Analysis of inhomogeneous Markov reward models