It seems to me that there are three options:
Focus on the same spot in the list, but on the new item that has now "slid" into the spot after the deletion. Presumably the list "collapses" and all of the items after the deleted one get moved up a spot closer to the top of the list. If you leave the focus location the same, the new item that occupies that position in the list would be in focus
Move the focus to the item before the deleted item. Instead of leaving the focus on the position in the list, move the focus back to the item BEFORE the deleted item.
Lose focus on the list altogether and declare focus on some higher-level page element.
I believe that if you go by a more standard database model or list-and-pointer model, if you have a pointer pointing at a position on a stack, and you delete an item from the stack, the pointer remains pointing at the same position, which will now be populated by a different item (option one above). I don't know if that is the best model to use for a visual list, but it does present a precedent that might map to some users' understanding.
I'm not terribly familiar with screen readers (shame on me), but where does the page place focus when the user first lands on the page? Might it make sense to go with the third option above and put the user back outside the context of the list? Or does that imply a lot more work for them to get back into the list?