Using ARIA to control screen readers

Hi
 
I want to provide additional contextual information about the HTML element
currently being voiced by a screen reader if the user presses a key as it
reads through the content on a page. For example, the page might contain a
list of used cars for sale and as this list is being read by a screen
reader, I want the user to be able to press a key, say 'd', and have the
screen reader read a description of that particular car before continuing
reading through the list. 
 
I know I could simply include the description in the visible content and the
screen reader would read this out, but there may be many cars in the list
and do not want the user to have to keep skipping the descriptions of cars
they may not be interested in.
 
I do not know of any way of finding out which element the screen reader is
currently reading at the time the user hits a key. I would be very
interested to hear if anyone thinks this may be possible and how to achieve
it.
 
I have been looking at ARIA and thought it may be possible to loop through
the elements on the page and update the content of an aria-live region as it
progresses. But this wouldn't wait until the screen reader had finished
reading the content of the live region before updating it with the new
content and you would probably only hear the first and last elements read
aloud.
 
I believe that some screen readers may focus the element being spoken in
certain modes which could then be used to determine the element being spoken
but suspect this will not work for all screen readers and may require the
user to switch to a particular reading mode that moves focus with speech.
 
It would be straight forward to simply require the user to press a key to
move to the next element and update a live region at the same time and leave
the control to the user, but this would require the user to manually press a
key to move through the list rather than simply sit back, listen, and only
interract when they want to now more. It's a minor inconvenience and suspect
wouldn't be a big issue for most screen reader users but thought I'd ask
anyway. To be able to do something based on what is being read, as it is
being read, would seem like a useful thing to do as well more generally.
Maybe this is something more for AT though.
 
I'm currently using NVDA to test this concept if it makes any difference.
 
Thanks in advance.
Cheers
Ian