In this paper the viscoelastic responses of four experimental steel-concrete composite beams subjected to highly variable environmental conditions are investigated by means of a finite element (FE) model. Concrete specimens submitted to stepped stress changes are also evaluated to validate the current formulations. Here, two well-known approaches commonly used to solve the viscoelastic constitutive relationship for concrete are employed. The first approach directly solves the integral-type form of the constitutive equation at the macroscopic level, in which aging is included by updating material properties. The second approach is postulated from a rate-type law based on an age-independent Generalized Kelvin rheological model together with Solidification Theory, using a micromechanical based approach. Thus, conceptually both approaches include concrete hardening in two different manners. The aim of this work is to compare and analyze the numerical prediction in terms of long-term deflections of the studied specimens according to both approaches. To accomplish this goal, the performance of several well-known model codes for concrete creep and shrinkage such as ACI 209, CEB-MC90, CEB-MC99, B3, GL 2000 and FIB-2010 are evaluated by means of statistical bias indicators. It is shown that both approaches with minor differences acceptably match the long-term experimental deflection and are able to capture complex oscillatory responses due to variable temperature and relative humidity. Nevertheless, the use of an age-independent scheme as proposed by Solidification Theory may be computationally more advantageous.