The long lists of murders by people previously treated for mental illness are scary, but are basically long lists of anecdotes. The attempt to deliver some stat

The long lists of murders by people previously treated for mental illness are scary, but are basically long lists of anecdotes. The attempt to deliver some statistics isn't very helpful: Is there any way to estimate the frequency of these episodes of violence committed by mentally ill person who were not being treated? There was then, and continues to be, no national database that tracks homicides committed by mentally ill persons. However, a small study published in 1988 provided a clue. In Contra Costa County, California, all 71 homicides committed between 1978 and 1980 were examined. Seven of the 71 homicides were found to have been done by individuals with schizophrenia, all of whom had been previously hospitalized at some point before the crime. The 10% rate was also consistent with the findings of another small study in Albany County, New York. Therefore, by the late 1980s, it appeared that violent acts committed by untreated mentally ill persons was one of the consequences of the deinstitutionalization movement, and the problem appeared to be a growing one. This basically gives P(Schizophrenic|CommitsHomicide), but the number of interest is P(CommitsHomicide|Schizophrenic). Bayes' Rule: P(CommitsHomicide|Schizophrenic) = P(Schizophrenic|CommitsHomicide)P(CommitsHomicide)/P(Schizophrenic) P(Schizophrenic|CommitsHomicide) = 7 / 71 (Very rough estimate from a single county 35 years ago) P(CommitsHomicide) = ??? (I'm having trouble estimating this---the easily available figure is murder rate, but that's not the same as probability of being a murderer) But let's fudge this as being the murder rate of 4.7/100,000, which surely overshoots because some murderers kill multiple people [1] P(Schizophrenic) = 15.2/100,000 [2] Then we get P(CommitsHomicide|Schizophrenic) = 0.0986 * 0.000047 / 0.000152 = 0.0305 So very roughly three percent of people with schizophrenia would be expected to commit a murder. This is 649 times the probability of the average person being a murderer, which is quite a substantial difference. However, in my opinion that increased likelihood of violence would not justify involuntary commitment in all but the most extreme cases. Otherwise we would essentially have to imprison tens of thousands of people for crimes that 97% of them are not going to commit. It's like Minority Report's "pre-crime", but a horribly inaccurate version thereof. [1]: https://en.wikipedia.org/wiki/List_of_countries_by_intentional_homicide_rate https://en.wikipedia.org/wiki/List_of_countries_by_intention... [2]: http://epirev.oxfordjournals.org/content/30/1/67.full http://epirev.oxfordjournals.org/content/30/1/67.full