Follow 463,244 Mexican schoolchildren for four school years and the steepest climb in overweight wasn't among the heaviest children. It was in the third of pupils on the path just below them

Most national figures on child weight are snapshots: this year's children against last year's different children. Mexico's national registry of weight and height offered a rarer view. Researchers linked encrypted identity codes across four school years, from 2015–16 to 2018–19, and followed 463,244 children in 17,372 schools, each measured once a year for a total of more than 1.85 million measurements. Nearly nine in ten were 6 to 8 years old at the start.

At that start, 18.4% were classed as overweight, 15.7% as having obesity, 62.0% as normal weight and 3.9% as thin. A statistical model then sorted the children by the path of their BMI-for-age scores, and four groups came out: a low group (13.1% of children), a mid-range group (43.7%), an elevated group (32.0%) and a very-high group (11.2%).

The interesting one is the elevated group. Compared with the low group, the odds of overweight or obesity climbed with each school grade most steeply there, with an interaction odds ratio of 1.33 a grade, against 1.22 for the mid-range group. The very-high group, where overweight was already the norm, went slightly the other way (0.89). In practical terms, the children whose status was changing most sat in that large band below the heaviest, a group a one-off snapshot can't pick out.

The paths also split along social lines, in a direction that may surprise anyone who thinks of child obesity as a problem of poverty alone. In schools in Mexico's most marginalised municipalities, 3.1% of children were on the very-high path and 20.7% on the elevated one. In the least marginalised, the figures were 11.7% and 32.7%. Urban schools had 11.7% on the very-high path, rural schools 7.9%. Indigenous and community-run CONAFE schools had 2.7%, private schools 12.2%. Boys were much more likely than girls to be on the very-high path, at 14.0% against 8.4%.

Those gaps need care. Marginalisation and rural setting describe where a school is, not a particular family's income, and the registry held no information on diet, activity, parents' weight or puberty. The study can't say why any child took any path. Schools joined voluntarily, so the authors say the registry isn't representative of Mexican primary schools, and 84.7% of these children attended schools in the least marginalised municipalities, with only 1% in the most marginalised. Only children with all four yearly measurements counted, and the number left out for that reason couldn't be reconstructed. The groups themselves are model outputs, not natural categories, and the data end in 2019.

Even so, the method has something to teach. Prevalence tells a health ministry how many children are overweight this year. Repeat measurement of the same child shows who is moving, and when, which could help school programmes act before a child's path has settled.