Safety Risk Formulation — Suicide & Violence (Teaching Tool)

Compared with the average person in the general population

Include clinical/subgroup-comparator factors in the combined estimate (less valid)

How the illustrative combined estimate is calculated

Multiplying relative risks assumes each factor is independent. They are not: depression, hopelessness, insomnia, substance use and prior attempts cluster in the same people, so a naive product badly overstates risk. This tool therefore shows a range rather than a point:

Lower end: the single strongest selected factor alone (as if every other factor were fully redundant).
Diamond (central): factors sorted by strength, each added on the log scale at a geometrically decreasing weight (1, ½, ¼, …), so additional correlated factors add progressively less.
Upper end: the naive product (as if fully independent), shown only to illustrate how much independence assumptions inflate risk. Values are capped at 500×.
Protective factors are discounted the same way and then applied at half weight, reflecting SAFE-T guidance that protective factors do not negate acute risk.
Demographic baseline (sex) is a rate ratio to the overall population and is applied directly.
Only factors with general-population comparators are included unless you tick the box above. Factors without a usable figure (e.g., hopelessness, plan with intent, threats) are listed separately and do not lower the number, even though they may matter most clinically.

This damping scheme is a teaching heuristic, not a validated model. Validated multivariable tools exist for narrow populations: OxMIS for suicide in severe mental illness, OxRec for violent reoffending after prison. They cannot be reproduced from pooled odds ratios.

References for selected factors

Select factors to list their references.