Survey Evidence and Uncertainty in Global Adult Insufficient Physical Activity Estimates, 2000-2022

Objective. To characterize how the density and recency of country-specific survey evidence map onto the uncertainty of national insufficient-physical-activity estimates within the current global estimation system, and to assess whether these patterns persists after adjustment for statistical performance. Methodology. A secondary meta-epidemiological analysis linked 507 surveys from 163 countries to annual estimates and 95% uncertainty intervals for 197 countries/territories, 2000–2022. The outcome was uncertainty-interval width. Evidence available was characterized by survey count, distance to the nearest survey and surveys within five years. Cross-sectional and country-year fixed-effects models were used, with external validation against the World Bank Statistical Performance Index. Because the survey evidence is an input to the Bayesian model that generates the uncertainty intervals, the analyses were interpreted as a structural audit of evidence–uncertainty patterns rather than as independent estimates of information value or the causal effect of adding surveys. Results. Thirty-four countries had no eligible survey. The median uncertainty interval width in 2022 was 36.03 percentage points with no survey and 13.32 with four or more. In the 2000–2022 panel, each additional year from the nearest survey was associated with 3.1% wider uncertainty and each nearby survey with 5.3% narrower uncertainty. After adjustment for statistical performance, income and region, countries without an eligible survey had 79.4% wider uncertainty (95% CI 53.8% to 109.4%). Conclusions. Apparently comparable national estimates are generated from evidence bases that differ markedly in density and recency, and those differences are reflected in the uncertainty produced by the current estimation system. These associations are partly structural because the surveys are model inputs; they should not be interpreted as the expected effect of commissioning additional surveys. Reporting survey provenance alongside prevalence and uncertainty would improve transparency.