Comprehensive Knowledge of HIV among Women in Rural Mozambique: Development and Validation of the HIV Knowledge 27 Scale

Background: The relationship between HIV knowledge and HIV-related behaviors in settings like Mozambique has been limited by a lack of rigorously validated measures. Methods: A convenience sample of women seeking prenatal care at two clinics were administered an adapted, orally-administered, 27 item HIV-knowledge scale, the HK-27. Validation analyses were stratified by survey language (Portuguese and Echuabo). Kuder-Richardson (KR-20) coefficients estimated internal reliability. Construct validity was assessed with bivariate associations between HK-27 scores (% correct) and selected participant characteristics. The association between knowledge, self-reported HIV testing, and HIV infection were evaluated with multivariable logistic regression. Results: Participants (N = 348) had a median age of 24; 188 spoke Portuguese, and 160 spoke Echuabo. Mean HK-27 scores were higher for Portuguese-speaking participants than Echuabo-speaking participants (68% correct vs. 42%, p<0.001). Internal reliability was strong (KR-20>0.8) for scales in both languages. Higher HK-27 scores were significantly (p≤0.05) correlated with more education, more media items in the home, a history of HIV testing, and participant work outside of the home for women of both languages. HK-27 scores were independently associated with completion of HIV testing in multivariable analysis (per 1% correct: aOR:1.02, 95%CI:0.01-0.03, p = 0.01), but not with HIV infection. Conclusions: HK-27 is a reliable and valid measure of HIV knowledge among Portuguese and Echuabo-speaking Mozambican women. The HK-27 demonstrated significant knowledge deficits among women in the study, and higher scores were associated with higher HIV testing probability. Future studies should evaluate the role of the HK-27 in longitudinal studies and in other populations. © 2012 Ciampa et al. The study protocol was reviewed and approved by the National Committee of Bioethics for Health in Mozambique (Comité Nacional Bioética Para a Saúde) and the Institutional Review Board of Vanderbilt University. No financial incentive was provided to participants, and written informed consent was obtained for all study participants that included permission to review medical records relating to HIV testing. Zambézia Province is located in central Mozambique, is predominantly rural, and had an estimated adult HIV prevalence of 12.6% in 2009 [32]. During pregnancy, 58% of women in Zambézia seek prenatal care at least once [36]. HIV counseling, screening, and information regarding mother-to-child transmission are routinely offered to all women seeking prenatal care, and those who are HIV infected are offered antiretroviral prophylaxis or combination antiretroviral therapy (cART) when eligible by national guidelines [37]. The HK-27 was developed by adapting items from three existing HIV knowledge scales: the Behavioral Surveillance Survey (BSS) from FHI 360, [35] the Demographic Health Survey-AIDS (DHSAIDS), [24] and the 45-item version of the HIV-Knowledge Questionnaire (HIV-KQ-45) [30]. A brief version of the HIV-KQ-45 (the HIV-KQ-18) has been developed, [38] but was not used for this study. Both the BSS (13 items) and DHSAIDS (9 items) assess knowledge of HIV transmission risk factors, are orally administered, and are nested within the framework of a larger questionnaire designed to measure HIV-related stigma and sexual behavior. Items measuring HIV knowledge from the HIV-KQ-45, BSS and DHSAIDS were pooled and reviewed by experts in HIV care for those living in resource-limited settings, cultural norms and beliefs surrounding HIV in Mozambique, health literacy, and survey design. Items were excluded if they were judged to be redundant in content, less culturally relevant for the Mozambican population, or poorly constructed. The resulting set of items assessed HIV knowledge across three content domains: general disease knowledge, sexual transmission risk factors, and non-sexual transmission risk factors. Four original items were created to add a fourth content domain, assessing knowledge of HIV treatment. A total of 30 items were included after preliminary review; these were translated into Portuguese by a fluent speaker and back-translated into English to verify the accuracy of the translation. The items were also translated by a trained bilingual translator from Portuguese into a local Mozambican language (Echuabo), a principal local language for southern Zambézia Province [39]. The two sets of translated survey items were assessed for clarity and cultural relevance with cognitive interviews of 32 women (19 Portuguese-speakers and 13 Echuabo-speakers) who were awaiting prenatal care at two Zambézia health centers, in Quelimane and Inhassunge. Interviews were conducted in Portuguese or Echuabo, and took place in the language of choice for each participant. A trained interpreter fluent in Portuguese and Echuabo was used to conduct cognitive interviews of Echaubo-speaking participants. These interviews assessed participants’ understanding of each item, ensured translations were accurate, and verified all items included in the scale were culturally appropriate. As a result of the cognitive interviews, three items were discarded because of poor clarity, and the translation of 12 items underwent slight revision to better match local language usage, using the suggestions of interview participants. The final scale included 27 items; each item consisted of a statement for which a participant could respond “agree”, “disagree” or “uncertain”. As with the HIV-KQ-45, the response choice of “uncertain” was included to discourage guessing [30]. A convenience sample of participants was recruited during prenatal care at two health centers. Potential participants were approached after completing a prenatal care consultation. One site was urban (Quelimane) and served patients who commonly spoke Portuguese, a second site (Inhassunge) was rural and served patients who commonly spoke Echuabo; an effort was made to recruit participants proficient in each language at each site. Women were included if they were 1) pregnant, 2) ≥18 years of age, and 3) spoke either Portuguese or Echuabo as their primary language. After consent was obtained, women were interviewed in a private area of the hospital. All study measures were administered orally in the language of choice of the participant (Echuabo or Portuguese); a trained interpreter fluent in Portuguese and Echuabo was used to facilitate administration of the HK-27 to Echaubo-speaking participants. Sociodemographic characteristics, including age, number of children, education, and occupation, were ascertained by participant self-report. Each participant was administered the HK-27, after which any participant questions about content were addressed. Each HK-27 item was scored as correct or incorrect, with responses of uncertain counted as incorrect. HIV knowledge was defined as the percentage of HK-27 items answered correctly for each participant. Media item ownership was ascertained by self-reported ownership of television, internet, or cell phone and ranged from 0–3 items. HIV testing was ascertained by self-report to account for tests done at other clinical sites where medical records were inaccessible. HIV test results were obtained on review of the medical record at the study sites only. All data was recorded on paper and entered into a secure electronic database, REDCap™ [40]. Sociodemographic data were reported as medians with interquartile ranges (IQR) or as proportions where statistically appropriate for the total sample, and stratified by survey language. Mean HK-27 scores were tabulated along with standard deviations (SD) for the total sample, and separately by language. The proportions of women who obtained an HIV test and were HIV-infected were also tabulated. To ensure construct validity of the HK-27 in both languages, psychometric testing and validation of the Echuabo and Portuguese language scales were done separately. To establish construct validity of the HK-27, we hypothesized that higher HIV knowledge would be associated with characteristics that may indicate greater exposure to HIV-related information, such as older age, more children (since pregnant women receive HIV counseling as part of prenatal care), receipt of HIV testing, and ownership of more media items. Other factors that may indicate a higher capacity for understanding HIV-related information (more education, work outside of the home, HIV-negative status) were also included in the assessment of construct validity. We hypothesized that Portuguese-speaking participants would have higher scores than participants who spoke Echuabo. Kuder-Richardson 20 coefficients were generated to test internal reliability of the HK-27. Spearman correlations were calculated to assess the association between HK-27 scores and continuous, nonparametric variables; Wilcoxon rank-sum tests compared HK-27 scores across binomial variables, and Kruskal-Wallis tests compared HK-27 scores for categorical variables. Multivariable logistic regression models examined the relationship between HIV knowledge and self-reported HIV testing. A second set of logistic regression models tested the association between HIV knowledge and HIV-infection for those with medical records available to review. Multivariable models adjusted for study language and site, maternal work, and travel time to the hospital. Covariates in the multivariable model were chosen a priori based on the belief that these factors were related to health care access. All analyses were conducted using STATA™ statistical software package (STATAcorp, Release 11, College Station, TX).