No references I'm afraid. AHI is convenient because it's a single number, but just like BMI, it ignores a lot of context. RDI (RERA, respiratory effort related arousal index) might be a slightly better measure, but still imperfect. Intuitively, AHI doesn't make much sense. There are multiple sleep stages, if an AHI of 4 prevents you from entering REM sleep, or if it keeps knocking you out of deep sleep, you're going to have a bad time even though the numbers look good.