Can you pick a cohort who circumstantially have to socially interact outside the home? Then you can monitor interaction effect neutral to their tendency to self isolate.
Can you pick a cohort with a trusted buddy and double test their state?
Could you measure this indirectly by some other thing, like rate of use of masking scents or cleaning product?
This seems like a long shot suggestion but could you train a service dogs to detect the changes in the odor and to signal to the person for recording? Since the odor is very detectable even by humans then the dogs should have zero problems detecting it far sooner.