---

Title:

Projections of NCD mortality and risk factor data

Definition:

Projections of NCD mortality and risk factor indicators with time-series data for estimating future country attainment of targets.

Estimation method:

Projections were generated based on the methods used for the projections of the WHO Global Programme of Work (GPW13) indicators. For glucose, mortality, alcohol and salt projections, eight models were considered: random walk with trend (RW2), autoregressive (AR1), exponential smoothing, Holt’s linear trend, Holt’s linear trend (damped), flat extrapolation, linear extrapolation and annual average rate change extrapolation. As described in further detail in the GPW13 methodology document, models were tested using existing data, with part of these serving as test data. To determine which model performed best, the following statistical metrics were used: RMSE (root mean squared error), MAE (mean absolute error), MdAE (median absolute error), MASE (median absolute scaled error) and CBA (confidence bound accuracy: percentage of test points that lie within the predicted confidence intervals). Using all of these metrics, the best model was selected independently for each indicator and then used to project data to 2025 or 2030. For tobacco use, a Bayesian negative binomial meta-regression was used to fit trendlines and project to 2025. Projections are predominantly based on data from before the COVID-19 pandemic and are not adjusted for any possible impact on trends the pandemic may have had. Adult prevalence of obesity, prevalence of insufficient activity were projected assuming trends since 2010 continue. To do so, a regression was fit for each Bayesian model iteration, country and sex, with the probit of age-standardized prevalence as the dependent variable, and year as the independent variable. The regression was used to predict age-standardized prevalence for each model iteration, country, sex and projection year. The projected values shown are the mean of all Bayesian iterations for the country, sex and projection year. For obesity, on-track status was assessed based on the proportion of iterations for which the regression coefficient was less than or equal to zero, which is the posterior probability that the trend in age-standardized prevalence was flat or decreasing. Each country-sex was considered on track if the posterior probability of a flat or decreasing trend was greater than 0.5. For insufficient physical activity, each sex and country were considered on-track if the projected value for 2030 was less than 85% of the estimated value for 2010. For both obesity and insufficient physical activity, countries were further categorized by the likelihood that past trends were sufficient to meet the target. Raised blood pressure was projecting using a similar method, with the exception that regressions were fit for each country and sex.