From what I've read, it tends to be higher than at average weights, but doesn't approach the mortality rate among obese individuals.
Here's a little bit on the subject; they conclude more or less that slim individuals are at an increased risk of early death despite dietary/other life conditions suggesting otherwise, but they can't explain it.
I do think, however, that there's a difference between individuals with a low BMI because of habitual dieting, and those who are at their optimal weight. My BMI is at 16.8 (which is pretty far underweight), but for my frame it's a healthy build. Moreover, I essentially
cannot gain or lose weight at any significant level. When I'm in a period where I'm eating heavily without much exercise I don't gain weight, and when I'm eating one meal a day and hardly noticing I don't lose weight. The closest I can get is a few pounds of muscle gained or lost when I start or stop working. This is all purely anecdotal, of course, but my father is in a similar situation and is nearly 60 without any serious health concerns (despite having been a heavy smoker most of his life, to boot), while my paternal grandfather is the same, save for being close to 90 with no issues other than arthritis.
Basically, I think that the data may be skewed by people who are dieting excessively and forcing themselves below their healthy weights, as that pretty clearly
would lead to health complications. It's common sense that forcing your body to gain or lose large amounts of weight is going to put stress on it.