<span name="style_bold">Aim: </span>To establish an initial approach to individual and group variables that determine adequate interventions of diabetic patients of local health services of the Costa Rican Social Security local. <span name="style_bold">Methods: </span>The study design is non-experimental, cross-sectional and applies multi-level logistic regression. The data were obtained from a sample chosen from the Health Services Purchasing Direction, and includes diabetic patients cared for at local levels from January to December 2004. The information was used at 2 levels of analysis; level 1 (individual) and level 2 (group), and glycosilated hemoglobin was used as a dependent variable. <span name="style_bold">Results: </span>Forty-nine percent were controlled,patients with a median age above 60 years, 66 % of the sample were women, 76,6% with a body mass index (BMI) reflecting overweight and obesity. On the average, the local health services in the study had 7 years since initiating the reform process, and their average score on the “Commitment of Management” for the last 5 years was 87 %, the average population covered was 40 thousand inhabitants, and 22 % of them had high school education. Multilevel logistic regression revealed that as diabetic patients aged the probability of achieving control also increased, women had a lower probability of being under control as compared to men, diabetic patients belonging to the local health services that inititiated their reform earlier, and those belonging to health services achieving higher scores on the “Commitment of Management” had a higher proability of being under control, even though this finding is not statistically significant, given the other model variables. <span name="style_bold">Discussion: </span>Approximately between 6 % and 10 % of the variance in the diabetic patient control, is explained by differences within local health services, after controling other intervening variables.Approximately between 6 % and 10 % of the variance in the diabetic patient control, is explained by differences within local health services, after controling other intervening variables.

diabetes Mellitus; hemoglobin A; glycosylated; logistic models; Costa Rica