Development and psychometric testing of a barriers to HIV testing scale among individuals with HIV infection in Sweden; The Barriers to HIV testing scale-Karolinska version

This was a cross-sectional study within the Swedish national project “Late Presentation of HIV-1 infection” lead by Karolinska University Hospital, Sweden. Aim of the national project was to identify factors in HIV-infected patients and the health care system that contribute to late diagnosis. The project is described elsewhere [20]. Eligible for participation were adults living in Sweden, diagnosed with HIV infection from October 2009 to January 2012, with data collected within six months after their diagnosis.

For the present study, an 18-item scale was developed, based on twelve items from the existing barriers to HIV testing scale by Awad et al. [14] and six new items. Dimensions measured in the original scale are structural barriers, fatalism/confidentiality concerns, and fears. The scale was evaluated and translated into Swedish by a bilingual panel of HIV experts. One item concerning costs of treatment was excluded in the new scale due to the general availability of treatment for free in Sweden. The six new items were added, based on the literature and clinical experiences, to expand on different feared consequences of being diagnosed with HIV [12, 13, 20]. Appropriateness of the content and phrasings of the new items was discussed with professionals from participating clinics. Three of the new items concerned potential consequences in social contacts and relationships of being diagnosed with HIV: fear of losing one’s family or friends (e.g. [21‐24]) and fear of negative consequences in sexual life [25]. The remaining three items concerned other potential negative consequences for the individual of being diagnosed with HIV infection: fear of becoming ill [12, 13], worries about legal consequences [10] and fear of feeling like a failure. The topic of fear of legal consequences has been shown relevant from an international perspective [10] and is, according to clinical experiences and previous research [26] also relevant for a Swedish context where people living with HIV under the law are obligated to disclose their HIV status to sexual partners and when seeking health care. The eighteen items were phrased as statements about potential barriers and the respondents are instructed to rate the importance of the barriers described on a three-point scale from 0–2. The response alternatives and their respective scores are not important (0), somewhat important (1), and very important (2). The simple response format was chosen to make the scale feasible for a broad population of respondents, including people with limited language skills and literacy. Swedish and English versions of the scale were developed simultaneously (new items in English adapted by the bilingual expert panel). Items in the scale, titled the Barriers to HIV Testing Scale – Karolinska version, are presented in Table 1.

Patients from three of the largest HIV clinics in Sweden and eight county clinics distributed throughout the country were eligible for participation in the present substudy. Inclusion criteria were being ≥ 18 years old and diagnosed with HIV infection up to 6 month before completion of the questionnaire. Since a majority of those living with HIV infection in Sweden are migrants and since individuals not speaking and understanding a main language are often excluded from research, an explicit goal of the study was to also include individuals with low comprehension of Swedish.

The inclusion process is presented in Fig. 1. Of 445 eligible participants, 308 chose to participate in the study, and 292 completed the barrier questionnaire. Seven participants had missing values on half of the items or more and were excluded. The remaining 285 individuals had completed at least two thirds of the items and were included in the psychometric evaluation. The participants were between 18 and 70 years old (mean 40.5, SD 11.5), 39 % were females and 77 % born outside Sweden. Routes of transmission were heterosexual transmission (HT) 63 %, male to male sex (MSM) 20 %, intravenous drug use (IDU) 5 %, blood product/transfusion 2 %, and unknown 9 % (Table 2). Sixteen percent of the participants completed the Swedish version of the questionnaire with assistance of a professional interpreter (in-person or by telephone), who translated the items into the respondents’ language of origin. The English version of the questionnaire was used by 27 % of the participants. Participants did not differ from non-participants regarding sex, age or route of transmission, but were less likely to be born outside Sweden (OR 0.51, CI 0.32–0.80, p 0.004). Furthermore, the participants were representative for the total population of individuals with newly diagnosed HIV in Sweden during the study period (N = 827) [27] with regard to gender and origin, but were slightly older (mean age 40.5 vs. 38.9, t(284) 2.35, p 0.020) and had a lower representation of individuals with MSM as route of transmission (OR 0.68, CI 0.49–0.94, p 0.018).

Eligible patients were informed about the study by their treating physician in connection with a visit at their outpatient clinic. They were given oral and written information about the purpose of the study and the voluntary nature of participation. The barrier questionnaires were administered by staff at the clinic, who could also answer questions regarding the questionnaire. Professional interpreters were available in-person or by telephone for participants who did not speak Swedish or English. Demographic data was collected from the Swedish InfCare HIV registry [27]. The study was approved by the Regional Ethical Review Board of Stockholm, Sweden (2009/1029-31/1-4).

Participants were compared with non-participants and population data with Pearson’s chi-square tests for dichotomous data (each route of transmission dichotomized as present/absent) and t-tests for continuous data (independent two-samples t-test and one-sample t-test for comparisons with non-participants and population data, respectively). Data quality was assessed by inspection of missing values, means and standard deviations as well as floor and ceiling effects. Items were considered feasible if they had less than 3 % missing values [28]. Questionnaires with at least two thirds of the items completed were considered acceptable for the data analyses, except for the factor analysis which was calculated on complete questionnaires. Evidence for construct validity was investigated with exploratory factor analysis. The sample size was considered adequate for exploratory factor analysis of the 18-item questionnaire [29]. The Kaiser-Meyer-Olkin measure of sampling adequacy (KMO) [30] and Bartlett’s test of sphericity were used to determine adequacy of the data for factor analysis [31]. Factors were extracted with principal axis factoring. Oblimin rotation was used since the factors were expected to be related [32]. The final factor solution was based on comprehensibility and interpretability together with level of explained variance. Scales were constructed from the factors, where items with loadings ≥ 0.32 [32] were assigned to the scale with the highest loading. Items with loadings ≥ 0.32 on two or more scales were considered as cross-loading items [32]. Scale means were calculated by averaging completed items on each scale. Scale reliabilities were assessed with Cronbach’s α [33], where the reliability was interpreted as; α ≥ 0.9 excellent, 0.7 ≤ α < 0.9 good, 0.6 ≤ α < 0.7 acceptable, 0.5 ≤ α < 0.6 poor, and α < 0.5 unacceptable [34]. Bivariate correlations between scales were calculated with Spearman’s rho. All statistical analyses were conducted with IBM SPSS 22 (IBM Corp., Armonk, NY).