Study design and setting
The study had a cross-sectional design and was carried out in 2013 in four counties, in different demographic and geographic areas of Sweden. All asylum seekers in Sweden must by law be offered a health examination, unless it is clearly unnecessary, free of charge [30]. The term asylum seeker in this article is used for someone whose refugee claim has yet to be definitely evaluated. The HEA can be carried out either before or after having received a permanent resident permit as a seeker of asylum. The HEA must include a dialogue about the participant’s earlier physical and psychological health and vaccinations and questions of importance from the point of view of infection control [4]. Physical examinations and clinical tests are based on what emerges in the dialogue. The participants should further be given information about their right to health care and how to access health care in Sweden.
Ethics approval was sought at the regional Ethical Committee of Clinical Investigation in Uppsala, Sweden, registration number 2012:506, but a committee judgment was not deemed necessary or applicable according to Swedish law, since data collection was performed anonymously, leaving no possibility of individual identification.
Study population
The target group was adult refugees in schools offering Swedish for immigrants (SFI). Inclusion criteria were: respondents eligible on the day of the data collection; speakers of Arabic, Somali, Dari or English; born outside a Nordic country or the EU; and having received a permanent resident permit as a seeker of asylum. This means that refugees were asked about their participation in a HEA that was done either before or after having received a permanent resident permit as a seeker of asylum. Of 455 eligible SFI students, 360 fulfilled the inclusion criteria as well as participated in HEA and, thus formed the study population in the present study.
Data collection
A strategic selection of SFI schools was based on the number of people who had received resident permits as asylum seekers in each municipality in 2012, in each county [9]. Eligible SFI schools with more than 30 students were selected from each municipality in the four counties. If more than one school in a municipality fulfilled the criterion, a school was randomly drawn. Nineteen schools agreed to participate and were visited to collect data.
A team consisting of a researcher (first author) and a number of language supporters visited each school to collect data. The SFI students were verbally informed about the project and voluntary participation. Eligible participants were grouped on the basis of their native language. They were then informed about the project, again in their native language, and asked for verbal consent. Those who consented to participate were given the questionnaire, which was completed on site. Twenty one percent of the participants (70 out of 360) who had difficulty reading or writing were assisted by language supporters that read and supported them in filling in the questionnaire.
Material and analysis
Data were collected using a questionnaire with 60 questions that focused on health, HL and experiences of HEA. The questions that concerned HEA were based on the results of an earlier qualitative explorative pilot study (Åkerman E and Wångdahl J, 2014, “Unpublished observation”). The questionnaire was translated by language supporters into Arabic, Somali, Dari and English following guidelines for the translation of instruments [31].
Characteristics
The socio-demographic and health-related characteristics under examination were sex, age, country of birth, education level, self-assessed general state of health and years of having had a resident permit. Background questions related to the HEA were whether the individual had participated in an HEA in Sweden, when he or she had participated and whether support by an interpreter had been given. FHL was measured by the Swedish FHL scale (S-FHL scale) which consists of five items assessing different aspects of FHL [32] (Additional file 1).
When analysing the data, an overall level for FHL was calculated for each respondent [9]. Persons responding “never” or “seldom” to all items were categorized as having sufficient HL. Persons scoring “often” or “always” to one or more of the five items were categorised as having inadequate HL. The rest, those who responded “sometimes” to at least one item and not “often” or “always” to any items, were categorised as having problematic HL. The cut-offs used when dividing the respondents into the three groups were based on definitions describing the abilities needed for sufficient FHL, i.e. basic skills in reading information and instructions about health [7]. Those lacking any basic skills were classified as having inadequate FHL, those having all the basic skills were classified as having sufficient FHL and those in between were classified as having problematic FHL.
CHL was measured by a slightly modified Swedish version of the short European HL questionnaire (HLS-EU-Q16) [9, 33] (Additional file 2). This focused on four HL dimensions: ability to access, understand, appraise and apply health information. An overall HLS-EU-Q16 index (CHL) was calculated in three steps when the data were analysed, following the manual for the instrument [33]. The response categories were first dichotomized [9]. The responses “fairly easy” and “very easy” were put together into one category, which was given the value of 1, the responses “fairly difficult” and “very difficult” were put together into one category which was given the value of 0, and the response “don’t know” was treated as missing. Second, a sum score of the response values was calculated and divided into three categories: sufficient CHL if there were 13–16 score points, problematic CHL if there were 9–12 score points or inadequate CHL if there were 0–8 score points.
In the case of missing values in two or more items of the HLS-EU-Q16 scale, any missing value on an item was substituted with “difficult” and a new CHL value was calculated, i.e. all participants thereafter had a CHL value (CHLm). CHLm was used in analyzing associations between experiences of HEA and HL, and what could predict poor experiences of communication and the usefulness of HEA. The treatment of missing values was chosen in a dialogue with Florian Röthlin, statistician at the Ludwig Boltzman Institute for Health Promotion research (“personal communication”, October 28, 2014), who worked with the development of the HLS-EU-Q16. The multiple imputations method was considered, but this would have changed the scoring algorithm (ibid). Thirty six percent (123 out of 360) of the participants had an incomplete CHL value; thus the CHLm was calculated for those participants.
The S-FHL scale and HLS-EU-Q16 measure quite different aspects of HL. The S-FHL scale is focusing on the individual’s abilities to read and understand written health information. The HLS-EU-Q16 focuses on cognitive abilities in order to access, understand, appraise and apply oral health information as well. In dichotomizing FHL/CHL, inadequate and problematic FHL/CHL were merged into limited FHL/CHL. The scales measuring health literacy have been validated by carrying out cognitive interviews concerning the content with refugees comparable with the study group.
Experiences of communication and the usefulness of HEA
Experiences of communication and the usefulness of HEA were measured in four dependent variables (Additional file 3). Quality of communication, measured the experience of communication, used in this study as an umbrella term for different aspects of communication. Quality of communication was measured in four questions in an attempt to omit different aspects of communication. The response alternatives “no”, “partly” and “yes” were assigned values from 1 to 3, respectively, yielding an index with a maximum total score of 12. The usefulness of HEA was measured by three dependent variables: receiving health care information, receiving new knowledge and receiving help. Receiving health care information was measured in three questions. The response alternatives “no” and “yes” there were values from 1 to 2, yielding an index with a maximum total score of 6. Receiving new knowledge and receiving help were measured using one question each. The response alternatives “no”, “partly” and “yes” for those dependent variables were assigned values from 1 to 3, respectively, yielding indexes with a maximum total score of 3. The responses of participants who answered “no” or “don’t know” to the questions on participation in HEA were not included in the analysis. Values in the response categories “don’t remember” in the questions on the quality of communication and usefulness of HEA were treated as missing.
Statistical analysis
The statistical analyses used SPSS version 21.0 (Chicago, IL, USA). Chi square tests or Fisher’s exact tests when the expected frequency in each cell was lower than 5, were used to compare proportions of experiences of various aspects of HEA in groups with different FHL and CHL levels. Binary logistic regressions were done to calculate the crude odds ratios for the effect of socio-demographic factors, support of an interpreter and HL levels in the four dependent variables. Cronbach’s alphas were calculated to explore the internal consistency of the questions used for quality of communication and receiving health care information. The internal consistency was satisfactory (Cronbach’s alpha = 0.79 and 0.71, respectively).
Multiple binary logistic regression was used to investigate the association between FHL, CHLm and the four dependent variables focusing on different experiences of HEA, adjusting for sex, age, education, country of origin and support of an interpreter. The multivariate model (Tables 3 and 4) included sex, age, education, country of birth, interpreter support, FHL and CHLm. Results are presented as crude odds ratios and adjusted odds ratios (OR) with 95 % confidence intervals (CI) and p-values. A p-value of < 0.05 was considered statistically significant, and all analyses were two-sided. The dichotomized cut-offs for the dependent variables in the logistic regression analyses were suggested by the data. Each dependent variable was dichotomized into a higher (median and above) and a lower (below median) group based on the median.
In addition to the binary logistic regression analysis including CHLm (all participants), logistic regressions were performed with CHL, only including participants having a valid CHL from the beginning. The results for HL were not changed in any major sense.
Sensitivity analyses
Alternative cut-offs for the dichotomous dependent variables were tested to identify the robustness of the results of the multivariate analysis. The cut-off levels for quality of communication and receiving health care information were moved downwards one step, and the cut-off levels for receiving new knowledge and receiving help were moved one step upwards. Thus, the results for HL were not changed in any major sense.