There are some long term diseases where the severity of your symptoms tend towards a 'normal'. So imagine plotting out the severity of the symptoms say, every day or every week, then drawing a line of best fit through them.
So over the long term, severity may change slowly. However in the short term, statistically speaking, it'll tend to be that if symptoms have been super bad for a short while, it'll tend to get better. And if symptoms have been very good for a short while, it'll tend to get worse.
There's a name for this trend. I'm not sure if it's actually a cognitive bias, it could be more of a statistical fallacy type thing. I can't remember the name... and I've searched for it in every way I can think of in google but no luck. I originally read about it in the context of pseudoscience treatments. So someone has a long term condition, they have a very 'bad year', they go see say... a homeopath, and then they get a bit better and they ascribe their improvement to the homeopathy where as in fact, statistically speaking they were likely to have gotten a bit better on their own.