From national mortality registers, not estimates: 4,509,797 suicide deaths and 2,480,158 undetermined-intent deaths across 139 countries.

The measure is undetermined ÷ (suicide + undetermined) — the share of deaths that could have been called suicide but were not. It needs no population denominator, so it is immune to every assumption in the population estimates.

	value
countries compared	87
median share, women	0.2542
median share, men	0.2097
countries where women's share is higher	64 of 87
Wilcoxon signed-rank, paired within country	p = 0.000013

and what it fails to explain

Recounting every undetermined death as a suicide shifts the ratio in most countries, but shrinks the variance between countries by only about 6%. real and systematic, but does NOT explain the cross-country spread