The purpose of this thesis is on the one hand to study stability of equilibria of some epidemic models and secondly to construct an observer to estimate the non-measured states and a key parameter in a within host model. We propose extensions of classical models SIR, SIRS and SIS and we study the global stability of their equilibria. In presence of multiple pathogen strains, we proved that competitive exclusion principle holds: the strain having the largest threshold wins the competition by eliminating the others. It turns out that the winning strain is the one for which the equilibrium gives the minimum of the susceptible host population. This can be interpreted as pessimization principle. By considering the same model with two strains and a frequency-dependent type of the contact law, we prove that dynamics changes and a coexistence equilibrium exists and it is globally asymptotically stable under some conditions. The asymptotic behavior of the two other boundary equilibria is also established. The stability study of equilibrium states is mainly done by construction of Lyapunov functions combined with LaSalle's invariance principle. We consider an age-structured within-host model of the \textit{Plasmodium falciuparum} parasite with a general infection force. We develop a method to estimate the total parasite burden that cannot be measured by the current methods. To this end, we use some tools from control theory, more precisely observers with unknown inputs, to estimate the non measured states from the measured ones (data). From this, we deduce a method to estimate an unknown parameter that represents infection rate of healthy reed blood cells by the parasites.
