Data source
The study is based on a secondary analysis using nationally representative data from the Pakistan Demographic and Health Survey (PDHS) 2012–13 [39]. The PDHS 2012–13 is the third survey conducted as part of the MEASURE DHS (Demographic and Health Survey) international series with the financial support of USAID. The National Institutes of Population Studies (NIPS) completed the PDHS with technical support from ICF International and the Pakistan Bureau of Statistics. The PDHS 2012–13 is the largest publicly available household dataset in Pakistan to have collected information on variables related to HIV/AIDS awareness and attitudes among the general population. The cross-sectional study used a two-stage cluster sampling technique for data collection. During the first stage, sampling areas of 248 urban and 252 rural units were selected; in the second stage, 14,000 households (6944 from urban areas and 7056 from rural areas) were selected through systematic random sampling. The fieldwork was completed between October 2012 and March 2013, during which time a total of 20 field teams, each comprising a supervisor, a field editor, one male and three female interviewers, collected data. These teams were supervised by quality-control interviewers, field coordinators and senior NIPS team members. Along with the fieldwork, the data processing, including editing and entry of completed questionnaires, was initiated simultaneously. Moreover, all data was entered twice using the CSPro computer package within the NIPS office for 100% verification [39].
The PDHS 2012–13 used four types of questionnaires for data collection, consisting of: a household questionnaire, a women’s questionnaire, a men’s questionnaire, and a community questionnaire. The standard women’s questionnaire, used in this analysis, was administered to 13,558 ever-married women of reproductive age (15–49 years) through face-to-face interviews, with a response rate of 93% [28]. A series of questions related to overall knowledge and attitudes around HIV/AIDS were also part of the standard women’s questionnaire to assess respondents’ knowledge regarding modes of HIV/AIDS transmission and the ways in which HIV/AIDS can be prevented, as well as their attitudes towards PLWHAs [39].
Variables
Outcome variables
Women’s overall knowledge about HIV/AIDS and their attitudes towards PLWHAs are the outcome variables for this research. Both variables were inferred from a series of questions used in the PDHS 2012–13 questionnaire [39], consistent with previous similar studies [16, 40] and also aligned with the global AIDS monitoring indicators of 2018 [27]. All the women who had ever heard of HIV/AIDS further responded to questions on their knowledge and attitudes towards PLWHAs. The construction of both variables is published elsewhere [40].
Women’s overall knowledge on HIV/AIDS was inferred from five questions, including knowing about the two most common methods to prevent HIV/AIDS infection: 1) consistent condom use and 2) limiting the number of sexual partners to one uninfected partner who is faithful. Additionally, it was assessed whether the respondents were able to reject three common misconceptions about HIV/AIDS: 3) a person can get HIV from a mosquito bite, 4) a person can get HIV by sharing a meal with an infected individual, and 5) a healthy-looking person can have HIV/AIDS. The answer categories for the above questions were “yes”, “no”, and “don’t know”. Hence, for this research, the incorrect and “don’t know” responses were re-coded as 0, while correct responses were re-coded as 1. Thus, the score for women’ overall knowledge on HIV/AIDS ranged from 0 to 5; where a woman who answered all five questions incorrectly had a score of 0 and a woman who answered any of the five questions correctly got a score between 1 and 5. The mean value was taken as a cut-off value for dichotomisation of women’ overall knowledge on HIV/AIDS scores into high vs. low knowledge. According to this, women who had a score of 3–5 were coded as having high HIV/AIDS knowledge, while the women who had a score of 0–2 were coded as having low knowledge.
Similarly, women’s attitudes towards people living with HIV/AIDS was measured through the following four questions: 1) “Would you want to keep a family member’s HIV infection a secret?”, 2) “Would you care for a relative who is infected with HIV?”, 3) “Would you buy vegetables from a vendor who has AIDS?”, and 4) “Should a female teacher infected with HIV be allowed to continue teaching in school?” The answer categories for these questions were “yes”, “no” and “don’t know”. Here, the “no” and “don’t know” responses were also re-coded as 0, having a negative attitude, and “yes” responses were re-coded as 1, having a positive attitude towards PLWHAs. Thus, the score for each woman’s attitude towards PLWHAs was computed for this analysis, ranging from 0 to 4, where a total score of 0 represented a negative attitude and a score of 1–4 indicated a positive attitude in any of the four scenarios. Again, the mean value was taken as a cut-off value for dichotomisation of scores into positive vs. negative attitudes. Thus, the women, who acquired 3–4 score, were coded as having a positive attitude, while the women, who had a score of 0–2 were considered as having a negative attitude towards PLWHAs.
Autonomy variables
Women’s autonomy is measured through two variables: a) their overall participation in multiple household decisions, and b) emotional autonomy, referring to their attitudes towards domestic violence, highlighting their opinions (agreement or disagreement) on wife beating. The construction of both these variables is published elsewhere [40, 41].
Women’s autonomy in terms of participation was inferred from the five questions: 1) “Who (in your family) usually decides how to spend your earnings?”, 2) “Who usually decides on making large household purchases?”, 3) “Who usually decides on your visits to family or relatives?”, 4) “Who usually decides on your healthcare?”, and 5) “Who usually decides what to do with your husband/partner’s earnings?” Possible responses to the first three autonomy questions were: “respondent alone”, “husband/partner alone”, “respondent and husband/partner jointly”, “respondent and other person”, “someone else”, “family elders” or “others”. Possible responses to who usually decides what to do with the husband/partner’s earnings were “respondent alone”, “husband/partner alone”, “respondent and husband/partner jointly”, “respondent and other person”, “someone else”, “family elders”, “the husband/partner does not bring in any money” or “other”. For this study, responses to the above autonomy questions were dichotomised into two categories: whether the woman has “a say at all” (either alone or jointly with her husband/partner or jointly with another person) coded as 1, or whether she has “no say at all” coded as 0 (in cases where the husband/partner, family elders or someone else makes the decision). This dichotomisation of autonomy/decision-making is consistent with previous research work done using the DHS datasets [40, 41]. Based upon these five binary household decision-making questions, the score for women’ autonomy was computed for each woman ranging from 0 to 5, where if she had no say in any of the five decisions, her total score was 0 and if she had a say in any of the five household decisions her total score ranged between 1 and 5. Further, the mean value was taken as a cut-off value for the dichotomisation of autonomy scores into high vs. low autonomy. Women who had scores of 2–5 were coded as having high autonomy, while women with scores of 0 or 1 were considered to have low autonomy.
Women’s emotional autonomy was assessed through their attitudes towards domestic violence (wife beating). PDHS asked the women about situations when sometimes a husband is annoyed or angered by things that his wife does. In the respondent’s opinion, is a husband justified in hitting or beating his wife in the following situations: 1) “If she goes out without telling her husband?”, 2) “If she neglects the children?”, 3) “If she argues with her husband?”, 4) “If she refuses to have sex with her husband?”, and 5) “If she burns the food?”. Response categories for the wife-beating questions were “yes”, “no”, or “don’t know”. For the purposes of this study, the “yes” and “don’t know” response categories were re-coded as 0, while “no” responses were re-coded as 1. Based upon the five questions above, the scores for women’s emotional autonomy (attitude towards wife beating) was computed as ranging from 0 to 5, where a woman with a score of 0 agreed with all five circumstances of wife beating, whereas a woman having score of 1–5 disagreed with wife beating under one or more of the five circumstances. Furthermore, the mean value (i.e. 3) was taken as a cut-off value for dichotomisation of emotional autonomy or attitudes towards wife beating into disagreement vs. agreement. Women who had a score of 3–5 were coded as having high emotional autonomy, disagreeing about wife beating situations, while women with a score of 0–2 were coded as having low emotional autonomy, agreeing with the wife beating circumstances. For the sake of this analysis, women’s agreement or disagreement on wife beating will likewise represent emotional autonomy.
Socio-demographic variables
Based on the existing literature and available data within the PDHS 2012–13, a number of socio-demographic variables were included in the analysis [39, 40]. These were: respondents’ region/province (Punjab, Sindh, Baluchistan, Khyber Pakhtunkhwa, Gilgit Baltistan, Islamabad), the geographical classification of their residence (urban/rural), respondents’ age (15–24 years, 25–34 years, 35 years and above), educational level of the women and their husbands (each was grouped into four categories: uneducated/no formal schooling, primary, secondary, higher education), occupation of respondents and their husbands (each was grouped into four categories: unemployed; working in professional/managerial positions, including sales & services; agriculture; and unskilled or manual/household workers) and respondents’ exposure to mass media, including newspapers, TV and radio to access information (yes/no). Moreover, a composite index of household wealth was grouped into five quintiles (richest, richer, middle, poorer, poorest), measured on the basis of household assets and ownership of a number of consumer items [42].
Statistical analysis
IBM SPSS® version 21 was used for data analysis. Sampling weights were used. Descriptive statistics for the variables of socio-demographics, autonomy and comprehensive knowledge and attitudes towards PLWHAs were compiled, and frequency distributions and percentages were presented. Cross-tabulations and chi-square tests were performed to assess the significance. A significance level of p < 0.05 was chosen. Simple binary logistic regression was used to determine the association of the predictors with women’s overall knowledge of HIV/AIDS and their attitudes towards PLWHAs. Afterwards, a multiple logistics regression was conducted using only those variables that were found to be significantly associated with both outcome variables. Further regression models are presented in the Additional file 1.