Diversity of influences on infant feeding strategies in women living with HIV in Cape Town, South Africa: a mixed methods study

Authors

  • Rose Zulliger,

    Corresponding author
    1. Department of Health, Behavior and Society, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA
    • Corresponding Author Rose Zulliger, Department of Health, Behavior and Society, Johns Hopkins Bloomberg School of Public Health, 624 N. Broadway, Baltimore, MD 21224, USA. Tel.: +1 614 937 0128; Fax +1 410 502 4333; E-mail: rozullig@jhsph.edu

    Search for more papers by this author
  • Elaine J. Abrams,

    1. ICAP, Mailman School of Public Health and College of Physicians & Surgeons, Columbia University, New York, NY, USA
    Search for more papers by this author
  • Landon Myer

    1. Division of Epidemiology and Biostatistics, School of Public Health & Family Medicine, University of Cape Town, Cape Town, South Africa
    Search for more papers by this author

Abstract

Objective

To explore influences on infant feeding intentions and practices in women living with HIV in South Africa.

Methods

Structured questionnaires were completed by 207 pregnant women and 203 post-partum women in Cape Town, South Africa. Concurrently, 34 semi-structured, qualitative interviews explored the influences on infant feeding strategies in women living with HIV.

Results

Overall, 50% (104) of pregnant women intended to breastfeed and 22% (45) of post-partum women ever breastfed. Women who breastfed or intended to breastfeed were significantly more likely to have running water in their homes, to have formal housing and to receive advice in support of breastfeeding. Advice from clinic staff was the strongest predictor of breastfeeding [adjusted relative odds (ARO) in pregnant women: 6.87; 95% confidence interval (CI): 2.67, 17.66; ARO in post-partum women: 4.04; 95% CI: 1.60, 10.19]. Other important influences included previous infant feeding experiences, desires to protect the infant from HIV and involvement of other care providers. Many women also noted that breastfeeding was not feasible due to work commitments and highlighted concerns around the discontinuation of the free provision of infant formula.

Conclusion

These results suggest that women living with HIV balance complex influences in deciding on their preferred infant feeding strategies. This underscores the need for extensive provider, patient and community education to ensure consistent messaging, while allowing for adaptation to the circumstances of individual mothers.

Abstract

Objectif

Explorer les diverses influences sur les intentions et les pratiques d'alimentation du nourrisson chez les femmes vivant avec le VIH en Afrique du Sud.

Méthodes

Des questionnaires structurés ont été complétés par 207 femmes enceintes et 203 femmes en post-partum à Cape Town, en Afrique du Sud. Parallèlement, 34 entretiens qualitatifs semi-structurés ont exploré les influences sur les stratégies d'alimentation du nourrisson chez les femmes vivant avec le VIH.

Résultats

Dans l'ensemble, 50% (104) des femmes enceintes avaient l'intention d'allaiter et 22% (45) des femmes en post-partum avaient allaité au sein. Les femmes qui ont allaité ou avaient l'intention d'allaiter étaient significativement plus susceptibles d'avoir de l'eau courante chez elles, d'avoir une habitation officielle et de recevoir des conseils de soutien à l'allaitement maternel. Les conseils du personnel clinique était le meilleur prédicteur de l'allaitement maternel (risque relatif ajusté [ARO] chez les femmes enceintes: 6.87, intervalle de confiance [IC] à 95%: 2.67–17.66; ARO chez les femmes en post-partum: 4.04; IC 95%: 1.60–10.19). D'autres influences importantes comprenaient les expériences antérieures dans l'alimentation du nourrisson, le désir de protéger l'enfant contre le VIH et l'implication d'autres prestataires de soins. Beaucoup de femmes ont également noté que l'allaitement n’était pas toujours possible à cause des engagements professionnels et ont souligné leurs préoccupations au sujet de l'arrêt de la provision gratuite de lait maternisé pour nourrissons.

Conclusion

Ces résultats suggèrent que les femmes vivant avec le VIH jaugent des influences complexes dans la décision de leurs stratégies préférées pour l'alimentation du nourrisson. Cela souligne la nécessité d'une prestation étendue, l’éducation des patients et de la collectivité afin d'assurer un message uniforme, tout en permettant une adaptation aux circonstances de chaque mère.

Abstract

Objetivo

Explorar las diversas influencias sobre las intenciones y prácticas de alimentación de lactantes en mujeres viviendo con VIH en Sudáfrica.

Métodos

207 mujeres embarazadas y 203 mujeres en postparto completaron cuestionarios semiestructurados en Ciudad del Cabo, Sudáfrica. En paralelo, 34 entrevistas cualitativas semi-estructuradas exploraron las influencias sobre las estrategias de alimentación de lactantes en mujeres viviendo con VIH.

Resultados

En general, un 50% (104) de las mujeres embarazadas tenían intención de dar el pecho y un 22% (45) de las mujeres en postparto había dado el pecho a su lactante en algún momento. Las mujeres que practicaban la lactancia materna o tenían intención de hacerlo tenían mayor probabilidad de tener agua corriente en sus hogares, una vivienda digna y de recibir consejos en apoyo a la lactancia materna. Recibir consejos del personal sanitario era el principal vaticinador de lactancia materna (razón de probabilidades ajustada [aOR] en mujeres embarazadas: 6.87; IC 95%: 2.67, 17.66; aOR en mujeres en postparto: 4.04; IC 95%: 1.60, 10.19). Otras influencias importantes incluían el tener experiencia previa dando el pecho, el deseo de proteger al lactante del VIH, y la participación de otros proveedores de cuidados. Muchas mujeres también anotaron que la lactancia materna no era posible debido a sus responsabilidades laborales y enfatizaron su preocupación sobre la discontinuidad del suministro gratuito de leche de fórmula.

Conclusión

Estos resultados sugieren que las mujeres viviendo con VIH deben hacer balance entre influencias complejas a la hora de decidir la preferencia en estrategia de alimentación de su lactante. Esto subraya la necesidad de un suministro amplio, educación del paciente y comunitaria para asegurar un mensaje consistente, con suficiente flexibilidad para permitir una adaptación a las circunstancias individuales de las madres.

Introduction

HIV disproportionately affects women of reproductive age, and nearly one-third of all pregnant women in South Africa are living with HIV (National Department of Health 2012). Despite this high burden of disease, policies on how pregnant women living with HIV should feed their infants have been inconsistent. International and national policies were initially supportive of breastfeeding in all pregnant women, but later shifted to promote formula feeding in certain situations after evidence emerged that HIV could be transmitted through breast milk. Subsequently, South African (SA) policies promoted exclusive breastfeeding (EBF) or exclusive formula feeding (EFF) for infants (Doherty et al. 2011; Goga et al. 2012). South Africa began to provide free formula as a part of its prevention of mother-to-child transmission of HIV strategy in 2002 (Doherty et al. 2011). The SA guidelines changed again in 2011, and the Department of Health discontinued free provision of formula for women living with HIV in favour of a policy that promoted EBF with protection against post-natal HIV transmission through antiretroviral use by the mother and the infant (Motsoaledi 2011).

As a result of these changing policies, pregnant women living with HIV have received diverse and, at times, conflicting guidance on how to feed their infants, alongside suboptimal infant feeding counselling and support (Chopra et al. 2005; Goga et al. 2012). This has resulted in confusion among healthcare providers and women (Chisenga et al. 2011).

Recent studies have consistently shown that breastfeeding uptake is low within South Africa (Tylleskar et al. 2011; Ijumba et al. 2012). For example, the most recent SA Demographic and Health Survey found that only 8.3% of infants were exclusively breastfed in 2003 (National Department of Health 2007). Women in one study who reported early breastfeeding did not maintain the practice for very long (Doherty et al. 2006).

Previous research has shown that infant feeding choices of women living with HIV were influenced by a variety of actors and experiences. These actors included financial providers (Buskens et al. 2007), health workers and family members (Bland et al. 2002). Feeding choices were also driven by the desire to protect the infant from HIV (Doherty et al. 2006), ability to afford replacement feeding (Thairu et al. 2005; Buskens et al. 2007; Chisenga et al. 2011), cultural norms around mixed feeding (Doherty et al. 2006; Buskens et al. 2007) and perceptions that breast milk was inadequate or insufficient (Buskens et al. 2007). A perceived association of formula feeding with HIV infection has also been noted as influential (Thairu et al. 2005; Chisenga et al. 2011).

In light of changing policies around infant feeding and HIV in South Africa and the importance of improving feeding practices in women living with HIV, there is a clear need to better understand HIV-positive women's infant feeding motivations, experiences and practices. Previous quantitative research has demonstrated low breastfeeding uptake and qualitative research has shown diverse influences on infant feeding, but there have been few mixed methods studies that explore HIV-positive women's experiences with infant feeding. Additionally, exploration of the factors that currently affect infant feeding decisions can highlight areas for future interventions to address breastfeeding in women living with HIV.

Methods

This convergent parallel, mixed methods study used structured, quantitative questionnaires and semi-structured, qualitative interviews to explore infant feeding practices, intentions and influences of HIV-positive women in Cape Town, South Africa. A convergent parallel, mixed methods study is one in which researchers concurrently collect and analyse two separate strands of qualitative and quantitative data (Creswell & Plano Clark 2011). This mixed methods study design was selected because it allows for a more complete and rigorous understanding of women's experiences through methodological triangulation (Bryman 2006; Creswell & Plano Clark 2011). These methods also allow for a more nuanced exploration of the context (Wagner et al. 2012) of infant feeding in women living with HIV.

All women living with HIV who attended services at one antenatal clinic (ANC) and one antiretroviral therapy (ART) clinic in the urban township of Gugulethu were screened for participation in the study over 4 months in 2011. As described above, South Africa was in the process of discontinuing the free provision of infant formula to HIV-positive women during the time of data collection, but free formula was still available in these clinics. Women were eligible for the study if they had been diagnosed with HIV and were currently pregnant and/or had delivered a live baby within the past 2 years.

Pregnant and post-partum women were purposively invited to participate in the qualitative study to ensure diversity of recruitment clinic and infant feeding intentions and experiences. All participants provided written informed consent and the University of Cape Town Research Ethics Committee approved the study protocol.

Quantitative strand

Eligible participants were interviewed in Xhosa by trained field workers. Questionnaires covered the socio-demographic background, diagnosis and treatment experiences, and infant feeding practices or intentions of participants. Quantitative data were analysed using Stata (Stata Statistical Software: Release 10.1; Stata Corp., College Station, TX, USA). Participant characteristics were compared using t-tests and Pearson's Chi-squared tests, as appropriate; all tests were two-sided at α = 0.05. Simple logistic regression analyses determined associations between independent variables and the binary outcome of intention to breastfeed for pregnant women and ever breastfed for post-partum women. Multiple logistic regression models were used to assess the association of a priori exposures of interest and breastfeeding.

Qualitative strand

Semi-structured, qualitative interviews were conducted in Xhosa with a subsample of pregnant and recently post-partum participants who were included in the quantitative strand. Interviews explored participants' experiences during pregnancy, HIV diagnosis and disclosure, and infant feeding practices and/or intentions. Transcripts were imported into Atlast.ti 6.2.16 (Scientific Software Development 2010) and read twice prior to coding. Data were coded in a systematic manner using constant comparison to enhance comprehension. Preliminary coding used an open, grounded coding approach in which repeating ideas were identified and coded (Green & Thorogood 2009). These repeating ideas were then organised into themes of influences on feeding practices, which informed the theoretical constructs for analysis (Auerbach & Silverstein 2003). Key domains of interest were infant feeding lessons, infant feeding motivations, facilitators of breastfeeding, barriers to breastfeeding and information sources.

Results

Of the 406 women living with HIV who participated in the quantitative strand, 203 were recently postpartum and 207 were pregnant (four women were both recently post-partum and currently pregnant). The average age of pregnant women was 28.6 years and that of post-partum women 30.0 years, as shown in Table 1. Most participants had a partner (69% of pregnant women and 76% of post-partum women), and 77% of women with partners reportedly disclosed their HIV status to their partner. Participants in both groups had an average of a tenth-grade education. On average, pregnant women tested HIV positive more recently than post-partum women.

Table 1. Characteristics of pregnant and postpartum women by infant feeding plans and practices
 Pregnant women Postpartum women
Does not plan to breastfeed (= 103) Plans to breastfeed (= 104) Total (= 207) P-valuea Never breastfed (= 158) Ever breastfed (= 45) Total (= 203) P-valuea
  1. SD, standard deviation.

  2. a

    Two-sided, two sample t-test for continuous variables and Pearson's Chi-squared test for categorical variables.

Age (years), mean (SD)28.9 (4.8)28.4 (6.0)28.6 (5.5)0.5230.2 (5.6)29.6 (4.5)30.0 (5.4)0.54
Marital status, %   0.88   0.97
Single31.130.130.6 24.224.424.3 
With partner68.969.969.475.875.675.7
Education (Grade)10.9 (1.9)10.6 (2.2)10.7 (2.1)0.2310.4 (1.6)10.6 (1.8)10.5 (1.6)0.40
Positive test (Year) 2008 (3.5) 2009 (3.2) 2008 (3.4) 0.03 2007 (3.5) 2008 (3.3) 2007 (3.5) 0.07
Ever on ART, %43.732.738.20.1016.531.119.70.03
Home, %   0.04   0.11
Shack 68.0 51.9 59.9  62.0 46.7 58.6  
Hostel 3.9 9.6 6.8 8.2 6.7 7.9
House 28.2 38.5 33.3 29.8 46.7 33.5
Water source, %   0.01   0.02
Tap in Home 29.1 43.3 36.2 27.2 44.4 31.0
Tap in Yard 18.5 26.0 22.2 20.3 26.7 21.7
Tap off Property 52.4 30.8 41.6 52.5 28.9 47.3
Clinic advice, %    <0.01   <0.01
Breastfeed 64.3 90.2 76.8  45.2 82.2 53.5  
Formula feed 3.1 0.0 1.6 15.9 2.2 12.9
No Advice/choose for Self 32.7 9.8 21.6 38.9 15.6 33.7
Friend advice, %   <0.01    0.01
Breastfeed 5.1 16.8 11.0  5.7 20.0 8.9
Formula feed 19.2 0.0 9.5 30.6 20.0 28.2
No Advice/choose for self 75.8 83.2 79.5 63.7 60.0 62.9

Overall, 22% (n = 45) of post-partum women reported ever breastfeeding; the average time of cessation was 10 weeks post-partum (interquartile range: 2–12). Most (93%, n = 188) women reported that they had ever used formula, including 73% (33 of 45) of women who had ever breastfed. Eighty-four percent (n = 157) of women who used formula had received free formula from the clinic. Higher rates of breastfeeding intention were reported by pregnant women: 50% (n = 104) reported that they planned to breastfeed.

Significant differences existed between women who intended to or had breastfed and those that did not. Pregnant women who intended to breastfeed were diagnosed more recently and were more likely to live in a house or hostel, to have closer access to running water and to receive advice to breastfeed (P-values < 0.05). Post-partum women who breastfed, as compared to women who never breastfed, were less likely to access water from a public tap and were more likely to receive advice to breastfeed (P-values < 0.05). There was no significant difference in HIV disclosure by feeding group in pregnant or post-partum women.

Simple and multiple regressions of predictors of intention to breastfeed in pregnant women are shown in Table 2. Pregnant women who had a tap off of their property were 60% less likely to intend to breastfeed than those with a tap in their home [relative odds (RO): 0.40, 95% CI: 0.21–0.75]. Women who received clinic advice to breastfeed were roughly five times more likely to plan on breastfeeding than women who received no clinic advice or were told to choose for themselves (RO: 4.68; 95% CI: 2.08–10.52). No woman who was advised to formula feed by her clinic or her friends reported intention to breastfeed.

Table 2. Crude and adjusted relative odds of intention to breastfeed in pregnant women (= 207)
 Crude associationsAdjusted associationsa
RO 95% CI P-value ARO 95% CI P-value
  1. RO, relative odds; ARO, adjusted relative odds.

  2. a

    Adjusted model includes water source, education, year tested positive, clinic advice, and friend advice.

  3. b

    Friend advice and clinic advice to formula feed predicted failure to breastfeed perfectly.

Home
Shack (ref)1.00 0.05   
Hostel 3.24 0.96, 10.90   
House 1.79 0.99, 3.24   
Water source
Tap in home (ref)1.00 <0.011.00 <0.01
Tap in yard 0.95 0.45, 2.001.04 0.39, 2.76
Tap off property 0.40 0.21, 0.750.23 0.10, 0.54
Positive test, year 1.10 1.01, 1.200.03 1.11 1.00, 1.240.05
Education, grade 0.92 0.81, 1.050.23 0.78 0.64, 0.940.01
Clinic advice
No advice/choose for self (ref) 1.00  <0.01 1.00 <0.01
Breastfeed 4.68 2.08, 10.526.87 2.67, 17.66
Formula feed b  b
Friend advice
No advice/choose for self (ref) 1.00  0.04  1.00 0.06
Breastfeed 3.04 1.07, 8.633.12 0.94, 10.33
Formula feed b  b  

In multiple regression analyses, water source, year of diagnosis, education and receipt of clinic advice to breastfeed significantly predicted breastfeeding intention in pregnant women. Breastfeeding intention was highest in women who received clinic advice to breastfeed, as compared to women who received no advice or were advised to choose for themselves [adjusted relative odds (ARO) = 6.87; 95% CI: 2.67–17.66]. The ARO of breastfeeding were also lower in women who had no water tap on their property (ARO = 0.23; 95% CI: 0.10–0.54). Lower education was associated with breastfeeding in the adjusted analysis with a 22% decrease in the relative odds of breastfeeding with each additional year of schooling (ARO = 0.78; 95% CI: 0.64, 0.94).

Simple and multiple regressions of predictors of breastfeeding in post-partum women are shown in Table 3. Post-partum women who ever breastfed were significantly more likely to have a tap in their home, test positive more recently and receive advice to breastfeed. Strong crude associations also existed between experience on ART (RO = 0.44; 95% CI: 0.20–0.93) and breastfeeding.

Table 3. Crude and adjusted relative odds of breastfeeding in postpartum women (= 203)
 Crude associationsAdjusted associationsa
RO 95% CI P-value ARO 95% CI P-value
  1. RO, relative odds; ARO, adjusted relative odds.

  2. a

    Adjusted model includes water source, child age, year tested positive, ever on ART, clinic advice, and friend advice.

Home
Shack (ref) 1.00  0.11    
Hostel 1.08 0.28, 4.12   
House 2.09 1.03, 4.19   
Water source
Tap in home (ref) 1.00  0.02 1.00  0.04
Tap in yard 0.81 0.34, 1.890.82 0.31, 2.16
Tap off property 0.34 0.15, 0.740.33 0.14, 0.80
Positive test, year 1.10 0.99, 1.230.07 1.14 1.02, 1.270.02
Ever on ART 0.44 0.20, 0.930.03 0.44 0.18, 1.070.07
Clinic Advice
No advice/choose for self (ref) 1.00  <0.01 1.00  <0.01
Breastfeed 4.54 1.89, 10.924.04 1.60, 10.19
Formula feed 0.35 0.04, 2.980.28 0.03, 2.52
Friend advice
No advice/choose for self (ref) 1.00  0.01 1.00  0.12
Breastfeed (ref) 3.70 1.34, 10.242.96 0.87, 10.14
Formula feed 0.69 0.30, 1.590.71 0.29, 1.75

Multiple regression analyses showed that water source, positive test year and receipt of clinic advice to breastfeed significantly predicted breastfeeding in post-partum women. Breastfeeding was highest in women who received clinic advice to breastfeed (ARO = 4.04; 95% CI: 1.60–10.19) and in women who were diagnosed more recently (ARO = 1.14; 95% CI: 1.02, 1.27). The ARO of breastfeeding were lower in women who used an off-property tap (ARO = 0.33; 95% CI: 0.14–0.80).

Feeding choice influences

Qualitative interviews were conducted with 12 recently post-partum and 22 pregnant women who were purposively selected based on feeding intentions and practices. Qualitative interviews suggested that women's decisions on how to feed their infants were complex and shaped by a broad array of factors. The salient themes that emerged during analysis were that infant feeding choices were influenced by the health system, personal experiences, fear of HIV transmission, involvement of other caregivers and free formula stigma and discontinuation.

Health system

Women received varied infant feeding guidance, primarily from the clinics where they received care. The majority of women stated that they received counselling – and occasionally pressure – from ANC staff on the benefits of breastfeeding. ANC staff encouraged all clinic attendees to breastfeed, regardless of their HIV status or expressed feeding desires. One pregnant woman explained, ‘I said to the nursing sister, “I do not want to breastfeed because the baby will be infected” and she said we should breastfeed, all of us’. Clinic lessons emphasised the importance of breastfeeding in facilitating mother–infant bonds and in increasing child health. Some women were also advised by family and friends that breastfeeding was more nutritious and would lead to healthier babies.

Despite high levels of breastfeeding promotion in ANC, some women reported misinformation and mixed messages regarding HIV transmission through breast milk. Many believed that HIV was only spread through blood from cracked nipples or teething babies. Women also expressed surprise that the policy had shifted from promotion of formula feeding to breastfeeding, ‘For me, it was news that someone positive can breastfeed, because before they were not allowed to breastfeed’.

Women often concurrently attended services both at the ART clinic where they were generally told to formula feed and at the ANC where breastfeeding was promoted. Some women reported confusion over varied infant feeding guidance from different healthcare providers, while others felt they received no guidance at all. Women explained that they ultimately had to decide how to feed their baby based on their own preferences, ‘Counsellors can confuse you. The other says this and the other say that, but you must also use your own brain. You must decide which one you believe’.

Personal experiences

A strong influence on feeding choice was previous experiences with infant feeding. Women generally preferred to feed their infants in the same way that they fed their other children. One pregnant woman from the ANC explained, ‘They said it's up to a person if they want to breastfeed or to bottle feed and I said I am not used to bottle feeding. I will try my best again in breastfeeding because breastfeeding is the best and I did not get a problem’. Yet women who had experience with either their own or other women's HIV-positive children who were breastfed consistently reported that they would not take a risk and preferred to formula feed.

Fear of HIV transmission

Decisions to formula feed were influenced by fear of HIV transmission and a desire to avoid mixed feeding. One post-partum woman explained that she changed her feeding plans upon diagnosis, ‘I had chosen to breastfeed, but when I got my HIV results I told them that I had changed my mind, that I don't want to breastfeed anymore. I wanted to formula feed, because I didn't trust breastfeeding while positive. Even though breast milk is nutritious, I don't think it is safe for the baby’. These women heard the lessons from the ANC on breastfeeding and the use of ART to reduce transmission risk, but decided that they were not willing to risk transmission.

Involvement of other caregivers

Many women also felt that breastfeeding was not feasible because others were involved in the care of the infant who might provide the infant with food and other liquids. One post-partum woman on ART reported that her service provider alerted her to the risk and dangers of mixed feeding that could occur when the baby is breastfed.

The nurse who was advising was open enough to advise me that in whatever I'm doing I shouldn't put myself in a situation that may compromise my baby's health. I never breastfed because she explained that if I breastfeed it wouldn't be easy to leave my child with someone when I need to go somewhere without disclosing to them, because when you leave the child, the child minder might give the baby anything whereas you are supposed to be exclusively breastfeeding.

Women who were working or looking for work frequently reported this concern. They preferred formula feeding because it gave them more flexibility in what they fed the baby and in child care. A post-partum woman from the ART clinic explained, ‘One of my reasons I chose not to breastfeed was because when not breastfeeding, I can easily give her something else e.g. maize meal porridge, but if I breastfeed, that is not allowed’.

Free formula stigma and discontinuation

Despite women's appreciation for free formula, they indicated that the formula is stigmatizing because the community knew that HIV-positive women received free formula. ‘People talk. For example, as a positive mom, I do get formula at (the clinic). As you are sitting waiting as women, there will be comments like, “it's HIV positive people who get formula,” you see? There are a lot of comments. But then, you've got to because you want to save your baby … It hurts. It hurts so some would take this formula from their community centre and they would sell the (free formula brand) so as to buy different formula’.

Thus, women developed a range of strategies to discreetly obtain and store the formula, ‘When I get formula from the clinic, I put it in a different tin from the shop. The one that they know is for HIV positive people, I throw it away’. Women reported that friends and family collected formula from the clinic on the behalf of some HIV-positive women. Others explained that they decided to buy formula instead of being identified as HIV positive through getting free formula.

Finally, women expressed concern about the end of provision of free formula. They felt that formula allowed them to provide the best care for their children and to save their money for other expenses. Numerous women stated that they would buy formula if its provision was discontinued, but they felt this would create financial hardship and potentially mixed feeding and HIV-infected babies due to insufficient quantity of formula. They also stated that it will inhibit mothers' ability to feed their infants according to their preferences, ‘when milk is finished they will give the babies breast milk because they do not have money for formula and then they end up mixed feeding… It will affect a particular person's choice because she will feed the baby the way she did not want to and do it without care because it's not her choice’.

Discussion

This mixed methods study found substantial variation in perceptions, intentions and practices related to infant feeding among women living with HIV in South Africa. Similar to previous research (Tylleskar et al. 2011; Ijumba et al. 2012), women in this study reported low levels of breastfeeding, with only 22% of post-partum women reporting ever breastfeeding. The strongest predictor of breastfeeding or intention to breastfeed was clinic guidance in support of breastfeeding, a finding that has been shown previously (Doherty et al. 2006). Breastfeeding practices or intentions were also higher in women who had running water in their homes, an unexpected finding as these women were more likely to have water accessible to safely mix formula. Additionally, post-partum women were more likely to breastfeed if they were diagnosed more recently, and pregnant women were more likely to intend to breastfeed if they had less schooling.

Qualitative findings facilitated a more nuanced understanding of the diverse influences on infant feeding intention and practices. Similar to the quantitative findings, clinic advice had a salient influence on infant feeding. There were, however, other important influences on women's infant feeding decisions, most notably family and friends. This supports the notion that infant feeding is a social practice (Lazarus et al. 2013). The influence of actors outside of the healthcare setting underscores the need for infant feeding counselling and messaging that target entire families and communities. Women's ability to enact feeding decisions was dependent on their broader cultural, economical and social context. This was particularly true for working mothers who felt that their situation prevented breastfeeding. It is therefore important that interventions be developed to facilitate enabling environments for optimal infant feeding, such as involvement of partners in antenatal care (Chisenga et al. 2011).

Despite the importance of clinic messages on infant feeding, this analysis indicated that women encountered divergent guidance and ultimately decided based on their own priorities and experiences. Lessons within the ANC were generally promotive of breastfeeding while those in the ART clinic often promoted formula feeding. Given that ART services were not integrated into ANC at the time of data collection, this meant that women received discrepant guidance at the different clinics that they attended. As has been noted elsewhere, mixed messages were common that resulted in confusion (Doherty et al. 2006; Chisenga et al. 2011). Many women also expressed low levels of knowledge regarding HIV transmission and did not completely understand the mechanism and likelihood of MTCT through breast milk or the effect of ART. This demonstrates the need for tailored, accurate and accessible information about infant feeding (Thairu et al. 2005). It also highlights the value of closer alignment between ANC and ART to ensure consistent messages for pregnant women living with HIV. The recent SA policy shift to provide ART within ANC clinics holds considerable potential as a mechanism for ensuring consistent messaging across providers and increased ART uptake, as pregnant women starting treatment will only attend one clinic for both ART and ANC services in many settings.

This study also found that, despite local policies promoting breastfeeding, there are significant barriers to breastfeeding for many women. These include previous experience with formula feeding, the involvement of other caregivers and fear of HIV transmission. Women in SA were instructed for nearly a decade that HIV could be transmitted through breast milk. Most participants had previous pregnancies in which they received free formula to prevent this transmission. It is therefore unsurprising that they reported confusion and discomfort with the new guidelines, particularly given that the study was conducted when policy shifts in formula provision were being discussed, but not enacted. This underscores the need for patient counselling that is tailored to the individual woman's situation and background to facilitate optimal maternal and infant outcomes. Study participants emphasised that while free formula from the clinic was stigmatised, its discontinuation would be problematic and may lead to increased mixed feeding. Particular attention must be given to meet the needs of women who have used formula previously and to explain the rationale for the recent policy shifts within South Africa (Ijumba et al. 2012).

Many influences on feeding choices were difficult to quantify, highlighting the value of a mixed methods approach. This approach also allowed for a more complete and nuanced understanding of the diverse actors and experiences that influenced the infant feeding decisions of HIV-positive women. Despite the strengths of this methodology, there are limitations to the current study. A major limitation of this study is that women were interviewed at one time, limiting our ability to look at temporal trends and influences in feeding intentions and practices. It is not possible to determine whether feeding intentions of pregnant women changed post-partum. The study was also conducted only in one community, which may limit the generalisability of this study's findings.

Conclusion

This study found that health provider messages were an important influence on infant feeding practices in women living with HIV, but that these lessons were then interpreted and enacted in a broader social context that both enabled and hindered breastfeeding. These results suggest that women living with HIV balance complex influences in arriving at decisions regarding preferred infant feeding strategies. This highlights the need for extensive provider, patient and community education on infant feeding to ensure consistent messaging based on national policies, while allowing for adaptation to the circumstances of individual mothers. Particular attention must be given to meet the needs of women who are working, attending both ART and ANCs or who have used formula previously.

Acknowledgements

We thank the women who participated in this study. We are also indebted to the staff at the two study sites for their assistance. LM is supported by an International Leadership Award from the Elizabeth Glaser Pediatric AIDS Foundation.

Ancillary