The design of the study on which this article is based was longitudinal, descriptive and correlation. The study was part of a large project that aimed at exploring culture-specific influences of exclusive breastfeeding among HIV-positive mothers in Blantyre, Malawi. The large project employed both quantitative and qualitative methods. This paper is based on the baseline data from the quantitative phase of this project.
This study was conducted at Queen Elizabeth Central Hospital (QECH) maternity unit in Blantyre, Malawi. Blantyre is one of the districts in the Southern Region of Malawi. It has a total population of 1.01 million, 68% of whom live within the urban area . The district is served by QECH, a public hospital which functions both as a district hospital for Blantyre and as a referral hospital for the southern region of Malawi. QECH is also a teaching hospital for different health related professions. Blantyre is one of the districts in Malawi with high prenatal care coverage. Most of the women in Blantyre receive prenatal care from skilled health care providers with 92.2% of them cared for by midwives. Seventy percent of women in Blantyre deliver in public health facilities . QECH has organized services for PMTCT of HIV and prenatal HIV testing is an important gateway into these services. Availability of PMTCT services provided access for research to HIV-positive prenatal and postnatal mothers without asking them to go through HIV screening in order to participate in the study.
The study utilized a convenience sample of 110 HIV-positive pregnant women. To participate in this study, one had to be at least 18 years old and with a singleton pregnancy, at least 36 weeks pregnant, intending to breastfeed her baby (whether exclusively or not), and a resident of Blantyre district. Furthermore, one had to be able to hear, understand and speak Chichewa (a national language for Malawi), willing to give informed consent, and be re-interviewed at 12 weeks postnatal. Women who were experiencing tuberculosis (TB) or full-blown AIDS (self reported record of TB or a record of being chronically sick) were excluded from this study.
Sample size and power calculation
A minimum sample size of 87 was estimated by GPower 3.0.8 for multiple linear regression analysis using 13 variables: 3 belief constructs of TPB measured by 4 subscales of the Breastfeeding Attrition Prediction Tool (BAPT), and 9 socio-demographic variables (external influences) at alpha value of 0.05, power of 0.80, and effect size of 0.24. The effect size of 0.24 represented the smallest effect size from literature review of studies on breastfeeding from other, mostly industrialized countries that had used the Theory of Planned Behavior [23–25]. One hundred and ten (110) participants were recruited in order to allow for an attrition rate of 23%.
New York University Committee on Activities Involving Human Subjects and Malawi College of Medicine Research and Ethics Committee approved this study. In addition, the researchers sought permission from the director of Queen Elizabeth Central Hospital (QECH), Head of Obstetric and Gynecology Department and Head of Pediatric Department of QECH to access participants. Written informed consent was obtained from individual participants.
The researchers collected data between May 12, 2009 and March 22, 2010 utilizing a structured and modified exclusive breastfeeding attrition prediction tool. Care providers of participants were not involved in data collection.
Instrumentation and measures
Beliefs (behavioral, normative and control) were measured at interval level utilizing the four sub-scales of the exclusive breastfeeding attrition prediction tool (EBAPT). EBAPT was adapted from the original breastfeeding attrition prediction tool (BAPT) .
Negative exclusive breastfeeding beliefs (NEBB) sub-scale (12 items) and positive exclusive breastfeeding beliefs (PEBB) sub-scale (13 items) measured behavioral beliefs. Each item was a weighted item, which was obtained by multiplying each behavioral belief by its corresponding motivation to comply statement. Items included: (1) ‘Exclusive breastfeeding is more convenient than mixed feeding’; ‘Using a feeding method that is convenient is…’ (2) ‘It is difficult to exclusively breastfeed’; ‘Using a feeding method that is easy to do is…’. Response choices for each belief and its corresponding motivation to comply were on a 5-point Likert scale ordered on a continuum from 1 = strongly disagree to 5 = strongly agree; and 1 = not important at all to me to 5 = very important to me; respectively. Addition of all the weighted items in the sub-scale produced a total score for the sub-scale. Total possible scores ranged from 12-300 and 13-325 for NEBB and PEBB respectively. The higher the score of NEBB, the stronger the negative exclusive breastfeeding beliefs the HIV-positive woman had. Likewise, the higher the score of PEBB, the stronger the positive exclusive breastfeeding beliefs the HIV-positive woman had.
Normative belief (NorB) sub-scale (10 items) measured normative beliefs. As for NEBB and PEBB sub-scales, each item was a weighted item obtained by multiplying each normative belief by its corresponding motivation to comply statement. Items and their corresponding motivation to comply included: (1) ‘The baby's father thinks I should…’; ‘How much do you care about the baby’s father’s opinion on how you should feed your baby?’; (2) ‘My nurse-midwife thinks I should…’; ‘How much do you care about your nurse-midwife’s opinion on how you should feed your baby?’ Response choices for each normative belief and its corresponding motivation to comply were ordered from 1 = definitely not exclusively breastfeed to 5 = definitely exclusively breastfeed; and 1 = I do not care at all to 5 = I care very much; respectively. The response choices also provided participants with an option of 0 = not Applicable. A total score for the sub-scale was obtained by adding all the weighted scores. Total possible scores ranged from < 10-250. The higher the score, the greater the support the HIV-positive woman had for exclusive breastfeeding.
Exclusive breastfeeding control beliefs (EBCB) sub-scale (10 non-weighted items) measured control beliefs. Items included: ‘I have the necessary skills to exclusively breastfeed,’ ‘I am determined to exclusively breastfeed’ and ‘I am confident I can exclusively breastfeed’. Response choices were on a continuum from 1 = strongly disagree to 5 = strongly agree. Total possible scores ranged from10-50. The higher the score, the greater the perceived control the HIV-positive woman had over exclusive breastfeeding. (See the more detailed description of scoring the beliefs’ items of the breastfeeding attrition prediction tool in the Additional file 1).
External salient culture-specific influences
According to TPB, any factors that may have an effect on behavior do so by influencing behavioral, normative and control beliefs. Such factors are called external influences. Literature review and analysis of data from the qualitative component of the larger study from which this paper comes revealed some socio-demographic factors that may influence exclusive breastfeeding intentions and duration. These factors were socio-economic status (SES), maternal age, marital status, maternal education, previous experience of exclusive breastfeeding, place of delivery, pre-lacteal feeding practices, disclosure of HIV status, and timing of decision to breastfeed the baby. In the current study, these factors are referred to as external salient culture-specific influences. These factors were measured using the demographic items on the EBAPT.
Intentions to exclusively breastfeed
Two items that are part of EBAPT measured intentions prenatally. The item ‘How many weeks do you intend to exclusively breastfeed your infant?’ was used to measure intentions at a ratio level. Because, at the time of this study, all prenatal HIV-positive women who could not afford infant formula in Malawi were advised to exclusively breastfeed their infants for the first six months , measuring prenatal exclusive breastfeeding intentions on a Yes/No was likely to produce less reliable results due to social desirability . Therefore, the researchers decided to measure intentions as intended duration of exclusive breastfeeding in weeks to reduce the social desirability effect. However, since intentions are often measured on a Likert scale , a second item ‘How likely is it that you will exclusively breastfeed your baby for six months?’ was also used to measure intentions. The item was on a 5-point Likert scale with 5 = very likely; 4 = likely; 3 = not sure; 2 = unlikely; 1 = very unlikely.
Actual exclusive breastfeeding duration in weeks
The actual duration of exclusive breastfeeding in weeks was assessed at 12 weeks postnatal follow-up. We chose a 12-week follow-up point because it was beyond both national (Malawi) and local (Blantyre district) median durations of exclusive breastfeeding; 10 weeks (2.5 months) and approximately 5 weeks (1.3 months) respectively . Therefore, a follow-up at 12-weeks postnatal allowed the researchers to identify factors associated with mothers who were still exclusively breastfeeding their infants. If the postnatal follow-up was later than 12 weeks, for example at 24 weeks postnatal, few if any mothers would still be exclusively breastfeeding.
A 7-item questionnaire was used. The items included: (1) the birth date of the infant and (2) feeds that were given to the infant before lactation was established. This helped the researchers to determine initiation of exclusive breastfeeding; (3) feeds other than breast milk that were being given to the infant at the time of the follow-up helped the researchers to confirm exclusive breastfeeding for those who indicated that they were still exclusively breastfeeding their infants; and (4) age of baby in weeks when feeds other than breast milk were initiated to determine time when exclusive breastfeeding was terminated.
Reliability of the instrument
The original BAPT  has an overall reliability of 0.80, with negative breastfeeding sentiments (NBS) having a Chronbach alpha value of 0.83, positive breastfeeding sentiments (PBS) 0.79, social and professional support sub-scale (SPS) 0.85, and breastfeeding control (BFC) 0.81. The item loading for each sub-scale was: NBS 0.31-0.70, PBS 0.30-0.60, SPS 0.31-0.75, and BFC 0.41-0.76. In addition, the original BAPT has a predictive validity of 73% .
BAPT has been tested with both prenatal [28, 29] and postnatal [28, 30] women, and has demonstrated adequate reliability ranging from 0.77-0.93 for all the sub-scales in most studies. In the current study, estimates of internal consistency using Cronbach’s Alpha coefficient were computed for each sub-scale in the EBAPT instrument and EBAPT as a whole using the Statistical Package for Social Science (SPSS). Results revealed a good internal consistency in all the four sub-scales of the BAPT and BAPT as a whole with Cronbach’s alpha of greater than .80.
The 12-item NEBB, 13-item PEBB and 10-item EBCB sub-scales had a Cronbach’s alpha of .881, .893 and .917 respectively. There was no item with a corrected item-Total correlation (Item loading) of less than .3 in all of these sub-scales. This indicates that all the items in these sub-scales were measuring the same thing . The 10-item NorB sub-scale had an alpha of .846. There was one item ‘My Traditional Birth Attendant thinks I should EBF’ with a corrected item-Total correlation of less than .3. However, the suggested Cronbach’s alpha if this item was deleted was .856, which did not differ greatly from the overall Cronbach’s alpha of .846 for the sub-scale. Therefore, the item was retained in the sub-scale. Finally, the 45-item EBAPT instrument as a whole had a Cronbach’s alpha of .858. There were five items with a corrected item-Total correlation of less than .3. An evaluation of the suggested Cronbach’s alpha values if each of these items were deleted one at a time did not greatly increase the overall alpha value for the EBAPT instrument. These suggested alpha values ranged from .852-.862 compared to the overall alpha value of .858. This indicates that the 45-item BAPT instrument was a reliable instrument for measuring exclusive breastfeeding beliefs.
Data analysis was done using SPSS version 16.0 for windows. Pearson’s product-moment coefficient, Spearman’s rho coefficient and Pearson’s Chi-square test for independence were used to explore associations among the variables. To be consistent with previous studies that have used Janke’s original BAPT instrument and TPB, multiple linear regression analyses were run with number of weeks of intended duration of exclusive breastfeeding at baseline. A sequential modeling approach was used. All the external salient culture-specific influences were entered together at the first step, and then behavioral, normative and control beliefs, one at a time, to assess their effects on the outcome variables. Only the predictor variables that were significantly associated with prenatal intended duration of EBF were entered into the models. However, because some variables did not meet the assumption of normalcy, logistic regression analyses were also run to confirm the findings.