Association of education level with diabetes prevalence in Latin American cities and its modification by city social environment

Abstract

Background: Diabetes prevalence continues to increase in urban areas of low-income and middle-income countries (LMIC). Evidence from high-income countries suggests an inverse association between educational attainment and diabetes, but research in LMIC is limited. We investigated educational differences in diabetes prevalence across 232 Latin American (LA) cities, and the extent to which these inequities vary across countries/cities and are modified by city socioeconomic factors.

Methods: Using harmonised health survey and census data for 110 498 city dwellers from eight LA countries, we estimated the association between education and diabetes. We considered effect modification by city Social Environment Index (SEI) as a proxy for city-level development using multilevel models, considering heterogeneity by sex and country.

Results: In women, there was an inverse dose-response relationship between education and diabetes (OR: 0.80 per level increase in education, 95% CI 0.75 to 0.85), consistent across countries and not modified by SEI. In men, Argentina, Brazil, Colombia, Chile and Mexico showed an inverse association (pooled OR: 0.92; 95% CI 0.86 to 0.99). Peru, Panama and El Salvador showed a positive relationship (pooled OR 1.24; 95% CI 1.04 to 1.49). For men, these associations were further modified by city-SEI: in countries with an inverse association, it became stronger as city-SEI increased. In countries where the association was positive, it became weaker as city-SEI increased.

Conclusion: Social inequities in diabetes inequalities increase as cities develop. To achieve non-communicable disease-related sustainable development goals in LMIC, there is an urgent need to develop policies aimed at reducing these educational inequities.

Keywords: diabetes; education; epidemiology; social inequalities; urbanisation.

Figure 1

Age-adjusted percentage of people with diabetes by country and gender according to their individual education level. Age was categorised into three groups: 25–39, 40–64, and 65 and older. These categories were further included in Poisson regression models to estimate diabetes prevalence by education levels.

Figure 2

Predicted probabilities of diabetes based on the multilevel logistic regression models to assess the effect modification of education level by city-Social Environment Index (SEI). Education level: 0=less than primary; 1=primary; 2=secondary; 3=university or more. City-SEI values are based on the standardised variable used in the model so the range of values differs from the on presented in table 1.