The Caring Behaviors Inventory-24 (CBI-24)
The instrument includes four correlated subscales: (I) the Assurance subscale, being readily available to a patient’s needs and security (8 items); (II) the Knowledge and Skill subscale, demonstrating conscience and competence (5 items); (III) the Respectful subscale, attending to the dignity of the person (6 items); and (IV) the Connectedness subscale, providing constant assistance to patients with readiness (5 items). Subjects are asked to rate each item on a six-point Likert scale ranging from 1 (never) to 6 (always). The caring behavior for each subscale as well as for the overall scale is calculated as the mean value within each separate scale [11].
Design and setting
The project used a traditional forward and back translation process as described by Guillemin, Bombardier and Beaton in 1993 [13], and developed further by Beaton et al [14] in 2000. The process contains six stages: (1) two simultaneous translations by bilingual experts; (2) expert review committee synthesis of the two translations to one; (3) blind back translation to the source language by two new bilingual experts; (4) expert review committee deliberations for consensus on version ready for testing; (5) pre-testing with cognitive interviews, and (6) psychometric evaluations.
The forward translations in the first stage were performed by two native Swedish speakers, one clinically active nurse researcher familiar with the field and one researcher naïve to the field. The second stage synthesis of the two translations was achieved during an expert review committee meeting where the two translations were compared, discussed and reflected on, resulting in one translated version ready for back translation. A person not involved in the translations acted as chairperson and mediator. The third stage back translations were performed by two native English speakers, one naïve to context, originally from North America, and one familiar with clinical work, originally from the United Kingdom. The study was conducted by researchers in university settings, in an urban area in Sweden.
Expert review committee deliberations
The fourth stage entailed expert review committee deliberations where reports from stages I–III were read, discussed and reflected on. This was a crucial step for cross-cultural adaptation [14] since the processes of forward and back translation may reveal discrepancies and highlight inconsistencies. The expert review committee comprised three translators and two researchers, where one acted as moderator. In this group, there was expertise and competence regarding research, method and language as well as the clinical context through experience as healthcare professionals. Discussion ensued and consensus decisions were made in the four areas of equivalence [14]: semantic (words mean the same, no double meanings), idiomatic equivalence (idioms hard to translate), experiential (daily life experiences) and conceptual (concepts used may differ across cultures and languages). The proceedings and discussions were meticulously documented for transparency and to provide a record of the decisions made.
Cognitive interviewing
Using cognitive interviewing with the target population has potential to influence reliability and content validity by assessing the clarity and relevance as perceived by the target population [15], in our case women with experiences of homelessness, nursing students and registered nurses. Cognitive interviews, using a “think aloud” approach [12] to explore face and content validity, were performed with women having experiences of homelessness (n = 5), nursing students (n = 5) and registered nurses (n = 5). Participants were purposively recruited to meet the inclusion criteria of either being a woman with experience of homelessness, a nursing student or a registered nurse interacting with persons in homelessness in their clinical work. Women were chosen since the CBI-24 will be used in a larger project focusing on women in homelessness. Participants were encouraged to fill out the questionnaire while thinking out loud and to give words to the thoughts going through their minds during the process. Probes were used by the interviewer to elicit further information regarding thoughts about questionnaire items, instructions, and response options. The interviews were audio recorded with the participant’s permission. A small voucher, valid in grocery stores or movie theaters, was offered at interview conclusion in appreciation of participation.
For analysis, interviews were listened to and participants’ thoughts were documented in a template in writing, first one interview at a time, and secondly compiled as a comprehensive summary of the whole. The expert review committee convened to discuss outcomes and suggested minor revisions of wording or sentence structure in a few questionnaire items.
Data and setting for the psychometric analysis
For the psychometric analysis of CBI-24 SWE, registered nurses and nurse students were asked to participate in the study by answering an anonymous online questionnaire containing the CBI-24 questions as well as some questions about background characteristics. Participants were approached face-to-face by researchers, in two clinical units and one university setting. In total, 234 individuals answered the questionnaire during October and November 2019.
Statistical analyses
Categorical data are given as frequencies and percentages, n (%), while ordinal and continuous data are presented as means and standard deviations (SDs). In the psychometric analysis, the main interest was to examine if the hypothesized CBI-24 factor structure fitted the collected data for the CBI-24 SWE. To this end, a confirmatory factor analysis (CFA) was applied, using a χ2-test to assess the overall model fit. A normed χ2-value < 5.0 was considered to indicate an acceptable model fit. The goodness-of-fit indices Comparative Fit Index (CFI), Standardized Root Mean Square Residual (SRMR), and Root Mean Square Error of Approximation (RMSEA) were used as heuristic measures of model fit. Values of CFI > 0.90 and SRMR < 0.08 were considered to indicate good model fits, while values of RMSEA ≤ 0.05 and RMSEA < 0.08 were considered as close and acceptable model fits, respectively. [16, 17] An item reliability R2 ≥ 0.40 was deemed acceptable. The R package ‘lavaan’ version 0.6–5 (Rosseel, 2012) was used for the CFA analyses, applying the full information maximum likelihood (FIML) estimator to handle potentially missing data. All statistical analyses were performed in R 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria), with P-values < 0.05 considered statistically significant.
All proceedings adhered to ethical principles outlined by the World Medical Association, [18] and the study had ethical approval from the Swedish Ethical Review Authority.