This paper is devoted to analyze the sensitivity of input parameters (basic reproduction number, cure rate, hospitalization rate) on the evolution of the SIHR epidemiological model. Lagrange polynomials affords a convenient way for computing Sobol indices. Those indices furnish important information on the relative weight of each uncertain input parameters. We use the COVID-19 disease as a numerical application to illustrate the study for two different basic reproduction numbers R0.  Furthermore, we consider a sinusoidal and logistic infection rate and we give the Sobol indices when the entry parameters are assumed to follow the uniform law.