The Government of India, in collaboration with Measures DHS, conducted the National Family Health Surveys (NFHS) to produce consistent excellence data on health and sociodemographic issues of both women and men. India has NFHS-1 (1992–1993), NFHS-2 (1998–1999), NFHS-3 (2005–2006) and NFHS-4 (2015–2016) to support policymakers in the health sector.
Sampling and Data Collection
NFHS-4 had two stages of sampling techniques for the rural areas and three stages for the urban areas, using the 2011 population census. Using probability proportional to size (PPS), primary sampling units (PSUs) were used to select villages in rural areas. Households were then randomly selected from the PSUs [13]. Using PPS, municipality wards were selected as PSUs in the urban areas. Then, census enumeration blocks (CEB) were randomly selected from each PSU and households were randomly selected from the CEB.
From January 2015 to 4 December 2016, field interviews were conducted by 789 trained field teams, who collected the data from 28,522 clusters in India. Each field team had three female and one male interviewer, two health investigators and the driver under the field supervisor. Initially, the survey selected 628,900 household samples. Among the selected households, 616,346 had prospective respondents. Finally, the study included 601,509 households. All women of reproductive age (15–49 years) who lived the night before the interview day in those selected households were considered as the eligible sample. A total of 699,686 women (15–49 years) were interviewed (97% response rate) using the NFHS questionnaires. A more detailed description of the survey, including sampling, questionnaires, data collection and data handling, is available elsewhere [13].
Dependent Variable
Any maternal delivery at a woman's own home, parents' home or other home constituted the primary variable, called 'home delivery'.
Independent Variables
Individual and family level factors included: age (seven groups: 15–19, 20–24, 25–29, 30–34, 35–39, 40–44 and 45–49 years); residency (rural and urban); education (no education, primary, secondary and higher education); religion (Hindu, Muslim and others); economic status (poorest, poorer, middle, richer and richest); and sex of household head (female and male). The other variables were the husband's education (no education, primary, secondary and higher education), type of cooking fuel (solid or non-solid fuel), health insurance coverage (yes or no) and neighbourhood socioeconomic status (more disadvantaged and less disadvantaged).
Economic status was measured by the validated and widely used wealth index, a composite measure of the cumulative living standard of the household, introduced in India by Rutstein and Johnson [14, 15]. It primarily assesses the respondent’s ability to pay for healthcare facilities and the distribution of the health services among the poor. The wealth index includes ownership of household assets. Principal component analysis puts individual households on a continuous scale (standard normal distribution, mean = 0, SD = 1) of relative wealth. From the standardised scores, five different categories of wealth quintiles are estimated (poorest, poorer, middle, richer and richest).
Solid fuel includes wood, charcoal, straw, shrubs, grass, coal, ignited agricultural crops, and cow or buffalo dung. Non-solid fuel is electricity or natural/liquid petroleum gas, biogas or kerosene. Generally, using solid fuel indicates a low socioeconomic status [16].
Neighbourhood socioeconomic (NSE) status is widely used for reviewing the influence of neighbourhood socioeconomic status on health [15, 17, 18]. The NSE index was constructed to assess whether the respondent lived in a less or more disadvantaged socioeconomic neighbourhood. The NSE index included four variables: the proportion of rural respondents, proportion of respondents living below the poverty level, respondents living in slum areas and proportion of illiterate respondents. Using principal component analysis (PCA), the continuous scores were estimated to classify neighbourhoods into two categories: (1) more disadvantaged and (2) less disadvantaged socioeconomic neighbourhood status.
Economic and electronic empowerment factors included working status (working and non-working), employment status (employed year round, seasonal employment, occasional employment), having money that the respondent alone can decide how to spend (yes, no), having a bank account (yes, no), knowledge of a programme in the neighbourhood area that gives loans to women to start or expand a business (yes, no), owning a mobile phone (yes, no) and being able to use SMS (yes, no). Seasonal employment indicated a kind of temporary employment for specific seasons, mostly with some certainty, for example, during monsoon season employment in the paddy field. Occasional labour means great uncertainty about getting any employment when seasonal employment was not available.
Domestic control and violence factors included the experience of emotional violence (yes, no), experience of physical violence (yes, no) and experience of any sexual violence (yes, no). Controlling issues included whether the woman was usually allowed to go to the market and to visit a healthcare facility. Each question had three options: 'not at all', 'can go alone' and 'can go with someone else'.
Statistical Analysis
Chi-square tests were used to examine differences in proportions of exposure to IPV by demographic, socioeconomic and empowerment variables. Multivariate logistic regression analysis was performed with all demographic, socioeconomic and empowerment (including electronic) variables to assess their independent contribution in predicting exposure to IPV. IBM SPSS v25 was used for analysis. Statistical significance was considered at P < 0.05.
NFHS-4 used all necessary sampling techniques, emphasising consistency and comparability and ensuring the best quality of survey results [13]. The prevalence estimate of home delivery was estimated. For investigating the cross-relationship between home delivery and independent variables, we estimated proportions and conducted \(\chi 2\) tests including adjusted standardised residuals. Multivariate logistic regressions were estimated to determine the possible association between home delivery and independent variables.
Data analysis was done using IBM SPSS v 25.
Ethical Permission
The current study was conducted using secondary data from NFHS-4. NFHS-4 received ethical approval from the Institutional Ethical Review Board (ref. no./IRB/NFHS-4/01_1/2015) of the International Institute for Population Sciences (IIPS), Mumbai, India. No informed consent was required as this study used anonymised secondary data. However, the field staff and NFHS-4 had received informed consent from all participants.