Here is one that may poossibly help

http://www.stat.ucdavis.edu/~wang/paper/EOB3.pdf

I'm not sure why you need to use non-parametric data, unless you have a single ordinal item for quality of life. If you have multiple items that go into a quality of life scale, then you should look at a thread I started that provides some basic resources on scale construction.

https://www.researchgate.net/post/How_can_one_help_people_create_simple_scales_from_Likert-scored_items

It seems that you should be able to use an appropriate parametric longitudinal model for your analysis. A general estimating equation (GEE) model will allow you to choose the appropriate family and link, given your data type and distribution.