Cordeiro and Ferrari (1991) proposed a way to improve the chisquared approximation for test statistics which converge in distribution to a x2 distribution by multiplying the original statistic by a suitable polynomial in the statistic itself with coefficients that depend on certain moments of log likelihood derivatives. In the spirit of their work, we derive Bartlett type corrections to improve two Wald test statistics for testing the null hypothesis H:θ=θ(0) where θ is the scalar parameter of one parameter exponential family models. Our results are general enough to cover many important and commonly used distributions. We also obtain Bartlett type corrections for natural exponential families with some types of variance functions. Some simulation results illustrate the superiority of our corrected Wald statistics relative to the classical Wald statistics.