A community-based prospective cohort study of exclusive breastfeeding in central Nepal

Nepal is a small but diverse country in South Asia. Female literacy rate is 67% and total fertility rate is 2.6 births per women. The median age at first birth is 20.2 years. Most of the married women are occupied with home duties [20]. The location of this study was Kaski, a centrally located hills district of Nepal. The district has 75% female literacy rate and approximately 13,800 babies are born annually [21].

A large community-based prospective cohort study of utilisation of maternity services was conducted between December 2011 and November 2012 in the Kaski district. Details about participants, study design and sampling procedure had been described elsewhere [22]. Briefly, a total of 701 pregnant women of 5 months or more gestational age were recruited from five urban wards and seven rural illakas of the Kaski district, and were followed up for six months after delivery. Fifteen female data enumerators, who were recruited locally, searched for and identified pregnant women in their locality with the help of female community health volunteers and health facility registrations for antenatal care visits. They then visited the pregnant women’s houses.

The study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects/patients were approved by the Human Research Ethics Committee of Curtin University (approval number HR 130/2011), and the Ethical Review Board of the Nepal Health Research Council (approval number 88/2011). A STROBE checklist can be found in Additional file 1. Written informed consent was obtained from all subjects. An information sheet was distributed and read to each consented subject before obtaining her signature or thumb-print. Confidentiality of the information provided was maintained throughout the study.

The questionnaire used in the interview was adapted from the NDHS and another validated instrument on breastfeeding; see Additional file 2[23]. It was pretested on 25 postpartum women for cultural appropriateness, content validity and understanding. Fifteen female community health volunteers, trained by the first author, conducted the baseline interviews to collect information on socio-demographic and obstetric characteristics of the pregnant women. The participants were followed up with three subsequent household visits by the same female enumerators at 4 weeks, 12 weeks and 22 weeks postpartum, respectively. Breastfeeding at birth was assessed during the second follow-up interview.

The outcome variable was the reported duration of ‘exclusive breastfeeding’, defined as the infant being given breastmilk only without any other feeds (aside from medications) since birth [2]. The main exposure variable of interest was residential location (rural versus urban). Table 1 lists other maternal socio-demographic and obstetric variables. Four levels of education were recorded, namely, none, primary (1-5^th grade), secondary (6-10^th grade), and college (after 10^th grade). Caste was defined according to the government’s classification in the health system. ‘Upper caste’ and ‘lower caste’ referred to Indo-Aryan people, whereas ‘janajati’ referred to Tibeto-Burman people and the term ‘religious minorities’ denoted people who are Muslim or Christian. In this study, only three women belonged to the religious minority so they were merged with the janajati group. Employment status was categorised as employed (full-time salaried job), semi-employed (wage based labour, small business or employed abroad), and unemployed (agricultural, housewife or nothing). Breastfeeding information was recorded yes if the woman had reported receiving any breastfeeding education or information from anywhere on how to feed her baby. Encouragement to breastfeed was considered present if the woman had reported receiving encouragement or motivation to breastfeed from health workers, family members or relatives. Breastfeeding problem meant painful swelling of breast, and inverted, cracked or sore nipples affecting lactation.

In the presence of censoring observations due to lost to follow up or continuation of exclusive breastfeeding beyond the study period, Cox’s proportional hazards modelling was used to assess the effect of residential location on the exclusive breastfeeding duration, accounting for plausible demographic and obstetric confounding variables. Both crude and adjusted hazards ratios (HR) and their 95% confidence intervals (CI) were reported for each significant factor to estimate the corresponding risk of cessation of exclusive breastfeeding, and backward stepwise analysis was undertaken in view of the apparent collinearity between exposure variables. All statistical analyses were performed using the SPSS package version 21.

Figure 1 shows the flow chart of the cohort. Of the initial 701 pregnant women recruited, 639 (91.2%), 615 (87.7%), and 515 (73.5%) mothers were followed up at 4 weeks, 12 weeks, and 22 weeks after delivery, respectively. Table 1 presents the sample characteristics by residential location. About half of the mothers were primiparous (51.5%) within the age group 20–24 years (50.7%). The great majority of them were unemployed (79.7%), and had normal delivery (86.6%) at a health facility (85.5%). Almost all pregnant women (98%) had made at least one antenatal care visit. Most mothers reported receiving information about breastfeeding (73.6%) and were encouraged to breastfeed by health personnel or family members (80.8%), yet 35% of them experienced problems during lactation. Moreover, an examination using Spearman’s rho indicated that certain variables were highly correlated, e.g. caste and education, age and parity. When comparing urban versus rural mothers, the two groups appeared to be significantly different in terms of employment, caste, education, place of delivery, method of delivery, availability of breastfeeding information, and presence of breastfeeding problem; see Table 1. In particular, the prevalence of vaginal delivery was higher among rural women and more of them gave birth at home and received breastfeeding information relative to their urban counterparts.

Table 2 summarises reported breastfeeding rates at birth, 4, 12 and 22 weeks. Although nearly all women (98%) breastfed their infants up to six months, overall the exclusive breastfeeding rate declined from 90.9% at birth to 29.7% at 22 weeks. The main supplementary foods introduced were cow or buffalo milk while only a small proportion of the cohort (6% to 13%) fed their babies formula during the period. The data also indicated consistently higher rates of exclusive breastfeeding at birth and during follow up among rural mothers than urban mothers.

Of the 639 women in the cohort, 581 mothers fed colostrum to their infants at birth and 58 mothers gave other liquid as the first feed. This group of 581 women was eligible for survival analysis of exclusive breastfeeding duration. Kaplan-Meier log-rank test showed significant difference in exclusive breastfeeding between rural and urban women (p = 0.02), with rural mothers (mean 144.7, 95% CI 132.3 to 157.1 days) exclusively breastfed longer than urban mothers (mean 104.5, 95% CI 95.8 to 113.1 days). Table 3 presents the result of stepwise Cox regression analysis, which confirm that urban women were more likely to cease exclusive breastfeeding (HR 1.28, 95% CI 1.03 to 1.60) after adjustment for confounders. Other significant risk factors associated with the discontinuation of exclusive breastfeeding were reported breastfeeding problem (HR 2.07, 95% CI 1.66 to 2.57) and caesarean delivery (HR 1.88, 95% CI 1.36 to 2.62). The Kaplan-Meier curves corresponding to these three factors are shown in; see Additional file 3: Figure S1.

A major strength of the present study concerns the prospective assessment of breastfeeding practices from newly delivered women to minimise recall bias. However, all information obtained, including breastfeeding information and encouragement to breastfeed, was based on self-report by the participants. The occurrence of some recall error on breastfeeding practices and duration during the personal interview could not be ruled out, especially for breastfeeding at birth which was asked about 4 weeks later. Despite feeding status was assessed at periodic intervals, the response might not represent the actual feeding practice throughout the whole 6-month period, because some women merely reported what they were feeding around the time of each interview. This study did not specifically capture whether water was given to the infant. Water was not generally considered as a food so that some participants might perceive they were feeding only breastmilk even though water was given occasionally.

Loss to follow up poses as another limitation. Due to time limit and budgetary constraint, the exact duration of exclusive breastfeeding could not be recorded for all infants and follow up of participants were conducted at periodic intervals without covering the full 6 months (approximately 4, 12 and 22 weeks postpartum). Finally, the findings may not be generalised to the whole country as Nepal is diversified in terms of ecology, ethnicity and social development. The Kaski district was ranked third in human development index out of 75 districts in Nepal with high adult literacy (82%), and rural areas are connected to the central urban valley by gravelled roads [30].