The problem with Control-Shift-A is that most terminal emulation programs do not send a different code for it from the one sent by Control-A, so there would be no way for screen to tell the difference.
You might be able to use C-q if you do stty -ixon to disable “software flow control” on the tty. I usually do this so I can use C-s to search on the command line.
You might consider C-^, since it seems to be usually unbound. It is Control-Shift-6 on US keyboard layouts, but it may be difficult to type on other layouts.
Most people never use the tty control character C-\ to send SIGQUIT. To use it, disable the tty functionality with with stty quit undef and bind it as your prefix key. This is a bit risky though, since if you hit C-\ on a tty where the quit key has not been disabled with stty, you might kill the foreground process instead of activating screen.
I switched to tmux a while ago. Its default is C-b because it was developed under screen and did not interfere with screen's default. I have found that C-b works fairly well for me (I usually use the arrow keys for single character movement on the shell command line—though I use M-b and M-f to move by words).
Incidentally, the author of tmux uses C-a in his primary sessions and always types C-a twice (whether in a tmux session or not). If you are willing to adopt this habit, it seems like a decent compromise since the second C-a should be a no-op in most Emacs-like editing contexts. It would not work so will if you often use programs/contexts outside of screen where the second C-a would not be a no-op.
By default, C-a C-a is bound to the other command (switch to most recently active screen) in screen. To have C-a C-a send an actual C-a and have C-a o switch windows, put the following in your .screenrc:
escape ^A^A
bind o other
