The characteristics of children with diabetes readmitted to Children's Hospital during a 5-year period, 1984 to 1989, were compared with those characteristics of new-onset patients admitted for stabilization and education and to outpatients in the Children's Hospital diabetes program to determine which characteristics were associated with patients who were readmitted. Changes in the frequency of readmissions were examined to determine whether the introduction of a diabetes team and a program that emphasizes the importance of ensuring that patients at risk of readmission consistently received insulin injections resulted in a reduction of readmissions. Readmissions occurred more frequently in patients who were black (71% compared with 38% of new-onset patients and 31% of outpatients) (P less than .001), from one-parent homes (56% compared with 27% of new-onset patients and 24% of outpatients) (P less than .001), and without third-party insurance (45% compared with 18% of new-onset patients and 15% of outpatients) (P less than .001). Readmissions were very common at 14 to 15 years of age (39% of readmissions vs 18% of outpatients) and very uncommon in children younger than age 9 (6% of readmissions vs 27% of outpatients) (P less than .001). Fewer readmissions for ketoacidosis occurred in the summer than in any other season (P less than .05). Readmissions fell by 47% over the 5-year period while new-onset patients increased by 85%. The reduction in frequency of readmissions was due to fewer readmissions for ketoacidosis and fewer readmissions in blacks, in patients from one-parent homes, and in patients without third-party insurance.(ABSTRACT TRUNCATED AT 250 WORDS)