Because diseases are abstractions based on cause (or "etiology" in medical language), a diagnosis serves as an explanation of the symptoms involved. This might sound rather sophisticated and/or complicated, but it's really not. If you go to the doctor's office and complain that your stomach hurts, "indigestion" is an explanation because it refers to a causal process (that is, why your stomach hurts). Were the doctor to use an abstraction based on symptomology (such as "stomachache"), it would not.
Of course, the doctor's explanation for the symptoms can be wrong. Throughout most of medicine, this is referred to as "misdiagnosis". There are also syndromes and the like which we don't know the causes of and times when the doctor can't figure out what's going on. While we know some things about these syndromes and cases (e.g. epilepsy tends to be chronic problem), these "diagnoses" aren't explanations of the symptoms -- they're descriptions of them.
Once you get it, this is really pretty simple. An answer to the question of why your symptoms exist (in more technical language, an "etiological construct") can explain them; a description of the symptoms themselves (in more technical language, a "symptomolgoical construct") can't. Despite this, however, people often make this mistake in a wide variety of ways. There's even a formal name for doing so: "nominal fallacy".
Put yet another way, you cannot say that your stomach hurts because you have a stomachache. "Stomachache" is a symptomological construct -- a label for the stomach pain. You cannot say that you are having difficulty sleeping because you have insomnia. The statement that you "have insomnia" is simply another way of saying that you have trouble sleeping. Neither serves as an explanation. This isn't to say that terms and concepts like "stomachache" or "insomnia" can't be useful, but they can't answer most questions of "why"... because they have nothing whatsoever to do with cause.
I'm making this as clear as possible because there is one field of medicine where the definition of the term "diagnosis" I provided does not apply. That field is psychiatry.
"Mental disorders", as used in psychiatry, are not etiological constructs. They are symptomological constructs. To use my earlier analogy, they are not akin to "indigestion" and are more akin to "stomachache". When a psychiatrist "diagnoses" a mental disorder, they are emphatically not saying anything about the cause of the symptoms you present them with -- they are simply deciding how to describe those symptoms in the standardized and highly formalized language of psychiatry.
The psychiatric "diagnosis" of "major depressive disorder" is simply another way of saying that someone is depressed... only it's far more precise (among other things, it distinguishes "major depression" from less severe or more transient types of depression). The psychiatric "diagnosis" of "bipolar disorder" basically means that someone goes through 'episodes' during which his mood is different from normal (in a clinically significant way). Similarly, the psychiatric "diagnosis" of "autism" basically means that someone isn't following the developmental psychologists' often-bigoted (and why I call it that is a whole 'nother blog post) One True Developmental Path for human beings.
When looked at this way, the way that people tend to accumulate multiple psychiatric diagnoses is easily understood -- for many of the same reasons that I don't think people would be surprised to learn that people with stomachaches also have fevers much more frequently than people who do not. This is simply because fevers and stomachaches can be caused by many of the same things.
The easiest of these to resolve are simply matters of degree -- for instance, what constitutes "markedly diminished interest or pleasure in... activities"? Where do you draw the line between what's "markedly" diminished and what's just diminished?
For the most part, these represent a sort of diagnostic "fuzziness" which is... resolvable, albeit not necessarily easily. Statistical methods are pretty good at dealing with this sort of issue in a research setting, although the problem remains. It remains an obstacle, but hardly an intractable one. If this problem is not understood, however, it can create a very wide variety of misconceptions.
Other problems, however, are more noteworthy -- and fundamental. For one thing, the defining feature of a "symptom" in medicine is that it's viewed as an indicator of an underlying pathology. Stomachache is a symptom of indigestion because it provides evidence in support of the idea that you are having trouble digesting food. It provides this evidence because problems with digestion tend to cause stomachaches. "Stomachache", in general, is viewed as a symptom of disease because a stomachache is a pretty clear indicator that something is going wrong in the body (even if you don't know what, and even if the problem is fairly minor).
In other words, a "mental disorder" is a disorder because it is viewed as a sign that there is something wrong with the person who exhibits it. Our judgments of what constitutes something being "wrong" with someone, however, are notoriously problematic.
We human beings have a tendency to judge other people based on our expectations and our often-prejudiced personal (and/or cultural) views on what people should be. When people fail to live up to these, we tend to conclude that there's something wrong with them, rather than concluding that the problem was with our views and expectations.
For instance, homosexuality used to be a DSM mental disorder (and even though most sources will state that it was removed in 1973, this is not entirely accurate). Moreover, its official status as such has a long history of being used to justify the torture (via abusive "treatments") both of homosexuals and people judged as being "at risk for" homosexuality.
Then there's the rather infamous (and atrocious) example of the countless ways in which psychiatry and psychiatric diagnoses have been used as a tool of institutionalized racism and of racial oppression. We can even look at the ways in which attitudes about race have affected diagnostic patterns.
Then there's the issue of so-called "diagnostic redefinition", something which is rather hard to understand for people who don't understand that psychiatric disorders are symptomological constructs.
Diagnostic redefinition is relatively easy to understand if you look at approximate analogues involving symptomological constructs in the world of general medical practice. In this case, I'm going to use the construct of obesity for the purpose of explanation.
At present, obesity is most commonly defined in terms of something called "body mass index" (BMI) -- a calculated value based on height and weight. Neither BMI nor obesity, however, are etiological constructs -- they're descriptive constructs. In the case of obesity, it's a symptomological construct, presently defined by a BMI of thirty or higher (in most countries, anyway).
If, however, medical researchers were to find that a different cutoff point -- say twenty-five (which, incidentally, is the cutoff point in Japan) or thirty-five -- was more meaningful, the cutoff point would change to reflect this. If the cutoff point was lowered, a number of people would suddenly find themselves "obese" when they weren't before -- something which is called "broadening criteria" for obesity. If the cutoff point was raised, a number of people would find themselves no longer considered "obese", due to something called "narrowing criteria".
Note that nothing would really have changed with these people themselves. Only the terms used to describe them -- the label they receive, in other words -- would have changed. This is the essence of diagnostic redefinition in psychiatry.
To continue the analogy, if we were to find that some other measure of obeisity (e.g. total body weight, percentage body fat) was more meaningful than BMI, our definition of obeisity would shift to accomodate this. Obeisity would be redefined in terms of this new metric, and a number of people would suddenly "gain" or "lose" a "diagnosis" of obeisity without changing one whit themselves. The newly "diagnosed" or "undiagnosed" wouldn't have changed -- the language used to describe them would have.
This is precisely what happens every time a new edition of the DSM comes out. Sometimes it happens more often.
Note that none of this means that the "diagnosis" of "obesity" isn't useful or meaningful. None of it means that obesity isn't real (although if one is feeling particularly philosophical, one can point out that it's only a label or descriptor, and as such the phrase "for a certain value of 'real'" applies -- it's only "real" in the sense that "redness" is; similarly, "autism" is only real in the senses that "intelligence" is).
I just hope that this helps people understand certain matters and helps clear up some of the assorted confusion regarding the topic. Countless authors -- in academia, in the blogosphere, in the print media -- clearly don't understand a lot of what I try to explain above.
Hopefully, I did not just "try".