Pediatric Suicides in America

CDC WONDER, 2018–2025, NCHS 2023 urbanisation classification by county of residence. Deaths are as published.

These views restrict the data; they do not standardise it. The gradient is partly a fact about who lives where, so each of these holds race, ethnicity and sex still and lets only urbanisation vary — no weighting, no reference population, no model. Each one is not the rate at which children die in rural America: it sets aside every child outside its own group, which is the point of it and also the danger of it. Every chart shares one scale, and rate ratios take large central metro as the comparator. The dashed outline behind each bar is all children in that category over the same years, so switching view shows the restriction and not a different window. Boys are the group total minus girls, which is exact rather than modelled.

The denominator stops before the deaths do. CDC publishes deaths for White non-Hispanic, Black non-Hispanic and Hispanic children by urbanisation through 2024 and population only through 2021. For 2022–2024 the last published denominator is carried forward rather than projected. Measured against each group's own trend that moves the affected years by 2.6% at worst and under 1% pooled, and it moves them downward for White non-Hispanic children — the highest-rate group — so it narrows the gaps between groups slightly rather than widening them. Counts CDC suppresses for holding fewer than ten deaths are recovered by differencing two exports of the same query, one with suicide included and one without: both sides run to the hundreds, so neither is suppressed. Those recovered annual counts are never shown. Only the pooled figures are, and the smallest of those is eleven deaths. The mechanism split is where the gradient lives. Firearm suicide runs 3.2 times higher in the emptiest counties than in the biggest cities; every other method together runs 1.3 times higher. That ratio of gradients is 2.49 (2.22 to 2.79), and it is the same in boys (2.45) and in girls (2.45). What differs between the sexes is not how steeply either component climbs — boys against girls gives 1.03 for firearm and 1.03 for everything else, both intervals sitting on 1 — but how much of each sex’s deaths sit in the steep component: firearms are 47 to 69 per cent of boys’ suicides across the gradient and 18 to 36 per cent of girls’. Give rural children the big-city firearm rate and change nothing else and 87 per cent of the boys’ gap closes, and 64 per cent of the girls’. This is an ecological comparison of rates, not a claim about any individual death; means restriction is where that inference is best supported and it is still an inference. One firearm cell — small metro, girls, 2024 — is suppressed by CDC and taken here as 9, which is the largest it can be. It is one year of a seven-year pool and moves that pooled figure by under two per cent.

The denominators are modelled, and the modelling is the part to check. CDC publishes population for these categories only through 2021, four years against eight years of deaths. Rather than projecting each category's population forward — which would have carried the 2020/21 change of population vintage forward as though it were demography — each category's share of the classified population is projected and applied to the national person-time used elsewhere on this page. A share is invariant to a change of vintage, because numerator and denominator rescale together. The shares move about a tenth of a percentage point a year, in one direction, and a straight line through the four published years reproduces them with R² between 0.83 and 0.96 and reconstructs every published category population to within a third of one per cent. Figures from 2022 onward rest on that projection and are the least certain on this page.

The six categories account for 98.99% of the national population, stable to 0.016 of a percentage point across the published years; the remainder is children whose county CDC cannot classify. They are excluded here along with their 146 deaths, so the counts in this section sum to 22,206 rather than the national total.

Grouped to match the Canadian page so the two can be read against each other. 4 cells were suppressed by CDC for holding fewer than 10 deaths; they fall in categories totalling well under one per cent and are counted in neither the named groups nor the total, so the shares here are of 22,328 deaths with a published mechanism. No method is described beyond these categories, and none is ranked by lethality.

Rates are crude, not age-standardised, and are computed from counts and population like everything else here. Groups are pooled across the whole period rather than shown by year: Native Hawaiian and Other Pacific Islander children account for 62 deaths across eight years, which cannot be split by year without falling below CDC's threshold of 10. Wide intervals on the smaller groups are the honest consequence of small numbers and should be read as such.