Study population and procedures
An observational cross-sectional ex-post-facto study was conducted in which 1462 PLHIV participated. They were recruited by convenience sampling. The general inclusion criteria were positive HIV diagnosis, being at least 18 years old, on antiretroviral therapy (ART) for at least one year, and not having any severe psychiatric or cognitive disorder. Data were collected between October 2016 and April 2017.
An online survey was designed using the Qualtrics survey platform (available at: www.qualtrics.com). Qualtrics is a private online survey development platform that allows the creation of surveys which can be accessed through a link. In the present study, our survey was self-administered with the support of tablet computers. Service providers from 33 service delivery points across Spain (hospitals and NGOs) collaborated in the participants’ recruitment and data collection. During their medical consultations or when attending various services, the collaborating service providers explained the goals of the study to the participants, requesting their participation and obtaining their informed consent. The rate of refusal to participate in the study varied across centers, ranging from 0 to 18%, with an average around 7%. The main reasons argued for the refusal were not having enough time, the survey length or lack of skills to use tablets. Participants were compensated with 15 euros.
The Ethics Committee of the Hospital Clínico of Valencia approved the research protocol in March 2016. All study procedures were conducted in accordance with the 1964 Helsinki Declaration (revised in 1996) [12].
Measures
WHOQOL-HIV-BREF
WHOQOL-HIV-BREF has 31 items covering six domains: physical health; psychological health; level of independence; social relationship; environmental health; and spirituality, religion and personal beliefs (SRPB) [8]. Responses to all items are given on a 5-point scale. Items that ask about negative perceptions and experiences, such as “How much do you fear the future?” are reverse-coded for scoring. Thus, higher scores for all items indicate better quality of life. The average score for each domain is multiplied by four, yielding domain scores that range from 4 to 20 [13].
Several studies have examined the validity and psychometric properties of WHOQOL-HIV-BREF in different languages and countries. (A summary of these studies is presented in Additional file 1). They have found the instrument to have good psychometric properties and have also found evidence of its validity. The Spiritual, Religion and Personal Beliefs domain (SRPB) is the one that exhibited the lowest reliability (under .70) and discriminative power in most of the studies [8, 14,15,16,17,18], although it is the domain which contains more HIV-specific items measuring existential concerns relating to HIV.
Because WHOQOL-BREF has been validated in a Spanish study population [19], only the translation of the HIV-specific items was needed (see Additional file 2). The HIV-specific items collect information about the bother caused by physical problems related to HIV infection, HIV-related stigma, and fears related to the future and to death (“How much are you bothered by any physical problems related to your HIV infection?,” “To what extent are you bothered by people blaming you for your HIV status?,” “How much do you fear the future?,” “How much do you worry about death?,” and “To what extent do you feel accepted by the people you know?”). These items were extracted from five HIV facets of the WHOQOL-HIV long form and then integrated with the WHOQOL-BREF to complete the 31-item WHOQOL-HIV BREF [8]. They were translated following the criteria of the International Test Commission [20]. A backward translation was performed by two expert translators. In addition, a person with HIV reviewed the translation.
Questionnaire to evaluate the adherence to HIV therapy (CEAT-VIH 2.0 version)
The validated Spanish version of the Questionnaire to Evaluate the Adherence to HIV Therapy (CEAT-VIH 2.0 version) [21,22,23] was used. This scale is comprised of 17 items rated on a 5-point scale. Negative items were reverse-coded. A composite of all items (total score) was calculated, with higher the scores indicating higher treatment adherence. A systematic review of the psychometric properties of the CEAT-VIH including 20 studies revealed an adequate internal consistency as well as no floor or ceiling effects [23]. Additionally, evidence of validity comprised criterion-related validity (e.g., HIV viral load, length of time with continuous undetectable HIV viral load, days of missed doses, number of pills per day, and adherence assessed by the pharmacist or physician); responsiveness, sensitivity, and specificity; and patterns of convergence and divergence (e.g., negative mood, depression, anxiety and stress were negatively associated with CEAT-VIH scores whereas positive correlations with CEAT-VIH scores were found for perceived social support and quality of life outcomes).
General Health Questionnaire (GHQ-12)
The validated Spanish version of the General Health Questionnaire (GHQ-12) was used [24]. Items are rated on a 4-point scale, with higher scores indicating better psychological health. In previous studies, this scale has shown adequate reliability and validity in the Spanish general population [24].
HIV-related stigma
Five items of the Negative Self-image dimension and three items of the Disclosure Concerns dimension of the Spanish HIV Stigma Scale (HSSS) were used [25]. These items were selected for having higher validity constructs in the validation study of the scale. A previous study in Spain revealed that this scale shows good internal consistency and good construct validity, including content and criterion validity [25]. The items are rated on a 4-point scale.
Self-reported questions related to health status such years living with HIV, lymphocyte CD4 count and viral load copies were included in the survey. Moreover, the questionnaire also measured socio-demographic information.
Data analysis
Completing most items of the online questionnaire in the Qualtrics survey platform was programmed to be compulsory. Only items related to some sensitive characteristics of the participants were allowed to be skipped. Thus, there were no missing values in the tools used to measure the variables under analysis.
To test the construct validity, first-order confirmatory factor analysis (CFA) was used to assess the fit of the Spanish version of the WHOQOL-HIV-BREF to the six-dimension original structure (Model 1, [8]). Next, second-order CFA was performed to determine whether the six first-order factors could be explained by a higher-order latent factor associated with HRQoL. Previous studies showed that one of the items of the SRPB dimension presented a low facet-domain correlation [18] or factor loading [14, 17] and was saturated in the psychological domain rather than in the existential domain when exploratory factor analysis (EFA) was conducted [16]. Thus, we tested an alternative model (Model 2) allowing one of the items of the SRPB dimension—feelings of personal meaning—to load in the Psychological Health latent dimension. The robust unweighted least square method was used because the items in the scale did not meet the assumption of normality. Goodness of fit was evaluated using the goodness of fit index (GFI), the adjusted goodness of fit index (AGFI), the comparative fit index (CFI), the standardized root mean square residual (SRMR) and the standardized root mean square error of approximation (RMSEA). Also, the consistent Akaike information criterion (CAIC) was used to compare the alternative models. According to Hu and Bentler [26], the models are considered to have a good fit when the goodness of fit indexes (GFI and AGFI) and CFI > .90, RMSEA < .08 and SRMR <.08. The reliability of each domain was assessed using Cronbach’s α coefficient.
Pearson’s correlation analysis between domain scores and the general health dimension of WHOQOL-HIV-BREF was performed for convergent validity. To assess concurrent validity, we examine the association between domain scores and the criterion variables measured. We expected to find positive correlations between domain scores and CEAT-VIH and GHQ-12. We also expected to find negative correlations between domain scores and HIV-related stigma dimensions.
Known-group validity was used to assess the capacity of WHOQOL-HIV-BREF to discriminate among subgroups of participants according to their immunological (CD4 count) and virological (viral load copies) status. It was expected that participants with higher CD4 count and undetectable viral load would have higher HRQoL domain scores.
Finally, differences according to the socio-demographic and epidemiological profiles of study participants were analyzed. For the sake of simplicity and clarity, only differences in the domains were assessed in most of the characteristics of the participants, and only p-values are shown. However, differences in all HRQoL facets were tested according to age and sex. It was done because of the current relevance of analyzing HRQoL in aging PLHIV and the UNAIDS recommendations about disaggregation of data according to relevant socio-demographic characteristics [27]. To test differences by age, a cut-off point of 49 years (≤ 49 vs. ≥ 50 years) was established. T-test and one-way variance analysis were used for these analyses.
Regarding the data analysis software, LISREL (LInear Structural RELations) 8.7 program and its companion preprocessor program PRELIS for Windows were used for the CFAs. LISREL is an application for structural equation modeling developed by K. G. Jöreskog and D. Sörborm [28]. PRELIS is an application for data manipulation, data transformation, data generation, computing moment matrices and imputation by matching. A widely used program for statistical analysis in social science, IBM SPSS Statistics 22 [29], was selected for the remaining analyses.