This all depends on how much effort you want to put into this and whether you believe the next person will read it (and whether someone is paying for the documentation time). I hand document.
When documenting reports I use a variety of tools; a flowchart program for the table relations, comments inside the report, a PDF maker or equivalent to increase my publishing flexibility and a word processor.
I find the elements most critical to the process to include pretty much on a per-report basis; a db list, tables, a link diagram to show relations, the selection criteria, explanatory notes in any formulas through comment lines, a naming convention that remains consistent at least throughout the project, and notation on grouping.
I also try to find a way to leave explanatory notes around for any unusual choices that were made. If possible, notes as to what was asked for and what was delivered. Notes on particular permissions that had to be obtained to make a report work. Things like that.
If any db investigation was required to make the reports, those docs become part of the effort.
I don't feel that field lists are necessary, nor do I feel that I need formatting or layout information. These are irrelevant. I'd rather know why a statistical formula derives from a column B instead of column A than to know that it happens to be Magenta. Or even why it happens to be Magenta.
Any of these things being present when I walk into a project is an unexpected bonus, so even if you don't go whole-hog any of these can help the next developer. And cut the costs to everyone involved by not making the organization relearn the lessons that were learned.
And of course it helps to have them in an organized file structure on a network or in a binder. TOC if you have time.
All of this developed around the fact that I have a tendency to forget things myself, so when returning to a report I would become unhappy with having to relearn any piece of it.