[Surname as a marker of ethnicity. A study from child health services shows that immigrant respective Swedish families seem to be isolated in different ways].

A practical method was tested for identifying ethnicity through family names in the absence of precise demographic information. In a questionnaire study of children covered by the Swedish child health services in some different counties, all potential participants (the children's mothers) were classified as "Swedish" or "non-Swedish" using family names. Inter-rater reliability was assessed by comparing the scores of two independent raters (Cohen's kappa: 0.89). Cases in which the raters did not agree were settled through discussion. The method's validity was roughly measured by comparing assigned classifications against ethnicity information provided by the 63.8 percent of the mothers who returned the questionnaire (1039 of 1628). Assigned and self-reported classifications were the same in 86.5 percent of the cases. Information from child health services about mothers who had agreed to participate in the study revealed health-related differences between those who returned the questionnaires and those who did not. The family name classification method indicated an ethnic selectivity in drop-out rates, which was helpful in interpreting these health differences.