Menu
Menu
Menu
Menu
Background: Previous research has shown that developing countries account for the majority of maternal deaths around the world. Relatively high maternal mortality in developing countries has been linked to high HIV prevalence rates in these countries. Several studies have shown that women living with HIV are more vulnerable and are thus more likely to die during maternity than those who are not. Although there has been increased focus on this subject in contemporary research, the relationship between HIV status and maternal-care-utilization is not very well understood. It is not clear whether factors associated with professional maternal care utilization during antenatal, delivery and postnatal periods are similar for HIV positive and HIV negative women. It is also not known whether being HIV positive has an impact on the choice of care (professional care or traditional birth attendants). Thus the aim of this study is to investigate the differences in factors affecting choice of care during antenatal, delivery and postnatal periods between HIV positive and HIV negative women. We also investigate the effect of HIV positive status on choice of care. Methods: By using the 2013-2014 Zambia Demographic Health Survey Data (ZDHS), we performed two different quantitative analyses. a) Regression analysis: to identify and compare factors associated with the likelihood of utilizing professional care during antenatal, at birth and postnatal periods between HIV positive and HIV negative women. b) Propensity score matching: to investigate the effect of being HIV positive on the choice of care (Professional care or TBAs). Results: Our results show that reasons for choosing professional care during antenatal, at birth, and postnatal periods are the same for both HIV positive and HIV negative women. Further, we also showed that although the probability of utilizing professional care is slightly higher for HIV positive women, the difference is negligible. Conclusion: We demonstrated that in Zambia, utilization of professional care among HIV positive women is not particularly high. We also demonstrate that although institutional care is desirable and an ideal solution for HIV positive women, insisting on institutional care when the health facilities lack adequate trained personnel, drugs, and equipment is counterproductive.
This study made use of the publicly available and nationally representative survey known as the 2013–2014 Zambia Demographic Health Survey (ZDHS). This is a survey that was conducted by a consortium of institutions including the Zambian ministry of health, the University of Zambia and the Virology Laboratory at the University Teaching Hospital of Zambia ([2], p. 14). The ZDHS contains social demographic data obtained from women aged 15–49 years. The sample in the ZDHS 2013–2014 is a two-stage stratified cluster sample design. In the first stage, 722 Enumeration Areas (EAs) (305 in urban areas and 417 in rural areas) were selected from all the 10 provinces of Zambia. Stratification was done by dividing each of the 10 provinces into urban and rural areas resulting in 20 sampling strata. In the second stage, a complete list of households was used as a sampling frame upon which 25 households were selected for enumeration in each EA [14]. It was at this stage that a representative sample of 18,052 households was finally selected. Regarding HIV results; the HIV data were obtained by collecting blood samples during interviews from consenting participants (83.6% of the total sample) and later testing the blood samples for HIV and then linking the sample results to the unique IDs of participants. Specifically, after fulfilling the country’s ethical requirements, the DHS team collected blood spots on filter paper from a finger prick and transported them to a laboratory for testing. If the respondent did not consent to testing, their HIV status was left unstated in the data and treated as missing ([8], p. 14). The ZDHS does not make the HIV data publicly available, as such, these data were only availed to us after obtaining ethical clearance from the Ethics board of Zambia and the ministry of health in Zambia. By using unique IDs, we then linked the HIV data to the rest of the dataset. The next step was to select a sub-sample of women who had given birth at least once in the last 5 years preceding the survey and had a determined HIV status. We did not include women who did not take part in the HIV test or had undetermined HIV status. Our final sample thus consisted of 12, 225 women. These data enabled us to investigate the difference in maternal health care utilization (skilled delivery attendants or trained TBAs) between HIV-positive and HIV-negative women in Zambia. A summary of the characteristics of the data is provided in Table 1. A summary of characteristics (differences and similarities) between HIV positive and HIV negative populations. (Which we summarized in our previous study ***p < 0.01, **p < 0.05, *p < 0.10. Reported p-values are based on ttests of means for continuous variables and chi-squares for proportions/categorical variables For the selection of explanatory variables, we relied on previous studies on maternal health utilization [2, 9, 10], and also made use of the Andersen’s behavioral model of health which holds that usage of any given health service is based on three dynamics: predisposing factors (such as socio-demographic factors), enabling factors (such as wealth, access to health insurance) and healthcare needs (chronic illnesses, functional disability etc.) [15]. We therefore included the following predisposing factors in our study: age, religion, area of residence, level of education, distance to health facility, attitude of health workers and availability of drugs in facilities. The enabling factors were wealth and health insurance. While having given birth at least once in the last 5 years is seen as a proxy for the ‘need’ factor. The ZDHS included a section on maternal health service utilization for the most recent birth. In this section, questions were asked on who attended to the women during antenatal, birth and postnatal periods. The options in general included: Health professional and TBAs. Therefore, we separately use this dummy variable (health professional or TBA) for all the three stages of maternal health utilization (antenatal, birth, and postnatal stage). We undertook two different analyses: Probit analysis: to identify the factors that influenced the likelihood of utilizing professional care during antenatal, at birth and postnatal (for both HIV positive and HIV negative women). On the basis of this probit analysis, we further generated marginal effects to ascertain the probabilities of utilization of professional care during antenatal, at birth and postnatal. Propensity Score Matching: to investigate the effect of being HIV positive on choice of maternal health services between professional care and TBA during antenatal, at birth and postnatal. For this, we calculate the Average Treatment Effect on the Treated (ATT). More specifically, Propensity Score Matching (PSM) is used to allow us to statistically formulate a control group (HIV negative women) by matching the observed characteristics of the treated participants (HIV positive women) to the control group. This is based on similar values of the propensity score [16]. Heckman et al. (1998) define PSM as the probability of selection into the treated group, which in this case means the probability of being HIV positive. The variables that we use for matching include age, educational level, area of residence, and wealth. The selection of variables was guided by theory and consensus in extant literature regarding what factors are likely to increases chances of being HIV positive [17–19]. When conducting PSM, it is important to note that the “unbiased inference” in PSM is based on the assumption that outcomes are independent of assignment to the treatment group on the basis of observable characteristics [20]. In order to have valid results, it is important that between the propensity scores of treatment and control groups, there exists an area of “common-support” [21]. We used Stata to estimate the ATT using the Nearest Neighbor Matching technique [22]. This is a technique that matches individuals from control and treatment group with similar propensity scores and then drops all those that are not selected in the match [16]. The reliability of the estimated effect of the HIV positive status on choice of maternal care when using PSM depends on selection of observables [20]. For this purpose, we inspected the propensity score distribution to confirm the existence of common support area and we also checked for the balancing property using psmatch2 [23]. Regarding the robustness of the ATT, we compared our findings resulting from the Nearest Neighbor matching technique with two other matching techniques namely Kernel matching and Stratification matching [24]. Kernel matching makes use of more information which allows it to lower the variance [20]. Stratification matching on the other hand partitions the common support into different strata and then computes the impact of treatment within each of those strata [24]. What goes on specifically within the strata is that the effect of the HIV positive status is established as the mean difference in outcomes (use of skilled birth attendant or TBA) between the control and the treated individuals. An average of the interval impacts which is weighted gives the overall HIV status impact by taking the share of the individuals in each interval as weights [25]. PSM has a weakness of failing to correct for biases resulting from unobservables. Thus to deal with this challenge we use the Altonji et al. test. This test allows us to estimate the sensitivity of selection on the basis of unobserved characteristics [20].
N/A
