Who is running the asylum? That is the question raised by the case of Darren Harkin, the schizophrenic who absconded from a psychiatric hospital to rape a schoolgirl at knifepoint, after staff allowed him to build up a library of violent and pornographic films.

The answer given by almost every mental health expert that I have talked to is the same. The Harkin case is extremely unusual. Most mentally ill people pose a much greater risk to themselves than to others. Hysteria about this story will only increase public suspicion of psychiatric patients, who are mostly harmless and badly need to reintegrate into the community.

There is only one problem with this perfectly reasonable response. While Harkin himself is certainly unusual - at the age of 12 he stabbed his baby brother to death and cut off his arm, and he suffers from autism as well as schizophrenia, hence his ending up in the care of the National Autistic Society - the mistakes made in his case are not, sadly, as rare as we might like.

Yesterday the BBC revealed that 116 people have absconded last year from medium and low-security psychiatric hospitals or while under the escort of hospital staff. Sir David Ramsbotham, the former Chief Inspector of Prisons, branded this figure horrifying and called for more mental health facilities to be built inside prisons to keep the public safe.

I would not go that far. First, the BBC has provided no evidence that any of the 116 escapees has hurt anyone. Secondly, prisons already house far too many mentally ill petty criminals who, according to the Prison Reform Trust, are simply made more ill by jail. But I suggest that someone takes a good hard look at the management of institutions from which patients are absconding, and at why some people under supervision outside prison seem much more dangerous than those inside it.

Louis Appleby, the national director for mental health, explained yesterday that these “medium and low secure units” aim to rehabilitate people with low-grade, long-term mental illness, not to protect the public from killers. In other words, prisons are not hospitals and hospitals are not prisons. That is fair enough - although one could quibble with the lavish use of the word “secure”. But why then, was a proven killer, who had demonstrated aggressive behaviour and absconded three times from his medium secure unit, put into a “low secure” one?

Harkin was a restricted patient. The National Autistic Society says that tThe decision to transfer him to a low-security unit was, made by a multi-disciplinary team that included at least once person from the Ministry of Justice's mental health unit.

I am not party to the details of the decision to move Harkin and pander to his interest in pornography. But it seems to defy common sense, just as two years earlier, decisions by Springfield Hospital led to a schizophrenic patient, John Barrett, walking out, buying a kitchen knife and killing Denis Finnegan, a stranger cycling in Richmond Park.

Yesterday I read the latest of the inquiries - published in June - into homicides committed by those who have been previously in contact with mental health services. Sean Perry, who also killed a stranger, had also been at Springfield Hospital. He was “known to be capable of great violence”. But his psychiatrist was said to be “unduly focused on the wishes and desires of SP and his family rather than the risks he posed when unwell”.

Two years after the Barrett inquiry, the same hospital trust is said to have a missing person policy that “is not explicit enough about the need to ensure the return of patients who pose a risk if left at large”. That is dangerous complacency. Perry, Barrett and Harkin were violent men whose actions were predictable and avoidable. That does not mean that professionals are wrong to treat such people and to hope for improvement. But they are clearly failing accurately to assess the risks they pose.

Schizophrenics tend to be hospitalised rather than jailed, because they are deemed treatable. Most schizophrenics are not violent. But the Confidential Inquiry into Suicide and Homicide shows that schizophrenia is found in 1 per cent of the population and in 5 per cent of perpetrators of homicide. In the case of Perry, Barrett and Harkin, it looks as if the professionals' hope of medicating them back to normality overcame their duty to bear in mind the risks they posed.

Professor Anthony Maden, lecturer at Imperial College and forensic psychiatrist at Broadmoor, says that many general psychiatrists in Britain have an “attitude problem” which stops them assessing risks properly. Their fear of stigmatising patients, he says, makes them reluctant to talk about risk at all. In a culture that prizes “engaging” with patients to the point of indulging their passion for porn, risks are sometimes overlooked. Pandering to those who are not that ill in this way also leads to neglect of the neediest patients, Maden says, because it will not admit the seriousness of their problems.

The desire of the mental health lobby to accentuate the positive is understandable. They want to help people to gain acceptance in a society that is generally hostile. But this optimism has become a problem when the determination to overcome stigma leads people to rewrite reality.

Since I discovered that 50-60 people a year are killed by people recently in contact with the mental health services, I have repeatedly been told that this is less than a tenth of all murders (true), a stable number that is not increasing (true) and something that it would be “irresponsible to broadcast”.

Really? Disingenuousness will not allay public fears. Perhaps the psychiatrists should take a leaf out of the airlines' book. When there is a very small risk of something very nasty happening, blustering that “it's very rare” is far less convincing than saying that “even one is too many” - and to mean it.

Times Recommends

• Police beg us for food. What hope is there?
• Moral darkness that engulfed Baby P
• More to Advent than cheap chocolate