In a prospective, randomized intervention study, 36 children with diabetes mellitus type I were followed, the aim being to study if a family psychosocial intervention at diagnosis could improve glycemic control and minimise hospital admissions. The control group was treated initially in a hospital ward, while the whole family of the children in the study group received therapeutic and social support in an out-hospital training apartment. In the study group only, both parents reported a significant improvement of the family climate. An increased risk for poor glycemic control was recorded in children living in one-parent families and in families where the father had a low level of education. Younger age, a single-parent family and poor glycemic control were factors associated with more days of rehospitalization. The divorce rate in the whole group was at least as high as in the normal population but, surprisingly, maternal dysfunction was associated with lower levels of HbA1c. The conclusion is that the social family background is a most important factor for the glycemic control and need for hospital readmission of the diabetic child.