Can an Algorithm Prevent Suicide?

Probably no. All the evidence we have so far is that this kind of risk assessment does not work. It causes harm. It delays access to treatment for people who need it. It causes an intensification of unhelpful interventions for people deemed to be high risk. Sometimes those interventions increase risk. Often people labelled high risk do not go on to die by suicide (whether they have the interventions or not).

Most people who die by suicide will have been labelled as low risk.

> The algorithm is built on an analysis of thousands of previous suicides in the V.A.’s database, dating to 2008.

The article fails to explain what is meant here. Are these "suspected self-inflicted deaths" (a broad definition that will include many deaths) or are these "deaths ruled as suicide by a coroner / medical examiner" (a narrow definition that will exclude some deaths)?

Some links:

https://sites.manchester.ac.uk/ncish/reports/the-assessment-...

https://www.thelancet.com/journals/lanpsy/article/PIIS2215-0...

https://www.nationalelfservice.net/mental-health/suicide/sui...