117 individuals completed the survey, and 59 of these participants recompleted the BACE to provide test-retest reliability data. Those providing the test-retest data did not differ significantly from the remainder of the sample in age, gender, ethnicity, education, employment status, age at first treatment, and hospital admittance for psychiatric treatment, although those with non-psychotic conditions were more likely to be in the retest sample (66% vs 34%, 2
0.024) compared to those with psychotic conditions. The characteristics of the sample are shown in Table . Participants had a mean age of 36 (range 18 to 70), 80% were female, the majority (87%) reported their ethnicity as White British, and 42% were in full- or part-time employment. The most common self-reported primary diagnoses were depression (34%) and bipolar disorder (31%). Forty six percent had been hospitalised for a mental health problem.
Participant sociodemographic and clinical characteristics
Development of the BACE v1 and v2
There were 172 barriers items in the 23 papers assessing barriers. This was reduced to 30 items following deletion and amalgamation by the research team. The team identified eight additional items from their knowledge of the research literature. Consequently the BACE v1, sent to the expert panel, had 38 items. Four of these items were deleted following feedback from the panel because they were viewed as ambiguous or already covered by a similar item. Two new items (20 and 21 in Table ) were added at the expert panel’s suggestion making the BACE v2 a 36 item measure. Minor rewording was made in five items (revised items 3, 12, 16, 24 and 34 in Table ), and two items were reworded for conceptual reasons (revised items 9 and 11 in Table ). The research team designated 13 of the BACE v2 items as stigma-related, 23 as non-stigma related barriers, as shown in Table . One item (‘Difficulty taking time off work’) was viewed as potentially both instrumental and stigma-related because it risked disclosing the mental illness. A further two items (‘Concerns about the confidentiality of the information I share’ and ‘Dislike of talking about my feelings, emotions or thoughts’) were viewed as potentially both attitudinal and stigma-related. A decision was made to only designate items stigma-related if when there was no alternative potential designation for the barrier. The items that were designated stigma-related spanned anticipated discrimination in employment (items 6 and 33 in Table ) and in relation to parenting (item 29), social stigma (items 9, 16 and 31), disclosure concerns (items 19 and 25), stereotypes (items 3, 14 and 21), internalised stigma (item 11) and stigma by association (item 26).
Distribution of the BACE treatment stigma subscale
BACE treatment stigma scores varied from 0 to 3 with a mean of 1.43 (sd 0.73). The BACE treatment stigma scale was normally distributed (Kolmogrov-Smirnov Z
0.454), therefore parametric statistics are appropriate for use with this scale.
The majority (22/36) of the BACE v2 items had weighted kappa values from 0.61 to 0.80 indicating substantial agreement between test and retest [51
]; two items (12 and 29 of the BACE v2 shown in Table ) had values above 0.8 indicating almost perfect agreement; 11 items had values between 0.41 and 0.60 indicating moderate agreement, and one (item 3, ‘Concern about being seen as weak for having a mental health problem’) had a kappa of 0.346 meaning fair reliability, but not reaching the pre-specified criterion of 0.4. Lin’s concordance statistic for the treatment stigma subscale was ρc
0.816 exceeding the criterion of 0.70 for acceptable test-retest reliability. The Cronbach’s alpha value for the treatment stigma subscale was 0.89 indicating good internal consistency.
Twenty two participants gave free text answers in response to questions about other barriers. Content analysis the 36 additional barriers revealed 20 were already covered by the existing items, and 10 were proposed by one respondent only, thereby indicating good content validity. Provider delay and providers not responding to requests for help were mentioned as a barrier by four people, which suggests this is an aspect for further study. The hypothesised significant positive correlation between the BACE treatment stigma subscale and the SSRPH [41
] was supported (r
0.001), as was the same hypothesised relationship between the BACE treatment stigma subscale and the Internalised Stigma of Mental Illness Scale [47
0.001). Thus the subscale has convergent and hence construct validity.
Respondents gave a median overall evaluation rating of the BACE of 8 (IQR 7-9) on the 10-point scale, indicating a positive respondent opinion of the measure. The Flesch Reading Ease score for the BACE v3 was 78.8, indicating it is easier to read than documents at the general population level of 60-70. Its Flesch-Kincaid Grade Level was 5.9 indicating that it can be understood by the average 11 to 12 year-old.
Finalisation of the BACE v3
The research team deleted five items and amalgamated two items on the basis of item analysis and conceptual discussion, as well as making some final wording changes. Items 10 and 23 were removed because they were very highly correlated with another variable (rho
0.7) and were highly correlated (rho
0.5) with a further three variables. No other variables had this degree of inter-correlation. Items 22 and 32 were deleted because fewer than 10% of respondents rated the issue as a major barrier. Three other items (12, 17, 28, and 31) also met this criterion but were retained, the first two because they were relatively highly endorsed by non-White participants; the third because the literature suggests that informal care seeking is a common factor in delay in seeking professional care [57
] and perhaps would be more evident in less unwell populations; and the fourth because concern about stigma from friends has been shown to be an important factor in young populations [22
]. Item 3 which did not meet the moderate test-retest reliability criterion was considered for removal but retained as its reliability level was ‘fair’ [51
] and because it was a relatively highly rated barrier (ranked 12/36). Two stigma-related items (9 and 26) were amalgamated with each other because only 11% rated the latter as a major barrier and because it had some conceptual similarity to the former. Item 36 was removed because it was considered to have a large conceptual overlap with item 25. Thus the BACE v3 has 30 items including 12 that are stigma-related. Following team discussion conceptual rewording was made to items 9, 29, 31 and 33 and minor rewording made to item 10. When asked for free text comments on the BACE, several respondents pointed out the need to have a ‘not applicable box’. These have now been added to the items referring to employment and to children. The scoring for the treatment stigma subscale was consequently amended from the mean rating to the mean of ratings for applicable items. This scoring method has been used successfully for other measures with varying numbers of applicable life domains, such as the DISC [58
]. This finalised version of the BACE can be seen in Additional file 1
Prevalence of stigma-related and non-stigma-related barriers to access to care
The prevalence of stigma-related and non-stigma-related barriers to care-seeking for mental ill health reported by this sample is shown in Table . The top two barriers – ‘concern that it might harm my chances when applying for jobs’ (item 6) and ‘concern that I might be seen as a bad parent’ (item 16) - were both stigma-related with 39% and 38% reporting these as major barriers respectively and 89% and 88% experiencing them to some degree. The next most highly ranked barriers were difficulty taking time off work, being too unwell to ask for help, having had previous bad experiences, wanting to solve the problem on one’s own, and not wanting a mental health problem on one’s medical records with only the latter being a designated stigma-related barrier. Some types of stigma-related barriers were relatively rarely endorsed such as concern about it bringing shame or disapproval on one’s family and about what friends might think or say with 11% and 7% reporting these as a major barrier respectively.