Can we usefully stratify patients according to suicide risk?