Each group has is its own compositing pipeline so you can easily create sub-compositions if you assign groups to slices. In the main composition, each group ignores the layers below and becomes the input for whatever is the layer above.
In your case, if you have three different contents to send to three different screens, you could make each group a sub-composition and use slice transform to do the screen assignments. That gives you more creative options than using one layer per display.
I think this short video will give you a good idea of what I'm talking about: