This is a common claim which has been repeated enough times that one can find many sources claiming it to be true. However, this doesn't seem to be corroborated by accounts of the time or serious biographies. Indeed, Helen M. Walker, in her 1934 biography of De Moivre in Scripta Mathematica Volume II, Number 4, August 1934, (reproduced here freely on Google Books as an afterward) explicitly states the following:
It is often related that De Moivre, always interested in number series, had foretold that each day he should need 15 minutes more sleep than on the preceding day, and that his death would occur when the total reached 24 hours. The story, passed on as an interesting fiction by several writers, has some basis in fact, but the arithmetic progression and De Moivre's prophetic statement are purely apocryphal, and are not to be found in any of the accounts written by men living at the time of his death.
Therein, she cites a number of biographies of De Moivre, including by Fouchy (original available through Gallica) and by Maty (available translated through Arxiv). These are the only two major biographies to have appeared soon after his death. They're quite similar (see the annotations to Maty's biography for more details). Neither has any mention of this story. I have not checked every other biography cited therein, only those two linked above, but Walker extensively documents her sources and it can be reasonably safely assumed that none of them have anything related to this story or she would have noted it.
The story isn't completely false. It is true that De Moivre was, towards the end of his life, in an increasingly frail state. He required in excess of 20 hours per day of sleep, and his condition only worsened. However, the story about predicting his own death has no reliable source, and can reasonably be assumed to be at least partly apocryphal.
I have not traced this back to its origin. Wikipedia cites Cajori's 1893 text A History of Mathematics. This contains a slightly different version of the story without citation. It should be noted that this was intended as a popular presentation of an abbreviated history of mathematics, not an extensive scholarly work, so the lack of references is not surprising. There may be earlier sources for this claim (in fact, I suspect this is likely, as Cajori probably didn't just make it up), but none by any of De Moivre's biographers who knew him personally.