Nutritional predictors of chronic disease in a Central Australian Aboriginal cohort: A multi-mixture modelling analysis

• Multi-mixture modelling is a novel way of modelling the latent nutritional profile of subgroups of the population.
• For a Central Australian Aboriginal cohort, mixture- modelling revealed four subpopulation types based on nutrition profiles.
• Multi-mixture modelling appears useful in identifying population subgroups at risk of chronic conditions.

Background and aimsChronic diseases (including diabetes, cardiovascular disease, hypertension and chronic kidney disease) are major contributors to the total burden of disease for Aboriginal people. Here we used novel epidemiological modelling to investigate nutritional profiles at a population level associated with chronic disease.Methods and resultsMulti-mixture modelling, was used to group members of a Central Australian Aboriginal population (n = 444) based on their nutritional profile from a risk factor prevalence survey in 1995. Multi-mixture modelling assigned % membership to four classes; Class 1 (young, low adiposity and lipids, low dietary antioxidants; n = 171.7); Class 2 (older, greater adiposity and lipids; n = 22.6); Class 3 (predominantly female, greater adiposity and antioxidants, low smoking; n = 134.3) and Class 4 (predominantly male, greater lipids and adiposity, low antioxidants, high smoking prevalence; n = 115.4). For persons free of chronic disease (n = 285), incident chronic disease for classes 1, 3 and 4 was determined using follow up hospital, primary health care and death records collected in 2004/05. Fifty-four percent of Class 4 had incident chronic disease, an excess of 3355 events per 100,000 person years relative to Class 1. Incident CVD, hypertension, or CKD was highest for Class 4 and incident diabetes highest for Class 3.ConclusionMulti-mixture modelling appears useful in identifying population subgroups of an Aboriginal population at risk of chronic conditions.