Comparing two models with different (naive) baseline

I would like to compare a model with listwise deletion to a model with multiple imputation. However, the model with listwise deletion has a majority class of 70%, while the model with multiple imputation has a majority class of 60%. These classes differ due to the fact that the first model has deleted part of the observations. My accuracy results are 0.75 for the first model and 0.67 for the second model.

Can I conclude that the second model performs better because 0.67 compared to 0.6 is higher than 0.75 compared to 0.7? Or can the models not be compared?

I hope you could help me out!