Study setting and data source
Malawi is located in southern-central Africa with an agriculture-dependent economy. The data were extracted from the 2015–16 Malawi Demographic Health Survey (MDHS) . A total of 7970 children were included for analysis. The analysis for early initiation was restricted to the most recent birth in the past 2 years (n = 6351). However, analysis for EBF was restricted to the most recent birth and infants aged 0–5 months (n = 1619) .
Sampling and data collection
The methodology, design and sampling methods of the MDHS have been detailed elsewhere . Briefly, the survey used the 2008 Malawi Housing and Population census as its sampling frame. A two-stage cluster sampling design was employed with 850 clusters selected randomly and household listing conducted in the first stage. Using equal probability systematic selection, households from the chosen clusters were selected in the second stage.
Questionnaires, which were translated into two prominent local languages (Chichewa and Tumbuka), were administered to women of reproductive age (15–49 years) from the selected households via face-to-face interviews. The MDHS questionnaire collects data on a wide range of health indicators, including infant and young child feeding practices in addition to socio-demographic information.
Dependent variables (early initiation of breastfeeding, exclusive breastfeeding)
The dependent variable in this study was optimal breastfeeding measured using two binary variables namely “early initiation of breastfeeding” (early initiation) and “exclusive breastfeeding” (EBF). Early initiation was defined as children less than 24 months of age who were put to the breast within the first hour of birth [1, 12]. Those that reported having initiated breastfeeding within 1 h of birth were recorded as “1” while those that initiated breastfeeding after 1 h were recorded as “0”. Exclusive breastfeeding was defined as the number of infants aged 0–5 months (less than 6 months) who were fed exclusively with breast milk (including milk expressed or from a wet nurse) in the last 24 h [1, 12]. EBF allows the infant to receive oral rehydration salts, drops, syrups (vitamins, minerals, medicines) [1, 12]. Those who reported having fed their infant (0–5 months) with breast milk only or the allowed aforementioned liquids were recorded as “1” while those who had given infants other foods in addition to breast milk were recorded as “0”.
A number of child, maternal, and health-related independent variables were selected for analysis. Child factors included age in months (0–2, 3–5, and 6–23) and sex (male, female). Maternal factors included maternal age in years (15–24, 25–34, ≥35), marital status (married, unmarried), education (no formal education, primary education, secondary/post-secondary), and occupation (employed, unemployed). Wealth was calculated by the MDHS team using a principal component analysis model in which household items such as roofing materials and possession of bicycles were scored. The calculated scores were then divided into quintiles from poorest (lowest 20%) to richest (top 20%) . In this study, the top 40% were categorized as rich, the middle 20% as middle class, and the bottom 40% were categorized as poor. Other maternal factors included region (northern, central, southern), place of residence (urban, rural), parity (primipara, secundipara, multipara), and media exposure measured by exposure to any of the following; newspaper, radio or television (yes, no). Health-related variables included place of delivery (health facility, non-health facility), number of antenatal visits, categorized as adequate (4 or more visits) or inadequate (fewer than 4 visits) [21, 22], antenatal care attendant, categorized as skilled (health professionals) or unskilled (traditional attendants/no one), birth attendant, also categorized as skilled (health professionals) or unskilled (traditional birth attendant/self/no one), mode of delivery (caesarean section, vaginal delivery), birth type (singleton, twin or multiple), and perceived size of the child at birth (small, average, large).
All analyses were performed using Stata version 15.0 (Stata Corp LP, College Station, TX, USA) and considered the complex sample design. The “svy” command was used to adjust for cluster sampling design and sampling probabilities across clusters and strata. The distribution of participants according to early initiation and EBF status were analyzed using the Chi-squared tests. Unadjusted logistic models were used to examine the association between independent variables and early initiation and EBF, respectively. Variables with p-value ≤0.25 in the unadjusted model were manually included in the multivariate logistic models using purposeful selection method . Adjusted odds ratio (aOR) and their 95% confidence intervals (CI) were used to report the strength of association between independent variables and early initiation and EBF, respectively. Sensitivity analysis was conducted to include only women who were tested for HIV for the purpose of controlling for HIV status and the results were fairly consistent. In this sensitivity analysis, HIV status remained insignificant and the sample was reduced to 2080 and 515 for EBF and early initiation, respectively. Therefore, we present results from the whole sample regardless of the availability of HIV status. The level of significance was set at p < 0.05 (two-tailed).