Creep-fatigue crack growth models have been proposed employing $(C_t)_{avg}$ as a crack tip parameter characterizing the time-dependent crack growth. The basic assumptions made in these previous models were ideal creep reversal conditions such as no creep reversal and complete creep reversal condition. Due to this assumption, the applicability of the models was limited since they did not consider partial creep reversal condition which is usually observed in many engineering metals at high temperature. In this paper the creep reversal parameter, Temperature;$C_R$, which was defined by Grover, is critically evaluated to quantity the extent of partial creep reversal at the crack tip. This approach does not rely on any simplifying assumptions regarding the extent of the amount of creep reversal during the unloading part of a trapezoidal fatigue cycles. It is shown that the $(C_t)_{avg}$ value calculated for 9Cr steel agrees well with the experimentally measured one. It is argued that the extent of improvement is not significant when the result is compared with that of the conventional model which has an assumption of full creep reversal behavior.