An editorial in AJP opens,

In most countries of the world, older adults kill themselves at higher rates than any other age group. Given that the leading edge of the large post-World War II "baby boom" cohort will reach the age of 65 in 2011, demographers predict a rapid rise in the number of seniors taking their own lives in subsequent decades. The need for effective approaches to late-life suicide prevention is pressing.

No. 

First, I'd like to debunk the prevailing belief that almost all suicides are due to a psychiatric illness, a belief he supports using an article he himself wrote that references studies that don't actually show this.

Construction Of Pathology

Simply assuming anyone who commits suicide is therefore "somehow not normal"; or "only someone mentally ill would kill himself" is wrong.  

If a serial killer says, "kill yourself or I kill your kids" and you kill yourself, are you mentally ill?  Was Ajax mentally ill?

Consequently, the fact that you committed suicide is not proof of illness, in the exact same way that death cannot be proof of pancreatic cancer.

Attributing causality to a complex behavior is masturbation with words.  How is killing yourself from MDD different than killing yourself because of terminal pancreatic cancer?  Note that the syntactic construction forces me to say "from MDD" but "because of pancreatic cancer."  But is that a real, existent distinction?

Is it the same biological mechanism?  Different?  Note: "He was depressed, he killed himself, for no reason, his life was great." The presence of pathology is assumed because of the absence of causes; psychiatry abhors a vacuum. 

While it is clear that suicide is a risk in depression, the issue here is whether one can assume depression if they committed suicide?

Three Problems Of "Psychological Autopsy"

Hearsay:

The evidentiary support for the presence of mental illness in those who commits suicide is mostly determined not by a past history of diagnosed psychopathology, but by a post-mortem psychological evaluation ("tell me what he was like?") in which the deceased has drastically biased everyone around him by killing himself.

A psychological evaluation is basically interviewing "informants" (e.g. family), over three hours, asking psychiatric screening questions to determine diagnoses. Think about this, seriously think about this. The guy is dead, and you're asking the family if this guy ever exhibited signs of mental illness.

The closest analogy is doing a post-mortem of a marriage by only asking the ex-wife.   "The evidence strongly suggests 75% of divorced men are manipulative jerks."  Oh.

Validity of Symptoms Descriptions:

Example, from one of the sources of the above article:

Where subjects suffered both physical illness and depressive symptoms before death it was often difficult to judge whether a depressive episode was present. To ensure a consistent approach to diagnosis, we took possible depressive symptoms at face value; thus, if a subject was reported to suffer tiredness, this was included as a depressive symptom regardless as to whether it may have been due to physical illness.

Go figure: 77% of these people "had" a psychiatric diagnosis.

Generalizability:

So it's legitimate to ask, what percent of suicides have ever been diagnosed before they died? What percent have been to a psychiatrist or primary care doc for psychopathology? The answer varies from 30-60%, which is another way of saying 40-70% have never been. A full 75% had never attempted suicide in their lives-- this was their first and last attempt. It's important to keep this in mind because the point of the editorial is to offer the elderly "access through a care manager to algorithm-driven treatment"-- yes, that's what he wrote-- then you're going to help a maximum of 45% (the so far best response rate in psychiatry) of the 50% you actually ever meet.

Part 2 here.