Data for this study were collected as part of a survey conducted to provide data on key maternal, newborn, child and reproductive health indicators in Sindh province in 2013.
Study setting
Approximately 44 million people live in Sindh province, which includes the five districts of Karachi as well as 22 predominantly rural districts with limited health infrastructure. The 2012–13 Pakistan Demographic and Health Survey found that the neonatal mortality rate in Sindh was 54 deaths per 1,000 live births, the infant mortality rate was 74 per 1,000 live births, and the under-five mortality rate was 93 deaths per 1,000 live births. With a total fertility rate (TFR) of 3.9, Sindh had a TFR higher than the national average of 3.8. In Pakistan, maternal and child health services are provided by a mixed health care delivery system dominated by the private sector, complemented in rural areas by a network of public sector community health workers, including Lady Health Workers and community midwives [11].
Sampling
The study covered all 27 districts of Sindh. It used a multi-stage stratified sampling design to select 4,000 women who had had a live birth in the two years before the survey [23]. The survey oversampled rural districts of Sindh. Probability proportionate to size sampling was used to select the required number of villages (rural) and city sections (urban) in each selected district. Random sampling was used to identify households within each primary sampling unit (PSU). A total of 494 PSUs were selected for the survey, with 8 interviews being conducted per PSU. Detailed information about the survey methodology is available online [11]. The study provides a representative sample of Sindh.
Data collection
Data were collected in June and July 2013 through face-to-face interviews with female participants at their homes. A structured questionnaire based on the 2012–13 Demographic and Health Survey instrument for Pakistan developed by Macro International, Inc. was used for the survey [11]. Women were asked to provide basic information such as age, parity and educational status, along with information about the most recent pregnancy and health care received before, during and after delivery. Studies have shown that mothers’ recall of pregnancy-related events is valid and reliable over long periods of time [24]. The questionnaire was translated into Urdu and Sindhi, and data collected by female staff of a survey research firm that has been conducting national-level household surveys in Pakistan for more than 10 years. Verbal consent was obtained from all participants prior to conducting the interview. This study was approved by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board and the National Bioethics Committee of Pakistan.
Data analysis
Outcome variable
The outcome variable of interest for this study is the content of care provided to a woman during ANC visits. Women were asked about the services they were provided during ANC check-ups for their most recent pregnancy. These services included blood pressure measurement, blood sample testing, urine sample testing, body weight measurement, whether they were given iron tablets and whether they received tetanus immunizations. An outcome variable, comprising of a simple count of the number of elements of care received, was created. The variable had a minimum value of zero and a maximum value of six.
Independent variable
The key independent variable of interest for this study was the timing of the first ANC check-up. Women were asked how many months pregnant they were when they first received antenatal care for their most recent pregnancy which occurred during the last two years.
Control variables
Other independent variables were conceptualized as control variables. Several studies have found important effects of socio-demographic factors on maternal health services utilization. Income and wealth have been found to be powerful determinants of the utilization of maternal health services [25]. Maternal education is thought to influence the utilization of maternal health services through multiple mechanisms: education may increase awareness of the benefits of preventive health; education may also increase a woman’s confidence in dealing with health care providers [26]. Older ages at marriage, a reflection of greater autonomy of women, may be associated with higher use of maternal health services [27]. Although studies have rarely examined the determinants of early initiation of ANC in developing countries, it is likely that socio-demographic factors play some role in early initiation of ANC. Hence, a range of socio-demographic variables are included as control variables.
A binary variable was created measuring urban or rural residence. Categorical variables were created for woman’s age (15–24, 25–34, 35–49), woman’s age at marriage (12–15, 16–20, 21 or older), the number of children (one, two, three, four, five or more), the highest level of school attended (no formal education, primary or middle school, secondary or higher education). Principal component analysis was used to create a variable measuring household wealth. The data used to create this variable were based on the following assets and amenities: ownership of mobile phone, motorcycle, television, refrigerator, cupboard/cabinet, washing machine, bed, clock, sofa, sewing machine, livestock, construction material use for the floor, construction material use for the floor roof, construction material use for the wall, main fuel used by the household, whether the household has a water pump and whether the household has a toilet. This is approximately similar approach to the one used widely by the Demographic Health Surveys [28].
Statistical analysis
Data analysis was conducted using Stata 12.1 (StataCorp, College Station, TX.). Weights were attached to the data to adjust for oversampling of rural areas. Bivariate analysis included constructing a Kaplan-Meier survival curve to estimate the time to first ANC check-up. Bivariate analysis examined the relationship between the timing of the first ANC check-up and the content of care provided to women. Bivariate analysis also examined the relationships between socio-demographic variables and whether the first ANC check-up occurred within 3 months of pregnancy.
Ordinary least squares (linear) regression was used to determine whether the timing of the first ANC check-up was associated with a woman receiving higher quality of care during her pregnancy, controlling for socio-demographic factors and the number of ANC visits made during the pregnancy. The SVY suite of commands in STATA was used to conduct weighted analysis and control for clustering of observations at the PSU level. We tested for endogeneity, to see if unmeasured factors (such as a woman’s motivation) may be responsible both for earlier initiation of ANC and for seeking better quality of services. If unmeasured factors are responsible for both the seeking of better quality of services and for earlier initiation of ANC, biased estimates may be produced [29]. However, our analysis found no evidence of endogeneity (not shown).