Background
It is widely recognized that attrition from the prevention of mother-to-child HIV transmission (PMTCT) cascade is a significant obstacle to achieving UNAIDS’ and the World Health Organization (WHO)’s goal to eliminate mother-to-child transmission by 2015 [
1-
4]. The PMTCT cascade is a series of services that HIV-positive pregnant women and their infants need to receive in order to prevent HIV transmission, including antenatal care (ANC), HIV testing, and antiretroviral therapy (ART) or antiretroviral (ARV) prophylaxis [
5]. However, 49% of HIV-infected pregnant women in sub-Saharan Africa are lost between ANC registration and delivery and miss some or all essential PMTCT services [
6]. Furthermore, high rates of loss to follow-up among women initiating ART under ‘Option B+’, [
7] WHO’s PMTCT strategy whereby all pregnant and breastfeeding women receive lifelong ART, [
8] has renewed emphasis on the importance of reducing barriers to uptake of PMTCT services.
In many settings, economic factors are cited as barriers to ANC and PMTCT services, [
9,
10] including facility-based delivery, [
11] initiation of ART and ARV prophylaxis, and retention in HIV care [
12]. Here we focus on one aspect of socioeconomic position: food security. People are considered food secure when they have adequate physical, social, and economic access to sufficient, safe and nutritious food that meets their dietary needs and food preferences for an active and healthy life [
13]. Food insecurity is increasingly recognized as exacerbating the HIV/AIDS epidemic by increasing engagement in HIV-related risk behaviors, [
14-
17] and among people living with HIV, undermining ART adherence and retention in care [
18-
20]. However, few studies have examined the relationship between food insecurity and utilization of PMTCT services and MTCT [
19,
20].
There are several pathways through which food insecurity might affect women’s use of health services and increase vertical HIV transmission [
21]. First, food insecurity might result in avoidance or delay of maternal health services because of its overlap with socioeconomic position [
22,
23] and the real or perceived costs of ANC, facility delivery, and/or HIV prevention and care services. Second, food insecurity is associated with undernutrition, which among HIV-infected women is associated with preterm delivery, low birth weight, and MTCT [
24-
26]. In addition, food insecure HIV-infected women may be less likely to adhere to ART/ARV prophylaxis, [
27] and they may exclusively breastfeed their infants for shorter periods of time, heightening the risk of onward transmission as they resort to mixed feeding [
28,
29]. Lastly, food insecurity may influence women’s receipt of PMTCT services through mental health pathways related to the anxiety and stress associated with real or perceived hunger [
30,
31]. Stress and depression, in turn, may affect service utilization and obstetric and child health outcomes [
32,
33]. Together, these compelling pathways support the hypothesis that food insecurity may undermine global efforts to achieve elimination of mother-to-child HIV transmission.
We examined the relationship between food security and PMTCT in Zimbabwe, where a generalized HIV epidemic coexists with food insecurity, hunger, and undernutrition. Thirty-three percent of Zimbabwe’s population is undernourished, [
34] and the 2013 Global Hunger Index was 16.5, indicating serious levels of hunger [
35]. Moreover, 12% of pregnant women are HIV-positive [
36]. The objectives of our analyses were to: 1) determine the prevalence of food insecurity among women with a recent birth; 2) explore whether food insecurity is associated with receipt of PMTCT-related services; and 3) examine the association between food insecurity and MTCT.
Discussion
In this analysis of women with a recent birth in Zimbabwe, we found that more than half reported living in moderate or severely food insecure households in the month prior to the survey. Compared to women from food secure households, women from food insecure households were more likely to be HIV-infected. Consistent with previous qualitative studies, [
49] we found that food insecurity may be an important barrier to uptake of some PMTCT services: in unadjusted analyses, food insecurity was inversely associated with ANC, knowing one’s HIV status, facility-based delivery, and postnatal visit attendance. When services were examined together, and after adjustment for covariates, women who reported severe food insecurity were 14% less likely to complete all recommended maternal and infant health services for PMTCT compared to food secure women. Although the effect sizes are modest and absolute differences are small, these findings suggest that among a subgroup of pregnant women, severe food insecurity is an important barrier to some maternal health services.
Among HIV-infected women, we unexpectedly found that there was no association between food insecurity and completion of the PMTCT cascade. This might be due to the small sample size, women’s motivation to protect their infant from HIV infection, or nutritional support provided during pregnancy and postpartum that partially mitigates household food insecurity. However, women from severely food insecure households were 42% more likely to have an HIV-infected infant compared to women from food secure households, although this finding was not statistically significant after adjustment for covariates. Although food security’s association with vertical transmission has been speculated, [
21,
50] this analysis is the first, to our knowledge, to provide empirical data supporting this hypothesis. We must nevertheless interpret these findings with caution because a strong ‘dose-response’ relationship between food insecurity and MTCT was not observed, as women in moderately food insecure households had the lowest proportions of exclusive breastfeeding and MTCT.
Although we found that food insecurity was associated with ever attending ANC, we found no association between food insecurity and attending the WHO-recommended ≥4 ANC visits or the timing of ANC registration. There are several potential explanations for this finding. In facilities that charge a fee, the fees often cover the bundle of services from ANC booking through the 6-week postnatal visit. Thus, once engaged in ANC, women may be retained in the cascade and continue to receive ANC care and deliver at the health facility. Women who don’t attend ANC may therefore also be more likely to deliver at home and miss the postnatal visit. Although this study found only small but statistically significant reductions in utilization of each service among women who reported being moderately or severely food insecure, losses from the cascade are additive, as was demonstrated by a cohort study in Cameroon, Côte d’Ivoire, South Africa, and Zambia, where unremarkable levels of attrition of HIV-infected mothers from individual steps in the cascade coupled with poor adherence resulted in 49% of HIV-exposed infants not being protected by a prophylactic ARV regimen [
51].
An unanswered question is the pathway(s) through which food insecurity might impede service utilization. It is possible that food insecurity in this analysis is simply a proxy for poverty, and our findings are reflective of the difficult choices food insecure women must make between food (and other goods and services) and the costs associated with health care, including transport and fees. Certainly, food insecurity is highly correlated with socioeconomic position [
22,
23] and has been shown to be the strongest measure of socio-economic position associated with HIV and HSV-2 risk among young women in Zimbabwe, potentially due to engagement in risk behavior to obtain food or other essential goods and services [
52]. This might be the most likely explanation for the inverse association we identified between food insecurity and ANC and other services that require or are perceived to require payment. Although Zimbabwe is moving toward elimination of user fees for maternal and child health services, some facilities still charge a fee that may be cost-prohibitive.
However, this economic explanation does not fully explain the finding that severe food insecurity was not associated with receipt of maternal and infant ART/ARVs, but may nevertheless be associated with MTCT, although the width of the confidence limits suggests substantial uncertainty in this association. One explanation might be that although both food secure and insecure women and infants were equally likely to receive ART/ARV prophylaxis, women who were food insecure were less likely to adhere to treatment. Food insecurity is known to reduce ART adherence in non-pregnant populations, [
27] although this study is not able to test this hypothesis. Another possible explanation is more complex: severe food insecurity increases MTCT risk due to cumulative loss of women from the cascade coupled with an increased risk of MTCT associated with undernutrition [
25,
26] and the possible increased propensity for mixed feeding [
28]. This study was unable to explore more complex patterns of breastfeeding (such as duration of exclusive breastfeeding) to support or refute this hypothesis, and our simple measure of exclusive breastfeeding (ever) found that nearly all women exclusively breastfed at some point. Furthermore, this secondary data analysis used cross-sectional data from a study that was not designed to specifically examine this question, so we do not have prospective data nor information on several causal intermediates, including nutritional indicators such as women’s body mass index (BMI) and micronutrient status, to examine these pathways. Nonetheless, these data provide both compelling empirical evidence about these hypothesized relationships and also reveal important gaps for future research.
An important issue when considering these findings is the measurement of food security [
53,
54]. In this study, household food security was measured in the 4 weeks prior to the survey, a common reference period used by other scales [
30,
55,
56] in order to balance the tradeoffs between recall bias (which favors a shorter recall period) and ‘telescoping errors’ (a phenomenon associated with short recall periods whereby events outside of the recall window are erroneously reported) [
57,
58]. We made the essential temporality assumption that a household’s recent food security status was highly correlated with its food security status during the woman’s pregnancy, 9–18 months prior. The validity of this assumption is unknown; it depends on how much household food access changed over time and season, which was not measured in the survey. Nevertheless, this issue could be overcome in future prospective studies by conducting a simple food security assessment at ANC registration to identify food insecure women who are at risk of both undernutrition and of missing key PMTCT services.
Another key measurement issue is that food security was measured using only part of the HFIAS, a validated scale, which adds some uncertainty to the classification of food security. Nevertheless, our inclusion of a question from each dimension of food access [
30] in addition to the correlation between our parameterization of food security and other dimensions of household socio-economic position increases confidence in our classification scheme. A strength of our measurement of food security is that women themselves reported household food security status, the individuals who are typically responsible for a household’s food supply and meal preparation in Zimbabwe. This is also important because food insecurity at the individual level may be prevalent even in wealthier households due to unequal intra-household allocation of food, which, for example, can result in women eating last or having less access to fats, protein, or micronutrient-rich foods [
59-
61]. Nevertheless, as with all measures of food insecurity, food security may be subject to underreporting because of its sensitive nature [
62].
Our analysis has other important limitations. We used cross-sectional data and therefore cannot make inferences about causation. Further, women self-reported receipt of healthcare services. Moreover, although our data are representative of the communities from which the sample was selected, they are not representative of all regions in Zimbabwe, and it is possible that the relationship between food insecurity and service utilization are different in other parts of the country. In addition, although our strategy to create a sampling frame of 9–18 month old infants in the community was comprehensive, it is possible that some mother-infant pairs were missed. Lastly, women’s and infant’s HIV status was measured at the time of the survey, 9–18 months postpartum. Although we have assumed that women who were HIV-infected at the time of the survey were also HIV-infected during their pregnancy, it is possible that a small proportion were infected during pregnancy or postpartum. Likewise, infants who were still breastfeeding at the time of the survey remained at risk of MTCT, so we may not have captured all possible infant infections.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
SM RB, NP, and FC collaboratively designed the impact evaluation which was the source of data for this analysis. SM and RB conducted the data analysis, which was iteratively refined after discussion with all authors. SM drafted the initial manuscript and all authors participated in reviewing the draft for intellectual content and assisting with revisions. All authors approved the final version of the manuscript.