The CHF in PWR rod bundles is usually predicted by the local flow correlation approach based on subchannel analysis while difficulty exists due to the existence of spacer grids especially with mixing vanes. In order to evaluate the effect of spacer grids on CHF, the experimental rod bundle data with various types of spacer grids were analyzed using the subchannel code, COBRA-IV-i. For the Plain grid data, a CHF correlation was described as a function of local flow conditions and heated length, and then the residuals of the CHF in mixing vaned grids predicted by the correlation were examined in various kinds of grids. In order to compensate for the residual, three parameters, distances between grids and from the last grids to the CHF site, and equivalent hydraulic diameter were introduced into a grid parameter function representing the remaining effect of spacer grids predicted most of the CHF data points in plaing grids within $pm$20 percent error band. Good agreement with the CHF data was also shown when the grid parameter function for mixing vaned grids of a specific design was used to compensate for the residuals of the CHF data predicted by the correlation.