Anyone who knows anything about research knows that you have to include *everyone who starts* treatment (not just graduates) in your statistics in order for them to honestly demonstrate the success rate of your treatment.
This is known as "intention to treat" analysis-- and it is a critical element in evaluating health care. For example, let's say I'm trying to sell my "eat no food for 40 days" diet. I can honestly claim that 100% of people who religiously follow this diet will lose weight. Of course, virtually no one will do that-- and this is one reason why intention to treat analysis matters. it measures real-world outcomes, not outcomes of those with superhuman will power.
Or, say, I have a treatment for nail fungus that works great but causes severe diarrhea. Virtually no one will be willing to put up with that side effect so my clinical trials are going to have a high drop out rate: but I could possibly sneak the drug past the FDA if I didn't do intention to treat analysis, which is another reason why it is used.
Also of note: for a study to realistically measure treatment outcomes, not only does it have to be an intention-to-treat analysis, it has to reach a significant portion of those who started the study. In other words, mailing a survey and getting 30% of them back doesn't count. Unless you get 60, 70, 80% or more back, you are probably having a selection effect: in other words, the people who are replying are not a random sample.
Most likely, you are getting a larger proportion replying of those doing well-- for several reasons. One: people not doing well might not be reachable by mail-- homeless, in jail, etc. Two: people not doing well have been taught to blame themselves, so they are embarrassed to admit it. Three: people doing well like to boast about it and will take the time to fill out the survey. Etc.
Unless you have a significant response rate (or you control for lack of response by assuming a large negative bias-- for example, assuming all nonresponders have bad outcomes), you are not going to say anything real about effectiveness.
And, of course, you should also have a control group that was either untreated or treated with a different method in order to show that people didn't just get better with time, which is very, very common with kids.