Epidemiologcially, similar symptoms that progress to different outcomes is one of the key tools in understanding the underlying mechanisms of a disease, so to say "you can't reliably classify based on outcomes" is incorrect. Sometimes you can, sometimes you can't, and it is always worth asking the question "how come?" when you get very different outcomes from the same disease-causing agent.
In the general case there are four possibilities:
1) the different outcomes arise because of a difference in the individuals involved (this is useful for exploring treatment options: is there any way to make people with bad outcomes more closely resemble people with good outcomes?)
2) the different outcomes arise because of a difference in the disease-causing agent (this is useful when allocating treatment resources and preparing patients if the outcome is likely to be bad.)
3) some interaction between the first two factors
4) some other external factor is at play (you're a smoker, another patient is not.)
As I mentioned in another comment, I've done work on cancer genetics where the goal was to find out why some patients with what appeared to be the same disease "simply fell apart" (in the words of one of the physicians) while others lived relatively long and healthy lives). There certainly was a belief that we could classify the disease based on the outcome, and that belief motivated some fairly difficult research, but no one called it unreasonable because we had similar examples in the literature (Kahn's work on small round blue cell carcinoma being a famous case in point: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1867426/)
So: when you have populations with different outcomes, it is both reasonable and routine to ask if you have different conditions, despite superficial similarities in clinical presentation.
In the general case there are four possibilities:
1) the different outcomes arise because of a difference in the individuals involved (this is useful for exploring treatment options: is there any way to make people with bad outcomes more closely resemble people with good outcomes?)
2) the different outcomes arise because of a difference in the disease-causing agent (this is useful when allocating treatment resources and preparing patients if the outcome is likely to be bad.)
3) some interaction between the first two factors
4) some other external factor is at play (you're a smoker, another patient is not.)
As I mentioned in another comment, I've done work on cancer genetics where the goal was to find out why some patients with what appeared to be the same disease "simply fell apart" (in the words of one of the physicians) while others lived relatively long and healthy lives). There certainly was a belief that we could classify the disease based on the outcome, and that belief motivated some fairly difficult research, but no one called it unreasonable because we had similar examples in the literature (Kahn's work on small round blue cell carcinoma being a famous case in point: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1867426/)
So: when you have populations with different outcomes, it is both reasonable and routine to ask if you have different conditions, despite superficial similarities in clinical presentation.