The participants characteristics are presented in Table 1. Their mean age was 50 years (from 26 to 68 years), and there were 186 women and 69 men. There were differences in education level: the nursing staff within the municipalities mainly had upper secondary school training, though few had university education, while a greater proportion of staff in the county council had a university education. A greater proportion of staff working in the municipalities reported having a neighbour, or having once had a neighbour, with mental illness. The multiple regression analysis with the dependent variable staff’s attitudes (CAMI-S) and the hypothetical independent variables staff members’ age, gender, education level, personal contact (currently or ever; lived with, worked with, neighbour with, friend with someone with mental health problems), stigma-related knowledge (MAKS) as well as employer revealed a significant model (Table 2). The model explained 16% of the variance, but only stigma-related knowledge (P < 0.001) and employer (P = 0.013) had significant beta coefficients. The GEE model with CAMI-S as the dependent variable revealed a significant effect for stigma-related knowledge (P < 0.001) and employer (P = 0.043), however in this model personal contact (currently having or having had a friend with mental health problems) also showed a significant effect (P = 0.024). This finding indicates correlations at the unit level in the county council and in the municipalities. Altogether, based on these analyses, the findings indicate that mental health nursing staff have more positive attitudes towards persons with mental illness if their knowledge about mental illness is less stigmatized, if their work place is in the county council and if they currently have, or once have had, a friend with mental illness. This finding deserved further attention, and we proceeded by analyzing and presenting differences between staff employed by county council and the municipalities, respectively, in the four CAMI attitudes factors (Table 3). Significant differences between the groups were found for all factors except for the factor ‘Fear and avoidance’.