The aim of this population-based, one-to-one matched-pair case-control study was to evaluate the factors concerning the markedly increased risk for dying among Japanese subjects with insulin-dependent diabetes mellitus (IDDM) from a social and behavioural perspective. The study was based on the population-based cohort of IDDM subjects in the Diabetes Epidemiology Research International Mortality Study. We studied 90 cases who died and 90 living control subjects, selected from the rest of the cohort, who were matched for sex, birth year, year of diagnosis and duration of diabetes. Socioeconomic and behavioural status were surveyed through a questionnaire. Conditional logistic regression analyses based on 55 respondent pairs revealed that the better educated patients (year of completing education: odds ratio = 0.66) who kept the same physician (number of times a patient changed physician: odds ratio = 2.77) and who attended a clinic specializing in diabetes (attendance at university hospital clinic: odds ratio = 0.18) injecting insulin several times a day (number of injections, odds ratio = 0.31) and more frequently attending the clinic (> or = 12 times per year, odds ratio = 0.23) were at substantially lower risk of death. The results begin to profile the patients with the highest risk of dying who could be identified earlier and undergo intervention treatment.