Study design and sample size
We conducted a cross-sectional community-based study from April-September 2016. Based on the study population size of approximately 460 (husbands in 21 urban slums whose wife had childbirth within one year), assuming that 50% of the husbands are involved in BPCR (an optimal prevalence to get maximum sample size), assuming a design effect of 1.5 (due to multi-stage random sampling), and an expected 90% response rate, this study required 214 eligible husbands for estimating the expected proportion with 15% precision relative to the expected proportion and 95% confidence.
Sampling : Fig 1 summarises the multi-stage random sampling method.
Selection of slums: Firstly, based on the slum size and eligibility for various health schemes, we stratified 21 urban slums as large (with ≥100 households), small (<100 households) and designated Scheduled Caste/Scheduled Tribe (SC/ST) slums. According to this stratification, there were four large, nine small and eight designated SC/ST slums. Secondly, considering the resource constraints, we selected 10 urban slums (two large, four small, and four designated SC/ST slums) by simple random sampling with proportional allocation in each category.
Selection of participants: In each of the selected slums, we obtained a list of women who delivered within one year from the Anganwadicentres. The Anganwadi centre is the Integrated Child Development Services (ICDS) scheme’s basic functional unit, which provides supplementary nutrition, non-formal education, and health services to under-five children. We decided the required number of women from each slum by proportional allocation and selected by simple random sampling from the slum wise list of women. None of the randomly selected women was a widow, divorced or separated. We approached the husband of the selected woman in the slum in the afternoon or early evening to ensure their availability and better participation.