My own experience in troubleshooting complex systems has made me very aware of the difference between events that cause faults and those that merely expose preexisting faults. The medical profession, despite constantly acknowledging that they are dealing with extreme complexity, for which they have had years of training, is surprisingly slow to catch on to this basic difference. We probably are not catching these diseases at onset, only at the point of clinical presentation. This makes reasoning about causation very tricky.
One rule every practicing physician learns is that "if you do not cause harm, 85% of the patients you see will get better no matter what you do." This means they are constantly ignoring preclinical conditions. Laboratory tests have generally been developed to support careful clinical diagnoses, and the rule above existed before most modern tests were available. This is a particular problem in dealing with chronic infectious disease. Doctors find it convenient to deal only with the acute phase, where infectious agents are replicating rapidly and numbers increase exponentially. This is also the phase in which antibody response is most prominent.
In chronic infectious diseases either immune tolerance for the pathogen develops or the patient likely develops an autoimmune disease. Antibodies indicating a particular infection are likely to become less prominent than in acute infectious disease by the same infectious agent. This leads to such things as patients with late-stage TB producing negative responses to tests based on tuberculin response. This also happens with neurosyphilis and patients displaying ambiguous response to a Wassermann test. Patients can also display positive responses after treatment, which makes it impossible to determine if the infectious agent has been cleared. The only way to make a definitive determination of infection of the brain is often at autopsy.
Autopsies of patients who die with "mental illness" are rare, unless there are questions of abuse. A high percentage of careful autopsies of patients who die with diagnosed dementia do show signs of chronic infection in the brain. Many neurological diseases with "unknown etiology" have been shown to spread from cell-to-cell, suggesting an undetected infectious agent.
There are statistical correlations between antibodies to toxoplasma gondii in schizophrenics, but prior clinical diagnosis of toxoplasmosis is very rare. The obvious inference, that some schizophrenic patients have a treatable chronic infectious disease meets surprising resistance. This even holds true after a few patients treated with minocycline immediately at onset of symptoms, for unrelated infections, seem to recover completely, or even permanently. (This simply does not happen with typical treatment protocols for schizophrenia, unless you use a definition of "recovery" which allows it to happen repeatedly.) Clindamycin is another antibiotic effective against toxoplasmosis which ought to be effective in some patients. Guess what, schizophrenia is a possible adverse reaction to clindamycin. This ought to tell people about the role of infection and immune response in schizophrenia.
Finally, there are substantial differences in diagnosis in different countries. Patients with ambiguous symptoms are more likely to be diagnosed with depression in Europe, while the same patient could be diagnosed as schizophrenic in the U.S. The "protective effect" could be nothing more than exclusion of one diagnostic category of unknown etiology by another which is more fashionable. In those rare cases where a strong genetic basis can be followed in particular families those members who got the seriously defective genes (DISC-1,2) all developed serious mental disorders. The catch here is that, even though the genes are named for schizophrenia, some family members exhibited unipolar or bipolar depression instead of schizophrenia.
(Before you go trumpeting the news that the genetic cause of schizophrenia has been found you need to also consider those studies which examined hundreds of carefully diagnosed schizophrenics without finding a single example of the defective genes.)
This whole subject is a window on the present primitive state of the art.