Introduction: Etiological models use identical estimation procedures as most predictive modelling (i.e., regression) to quantify the relative risk related to a selected exposure on an outcome. Though regression is usually used for both purposes, the way within which the model is built will differ thanks to the goals of the model. The goal of a prediction model differs in several important ways. Mathods: Using a cohort design that links baseline risk factors to a validated population-based diabetes registry, a model (Diabetes Population Risk Tool or DPoRT) to predict risk factors for diabetes using commonly-collected national survey data was developed and validated. The event cohort was the National Population Health Survey (NPHS) linked to the validated Diabetes Database, a provincial component of the National Diabetes closed-circuit television (NDSS). Variables were restricted to factors routinely measured within the population. The probability of developing diabetes was modelled using sex-specific survival functions for those __ampersandsigngt; 20 years, without diabetes and not pregnant at baseline (N = 19,000). Results: The age-standardized 5-year incidence rates in the development cohorts were 6.52 % for males and 5.42 % for females. The 3-year age-standardized incidence rates in the development cohort were 3.42 % for males and 2.41% for females. The age-standardized 5-year incidence rates in the development cohorts were 6.42 % for males and 4.20 % for females. The age-standardized 3-year incidence rates for validation cohort was 3.45 % for males and 3.22 % for females. Conclusion: Determinants of weight and weight change are essential when developing strategies to prevent or reduce the future diabetes burden. In monitoring trends over time researchers are often faced with the dilemma of separating trends between individuals and trends within individuals. Multilevel growth models allow us to model both these aspects which strengthen the ability to model trends that vary between and within individuals.