Developing a model for type 2 diabetes mellitus medication use prevalence at local level

Background: Socio-economic profile, demographic composition and access to care are important predictors of local health and health care use. How such predictors are associated with medication use at local level is unknown, but knowledge is essential to improve efficiency of preventive programs in risk areas. Objectives: We explored spatial patterns of type 2 diabetes mellitus (T2DM) medication use at local level and determined its association with local demographic, socio-economic and access to care variables. Methods: Using the University of Groningen pharmacy database IADB.nl, we estimated the five-year prevalence of T2DM medication use (2005-2009) in persons aged 45+ at four-digit postal code level for two Dutch areas, totaling 141 postal code areas. Statistics Netherlands supplied data on potential predictor variables. We used empirical Bayes smoothing to deal with the small sample problem. We composed variables for socio-economic profile, ageing and access to care and estimated first an overall multiple linear regression model followed by two models for each separate geographical area. Results: Socioeconomic profile, ageing and access to care were all significant predictors for T2DM medication use in the model incorporating both areas. In our first area (Groningen-Veendam) the smoothed prevalence of T2DM medication use ranged from 5.0% to 24.3%. Access to care and socio-economic profile were statistically significant predictors in this area. In the second area (Zwolle-Kampen-Noordoostpolder) smoothed prevalence ranged from 6.2% to 20.2%. Here, socioeconomic profile and ageing were significant predictors. The discriminative value of the model was higher in our first area (adjusted R-squared 0.41 compared to 0.25 in the second area). The unstandardized coefficients pertaining to the three composite predictors were similar in all three models, indicating that the relations with T2DM are roughly stable across space. Conclusions: Our results demonstrate the importance of socio-economic profile, ageing and access to care variables for explaining and projecting local health care use for type 2 diabetes mellitus