic outbreaks in different regions with a common model, that uses additional information about these regions to adjust its parameters and relieve us of mundanity of data collecting, and inverse problem solving for each region separately. To that end, we study tuberculosis and HIV dynamics in regions of Russian Federation from 2009 to 2023 in connection with number of socio-economic parameters. SIR-like model was taken and modified as a dynamic model for tuberculosis-HIV co-infection and inverse problem of transfer rates between compartments was solved, based on statistical data of diseases incidence. To shorten the list of socio-economic parameters we make use of Shapley vector that allows us to estimate importance of these parameters in reconstruction of differential model parameters using regression algorithms.