Design: cross-sectional accelerometer study with weighted percentile estimation
Population: 3,570 children aged 3 and 4 years in childcare and education centres across 44 countries
Primary outcome: daily step count and cadence-based percentile reference values
Effect: weighted median 18,542 steps/day (95% CI 18,318-18,812); 17,502 at age 3 and 18,928 at age 4

SUNRISE fitted accelerometers to 3,570 typically developing children aged 3 and 4 attending childcare and education centres across 44 countries, spanning six regions and every World Bank income band, between 2018 and 2025.

The weighted median was 18,542 steps a day (95% CI 18,318-18,812): 17,502 at age 3 and 18,928 at age 4. Boys exceeded girls at most percentiles, by roughly 1,600 steps at age 3 and 1,150 at age 4. The gradient that will surprise people is the socioeconomic one, and it runs the other way from the adult literature: children in low- and middle-income countries and in rural settings accumulated more steps than those in high-income countries and urban areas. Cadence-based measures followed the same pattern with smaller gaps.

These are reference values, not targets. A percentile chart says where a child sits among peers, and nothing about what is healthy - no step threshold in this age group has been linked to any outcome. What the distribution does provide is a comparator that did not exist, and one that includes the settings where most of the world's preschoolers live. For an Indian clinic the relevant reading is that urbanisation, not poverty, is what tracks with fewer steps.

Use these as reference percentiles, not as a target - no step count is linked to an outcome at this age.
Median is about 18,500 steps a day at ages 3 to 4, higher at 4 than at 3.
Boys walked more than girls across most percentiles; account for sex when comparing.
Rural and lower-income settings recorded more steps, the reverse of the adult pattern.
Ask about active travel and outdoor play time, which is what the urban gap most likely reflects.

Why it matters

Step counts are the one activity measure a cheap wearable can produce anywhere, and until now there was nothing to compare a preschooler's number against.

The statistics, in plain English

Reference values describe a distribution; they do not define health. The confidence interval around the overall median is narrow because the sample is large, but that precision says nothing about whether a child at the 20th percentile is at any risk. This is also cross-sectional and drawn from children attending childcare centres, so it is not a random sample of all preschoolers in any of the 44 countries.