Sleep apnea is a dynamic process over time with a lot of variability, capturing these dynamics and charactering the nature of variability are important for providing an objective characterization of the underlying physiological phenomenology of respiratory events. AHI is the useful measurement that clinically used nowadays, which looks at the apnea in an average aspect. Rather than looking at AHI as an average rate, we want to explore how these individual events evolve over time.

We apply a general point process framework to respiratory events, so that we are able to compute an “Instantaneous AHI”, which measures the moment-by-moment events rate as a function of clinical observations as well as event history.

It turns out that history dependence plays the most important role in predicting the event timing, while people with similar AHI may have very different history dependence structures. Moreover, the way in which previous events influence subsequent events can be used as a means of phenotyping, paving the way towards improved metrics for developing personalized treatment.