A theoretical model of secular body mass index dynamics among Russian adults under changing socio-economic conditions

Abstract Background Body mass index (BMI) is a key indicator of population health and often shifts alongside socio-economic change. Few studies have tracked these dynamics over long periods in transitional economies. Russia’s late-twentieth-century transformations offer a rare opportunity to examine such links. This study develops and validates a time-series model of BMI in 19 years old, relating changes in socio-economic and demographic indicators to BMI trends and producing scenario-based forecasts. Materials and methods We analyzed national time-series data published by the NCD Risk Factor Collaboration for 1975–2016 (males and females, age 19), along with indicators of urbanization, fertility, infant and all-cause mortality, life expectancy, and nutritional proxies (protein supply, animal-source calories, meat). Predictors were standardized. Per sex, we estimated the following: (i) first-difference OLS, (ii) dynamic regressions with a lagged BMI term (ARDL(1,0)), and (iii) smooth-trend models with a natural cubic spline in year. Diagnostics included augmented Dickey–Fuller and Durbin–Watson tests. Model selection triangulated elastic net, partial least squares, and stepwise regression. Rolling-origin one-step-ahead forecasts used only information available at time t; bootstrap resampling assessed sign stability. Results Across specifications, the urbanization share was the most robust correlate of BMI. For males, higher urbanization was consistently and inversely associated with BMI; for females, the association was small and model sensitive (frequently negative but not uniformly significant). Effects of life expectancy and mortality attenuated and often lost significance once smooth time structure was included, indicating shared long-run movement rather than distinct short-run covariation; infant mortality added little independent signal. Nutrition proxies contributed limited, non-robust information. ARDL(1,0) one-step-ahead forecasts outperformed random-walk and trend-only baselines. Under a baseline scenario (continuation of recent socio-economic patterns), projected BMI in 2050 is approximately 26.6 kg/m2 (males) and 26.7 kg/m2 (females). Forecasts use only information available at time t (lagged predictors/nowcasts) and are conditional on assumed exogenous trajectories; longer-horizon projections are scenario based rather than unconditional. Conclusions After explicit treatment of nonstationarity, macrodemographic structure, especially urbanization, shows the most consistent links to BMI at age 19, whereas national-scale nutrition proxies are weak at this grain. Findings are descriptive, not causal; forecasts should be interpreted with caution. Incorporating finer-grained behavioral, dietary, and environmental data will help clarify mechanisms and improve long-term forecasting.