The survey questionnaire was mailed to 10000 persons aged 15-80 who were randomly selected from the Finnish Population Register and resided in four hospital catchment areas in western Finland. The overall response rate was 51.6% without any incentives or reminders. Overall, the response rate among females was 60% and among males 43%, with the highest response rate in the 50-70 age group. The average age of the respondents was 50.6 (SD 17.3) years. Overall, 16.5% of the respondents were Swedish-speakers. The lowest response rate was among Finnish-speaking men (42.1%) and the highest among Swedish-speaking women (68.8%). Population means and percentages were weighted according to age, gender, language and hospital area to ensure representativeness of the general population in the research regions. According to Finnish legislation (Medical Research Act 488/1999, (English translation available at http://www.finlex.fi/en/laki/kaannokset/1999/en19990488
), ethical approval is needed only for medical research, that is defined as research involving interventions. Thus ethical approval is not needed for e.g. register-based research, opinion polls or anonymous general population postal surveys. The current study was part of a repeated anonymous general population postal survey, performed every three years. Neither this study, nor the repeated general population survey, are "medical research" according to Finnish legislation, and statutory ethical committees will not deal with studies that are perceived as not being "medical research". Thus ethical approval was not needed, nor applied for.
The socio-demographic background variables were gender (coded as 1 = male, 2 = female) and age (year of birth).
Respondents who fulfilled self-reported criteria for major depressive disorder (MDD) according to the Diagnostic and Statistical Manual, fourth edition (DSM-IV) within the last twelve months were identified using questions from the Composite International Diagnostic Interview Short Form (CIDI-SF)[25
]. With this instrument we can both estimate the occurence of depression and its severity.
Professional help-seeking was ascertained by asking: "Have you during the past 12 months used any health services because of mental problems?". Response choices included "yes" and "no" (coded as 1 = used services, 2 = not used services). We also asked about the use of different types of mental health services by asking: "During the last 12 months, did you seek help from any of the following service institutions in respect of a mental health problem" and gave respondents 12 alternatives.
Sixteen statements exploring attitudes to and stereotypes of mental health were developed based on earlier studies measuring public attitudes towards mental health problems and also on researchers' clinical experience (Table ). Eight of the statements related to mental health problems in general and eight to depression only. Three of the statements referred to perceived public stigma/stereotype awareness and the rest to personal stigma/stereotype agreement. A four-point rating scale was used with the response alternatives: "strongly disagree", "disagree", "agree" and "strongly agree"
Results of the Principal Components Analysis (followed by Varimax rotation) applied to the 16 items data collected in 5160 population sample
Our first scale in this analysis, "Depression is a matter of will", measures negative stereotypes about people with depression and the belief that people with depression are responsible for their illness and their recovery. It was built from following five statements measuring personal stigma:
1. "Depression is a sign of failure"
2. "People with depression have caused their problems themselves"
3. "Depressed people should pull themselves together"
4. "Mental health problems are a sign of weakness and sensitivity"
5. "Depression is not a real disorder"
These statements were extracted by principal component analysis (PCA)[26
]. Prior to performing the PCA the suitability of the data for factor analysis was assessed. Inspection of the correlation matrix revealed the presence of many coefficients of 0.3 and above. The Kaiser-Meyer-Olkin value was 0.830, above the minimum recommended value of 0.6 and the Bartlett's Test of Sphericity reached statistical significance (p = 0.000), suggesting that a factor analysis was appropriate.
The PCA revealed the presence of four components with eigenvalues exceeding 1, explaining 21.7%, 9.3%, 8.1% and 6.6% of the variance respectively (Table ). This model accounted for 45.7% of the total variance. To aid in the interpretation of these four components, a Varimax rotation was performed. An identical PCA was performed three years earlier in a similar population survey and it identified exactly the same structure of four components. This analysis is reported elsewhere [27
The main component, here called "Depression is a matter of will", consisted of eight items and accounted for 21.7% of the variance. If the three items with low loadings ("Patients suffering from mental illness are unpredictable","Depression can't be treated" and "You don't recover from mental health problems") are excluded, we have a feasible five-item-scale with an internal consistency of 0.70 and inter-item correlations from 0.38 - 0.50. A high score on this scale indicates a belief that a person is responsible for the cause and course of his or her depression, and also capable of recovering from the illness if sufficiently strong-willed.
Our second attitude scale in this analysis, here called "Antidepressant attitudes" consisted of the two items in PCA component 3 and accounted for 8.1% of the variance. This 2-items scale has a very low internal consistency of 0.42 but because these items are highly correlated, we use them as a measure of antidepressant attitudes/knowledge in this analysis. A higher score on this scale indicates a belief that antidepressants are addictive and have plenty of side effects.
Our third attitude scale in this analysis, here called "Desire for Social Distance", reflects personal desire for social distance. This scale was constructed from a different set of items contained in the survey questionnaire and is based on respondents' expressed willingness in four different imaginary situations to be in contact with a person who has mental problems:
1. "Would you be willing to marry or be in a common law marriage with someone, who has mental problems?"
2. "Would you be willing to give your child into the care of someone who has mental problems?"
3. "Would you be willing to choose someone who has mental problems as your work colleague?"
4. "You find out that a rehabilitation centre for patients with mental illnesses is being planned in your neighborhood. Would you object to the plans?"
The fifth question "A person you know is committed to psychiatric hospital care. Would you be willing to visit him there?" was not included in the scale to make internal consistency stronger and because of its poor ability to differentiate. A higher total score means less willingness to be in contact with a person who has mental problems. The internal consistency of this scale was 0.70 (Cronbach's alpha).
The connections between depression (as measured by the CIDI-SF) and components of personal stigma (as measured by the "Depression is a matter of will"-scale, the "Desire for Social Distance"-scale and the "Antidepressant Attitudes" - scale) were analyzed using logistic regressions. Age and gender were entered in this model simultaneously with attitude components.
The relative effects of these three attitude scales on 12-month help- seeking among persons with depression were also analyzed using logistic regressions. Age and gender as well as the degree of depression were entered in this model simultaneously with attitude components. All analyses were carried out with SPSS 16 software.