Seems really interesting e.g. PEM is highlighted as a feature of ME/CFS. Currently I assume "PEM" is based on response to questionnaires yet approaching it like this (actimetry - potentially even mobile phone data @Adrian ) could prove an objective basis to assess PEM. Potentially identifying cohorts for particular studies.
I'm wondering if data on sleep patterns could be collected and would it provide insight into sleep disruption?
Sure I'm failing to understand the potential use of this type of data - but it's interesting.
Jonathan:
"I am interested in documenting the changed pattern of activity and its relation to prior activity.
I am interesting in finding out if there is something unexpected about the pattern and relation.
At root I am interested in documenting the pattern of pathology. That is what got me to an answer in RA. I appreciate that there will be all sorts of confounding factors but underneath that there ought to be some sort of regularity to the phenomenon, whatever the cause."
I suspect that phone-data level is the only effective way (goodness knows how you encapsulate all the elements as cogntive/looking at things etc) because I know from having to spend so many years working whilst really ill that it 'collects' e.g. the classic weekend in bed radically resting (I guess at there is some word for it now) even tho you've done that every night you got home from work. And then still having to use annual leave to do further radical rest over a week or more in order to just about get back to the point where you can keep putting one foot in front of the other with the evenings and weekend rest.
Knowing enough maths, I know that we'd need someone who is capable of modelling these different sequences across the data - because the world gives us no leeway for the pretend 'stay on top of your PEM', but I'm not even sure that is how it operates with some of us (difficult to separate having had to do that for so many years from whether that influenced how the body works or whether that was how the PEM worked).
It will need some sort of 'look for patterns/sequences' mode for each individual set of data - because I suspect there is something underlying each on the surface-level obvious stuff and then post-clustering there will probably be explanations hidden in whatever might connect/underly these different groupings. ie those doing the keep going whilst under-resting then 'out' pattern vs 'nearer to pacing' shows the internal data of how that 'collects up' and the 'healing/resting process'. As someone who is 'knocked out' by my body during that phase there could be meat on bones from one of those head sensor things during one of said weeks of recovery vs just weekend vs almost conking out still dressed in work gear when you got through the door of an evening.