Star formation is an inefficient process and in general only a small fraction of the gas in a giant molecular cloud (GMC) is turned into stars. This is partly due to the negative effect of stellar feedback from young massive star clusters. Recently, we introduced a novel 1D numerical treatment of the effects of stellar feedback from young massive clusters on their natal clouds, which we named WARPFIELD. With this model we can show that the minimum star formation efficiency, i.e. the star formation efficiency above which the cloud is destroyed by feedback, is mainly set by the average cloud surface density and that a star formation efficiency of 1-6 per cent is generally sufficient to destroy a GMC. These results imply that feedback alone is sufficient to explain the low observed star formation efficiencies of GMCs. We have now coupled our feedback model to the plasma code Cloudy and the radiative transfer code Polaris, which enables us to predict more than 100 emission lines from young star-forming regions and to compare these synthetic observations to observational surveys. In this talk I will address how individual objects (out of our database of thousands of simulated star-forming regions) evolve in various diagnostic plots (e.g. standard nebular emission line diagrams, so-called BPT-diagrams). Furthermore, I will present the evolution of whole galaxies as seen in synthetic observations.