A cross-cultural adaptation and validation of a scale to assess illness identity in adults living with a chronic illness in South Africa: a case of HIV

Data source, collection and participants

The current validation study forms part of a larger research study which aims to investigate illness identity in PLHIV, entitled HIV Illness Identity, Household HIV Competence and Antiretroviral Treatment Adherence: an analysis of associations among treatment naïve adult population living with HIV and for which a detailed research proposal has been articulated. The illness identity in HIV research study is in turn embedded within the Sinako cluster randomised control trial [41, 42]. The Sinako trial and intervention are described elsewhere [41, 41, 42, 42]. The trial aims to use community healthcare workers (CHWs) to stimulate HIV competence among PLHIV and the households, and explore the impact of the intervention on ART adherence outcomes and in turn enhance the routine support provided by CHWs [41, 42].

Data was obtained from the baseline phase of the trial. Due to the global COVID-19 pandemic and subsequent national lockdowns that were instituted by governments however, data collection activities had to be suspended until it was deemed relatively safe to continue. The study setting is five sub-districts of the Cape Metro health district in the Western Cape, South Africa; Mitchell’s Plain, Khayelitsha, Klipfontein, Eastern and Western. Unemployment is widespread and these communities are largely poor [55]. HIV prevalence in the Cape Town metro was estimated at 21.6% in 2015. Khayelitsha sub-district has been identified to have the highest overall HIV prevalence rate, 29.5% and 34.3% among pregnant women within the Western Cape province [33, 61]

Following a process of adaptation and development of a baseline questionnaire, data was collected from adults living with HIV and relatively recently started on ART visiting the 12 participating healthcare facilities and were recruited for participation in the study. Baseline data was collected between 8th October 2019 and 13th March 2020 by trained fieldworkers from 152 PLHIV who all provided informed consent. Data collection activities were then suspended due to the global COVID-19 pandemic which resulted in a total national lockdown. Ethical clearance to conduct the study was obtained from both the University of the Western Cape’s Biomedical Research Ethics Committee (BM19/4/6) and the Ethical Committee for the Social Sciences and Humanities of the University of Antwerp (SHW_17_64). Permissions to access health facilities were also issued by the Western Cape Department of Health and the City of Cape Town [41, 42].

Process for adaptation and validation

This study employs an approach initially proposed by Herdman et al. [26] as a systematic guide to direct the cross-cultural adaptation and validation process for the IIQ (Table 10 in Appendix) [20, 52]. With this approach, the constructs: illness identity, and its sub-scales acceptance, rejection, enrichment and engulfment, are assumed to be different across contexts, in the current case, Belgium and South Africa. It is therefore necessary to explore whether the concepts exist and whether they are interpreted the same way in this context as was in the original context [26, 52]. This process involved four-phases.

The first phase employed a literature review to evaluate conceptual and item equivalence of the illness identity construct and its domains in the target context. For this, we undertook a general literature review using the EBSCOhost interface, which provided access to a number of databases and the PubMed database. We followed a process of keyword, backward and finally forward searching to retrieve relevant published articles. This process gathered information from studies that were of relevance, important and valid for the illness identity construct. Using critical analysis, we especially scrutinised the literature for papers on illness identity, its related or similar concepts and or its four domains of acceptance, rejection, denial and enrichment, published in both target (South African or similar) and original contexts. The approach not only provided basis and the rationale for the overall illness identity study, but in this way, we could evaluate whether the illness identity domains had the same relationship and meant the same thing in both the original and target context. Additionally, we conducted a workshop with research fieldworkers, who represented the target population for our research study, to explore meanings of items within the IIQ from an emic perspective and advise on their appropriateness, relevance and acceptability [26, 52].

The second phase explored semantic equivalence using the technique of adaptation, also referred to as cultural substitution [43], to translate the instrument from the original English into the relevant local languages; isiXhosa and Afrikaans. This involved replacing the original text with appropriate local language making the text easy to comprehend. In some cases, the borrowing translation technique was used [43]. This is when direct translation with a word or phrase made the translation unnecessarily formal or complex and local speakers are likely to understand the original language (in our case, English) word. The technique was used only in a few instances where it was impossible to substitute a word. The translation process involved both forward and back-translations. The initial translation of the IIQ into isiXhosa was conducted by a first language isiXhosa speaker who is also fluent in English. This version was reviewed by another person fluent in both languages. Thereafter, two independent translators back translated the isiXhosa version to English. The two back translated versions were then compared against the original IIQ using two measures; comparability of language and similarity of interpretability by a person fluent in English [59]. For this, a seven-item Likert scale ranging extremely comparable/similar), through moderately comparable/similar to not at all comparable/similar was used.

The third phase assessed the operational equivalence and explored the potential for using the original questionnaire format, instructions, mode of administration and measurement methods in the context.

In the fourth and final phase, we assessed validity by exploring the psychometric properties of the IIQ. This involved three steps: (1) evaluation of the factorial validity and reliability; (2) given that there were no differences in illness identity based on age, gender and illness duration in the original validation of the IIQ using adolescents and emerging adults with type 1 diabetes [47], we also aimed to investigate whether consistency of the results for these two demographic variables and one clinical parameter could be established in adults living with HIV, and (3) finally, we explored association of the different illness identity subscales with HIV related stigma and disclosure of HIV status. We hypothesized that acceptance and enrichment would be positively correlated and that rejection and engulfment be negatively associated with disclosure of HIV status and HIV related stigma [7, 73].

MeasuresIllness identity questionnaire

The IIQ has four illness identity subscales: engulfment, rejection, acceptance and enrichment, represented by 25 items and participant responses are captured on a five-point Likert scale ranging from “strongly disagree” to “strongly agree” [46, 47].

When engulfed by illness, chronically ill individuals describe themselves according to their illness, they perceive the illness to be intruding in all areas of their lives and the illness seems to control their identities and routines, compromising other significant self-assets [46, 47]. Rejection on the other hand describes the extent to which the illness is rejected as part of their identity [46, 47, 70]. As such, the rejection and engulfment scales capture the lack of integration into self of the chronic illness. The engulfment subscale consists of eight items and the rejection subscale five items [46, 47].

Acceptance is the degree to which individuals accept their illness as a part of their identity, regardless of other social roles and identity assets. With acceptance, chronically ill individuals acknowledge their illness, they are not overwhelmed by it and the illness does not pervade other areas of their lives. In fact, with the acceptance process, they try to live their lives as normal as their illnesses allow and the illness plays an insignificant role in their lives [39, 44, 46, 47]. Enrichment indicates the extent to which having a chronic illness has either resulted in positive life changes, benefited the ill individual’s identity, and or facilitated their personal growth [46, 47, 70]. To this end, the acceptance and enrichment states represent more adaptive illness integration. The acceptance subscale consists of five items and the enrichment subscale is comprised of seven items [46, 47, 70]. The IIQ has been validated and used previously to measure illness identity in high income contexts among individuals with chronic illnesses including diabetes, refractory epilepsy and congenital heart disease [39, 47, 69, 70].

HIV disclosure questions

Two aspects of disclosure were explored: participants’ inclination to disclose their HIV status and situational disclosure. Inclination to disclose was estimated by how strongly participants agreed with the statements that assess the comfort and perceived ability to reveal their HIV status and situational disclosure by participants indicating how strongly they agreed or disagreed with a statement about comfort but only when necessary.

HIV related stigma

For HIV related stigma, we used the truncated HIV stigma scale previously validated by Reinius et al. [53], adapted from a longer 40-item HIV stigma scale [9]. Whilst other stigma scales such as the “People Living with HIV Stigma Index” [19] exist, we found the utility of the shortened scale appropriate for our context since the focus of our inquiry was to determine external validity of the IIQ and not stigma per se. The 12-item version is comprised of four subscales; personalised stigma, disclosure concerns, concerns about public attitudes and negative self-image measured by three items each [53].

Statistical analysis

The IIQ has been previously validated [47] and therefore as an initial step, we conduct a confirmatory factor analysis (CFA). The condition is that if the original model fits the data then it will be accepted and adopted as is with the item and or semantic and operational adjustments that are made for the study context described above. To evaluate model fit, we use a 2-index presentation strategy and because our sample size was relatively small (N ≤ 250), we adopt the combinational rule based on the comparative fit index (CFI) > 0.96 in combination with the Standardized Root Mean Squared Residual (SRMR) < 0.09 [30]. For transparency, we also evaluate and compare other fit indices obtained in the current study against those that were used and assessed in the original IIQ validation. These include the normed chi square, which is the ratio of the chi-square statistic to the respective degrees of freedom (χ2/DF), the cut-off is set at less than 2 and the root mean square error of approximation (RMSEA), a parsimony adjusted index, which should be less than 0.08 [46].

As a contingency, we conduct an exploratory factor analysis (EFA) of the original 25-item model, evaluating it for any redundant items [32, 71]. For the EFA, to evaluate factorability of the data, sample adequacy and non-randomness of the correlation matrix criteria were used. For this, the Kaiser–Meyer–Olkin (KMO) statistic [34], which is required to be above 0.50 and the Bartlett’s test of sphericity [6], required to have a significant p-value, are used. The correlation matrix is then submitted for exploratory factor analysis. Because our intention was to examine the factor structure of the IIQ rather than a reduction exercise, we use the common factor analysis instead of the principal component analysis. For this we elect principal axis factoring extraction method with initial communalities estimated by squared multiple correlations which has been demonstrated to be relatively accommodative of non-normality of the data distribution and has capacity to retrieve weak factors [71]. As recommended we use the visual scree test and parallel analysis to determine the number of factors to retain [71]. Although a common practice, we elect not to use eigenvalues to determine the number of factors to retain as the method is cited as often inaccurate and its use is discouraged [71]. Furthermore, an oblimin rotation is employed because of the understanding that the factors are indeed correlated. Redundant items are defined as those items with: pattern coefficients < 0.5, or with cross/complex loadings that were salient (> 0.4) on more than one factor. In these cases, the item is removed from the model to fulfil the simple structure principle [66, 71]. For a factor to be considered adequate, it is to have a minimum of three theoretically meaningful and salient pattern coefficients > 0.4. We then recalculate the adjusted model and submit it for another CFA. We re-specify the model by progressively removing the items with poor standardised factor loadings and reassess its goodness of fit with the data. We eventually undertake an exploratory phase of inspecting the modification indices of all the pairs of error terms and correlating those pairs with the largest indices until the model fitted [2].

Reliability, convergent and discriminant validity of the final measurement model is assessed using different estimates including correlations, the average variance extracted (AVE) which should be ≥ 0.5 and composite reliability (CR) which should be ≥ 0.7. Although the CR is a less biased estimate of reliability than Cronbach’s Alpha, we also report on the latter estimate. Multivariate analyses of variance (MANOVA), using Wilks’ Lambda, are also used to test for mean differences in illness identity as dependent variable based on gender. For age and illness duration, Pearson correlation coefficients are calculated with the four illness identity states. To examine the associations linking illness identity to aspects of disclosure of HIV and internalised stigma, Kendall’s (τb) and Spearman’s correlation coefficients are calculated. All analyses were conducted with the IBM SPSS Statistics for Windows, (Version 27.0.; released 2020) and the IBM SPSS AMOS 27.0.0 package (Version 4.0.30319.42000).

留言 (0)

沒有登入
gif