Study design and data source
This was a cross-sectional study that utilized data from the UDHS 2016. Details of the study design, methods and sampling strategy are described in the UDHS 2016 final report . In summary, the sampling frame for this survey was adopted from the Uganda National Population and Housing Census (NPHC) conducted in 2014. The sample for the survey was selected in two stages. In the first stage, 697 enumeration areas (EAs) were selected from the 2014 NPHC frame: 162 EAs in urban areas and 535 in rural areas. In the second stage, a listing of all households in selected EAs was obtained. Through a systematic random sampling strategy, households were selected for the survey. All women aged 15–49 years who were either permanent residents of the selected households or visitors who stayed in the household the night before the survey were eligible to be interviewed. Data for the survey were collected from 20 June to 16 December, 2016. The data collection tools were translated into eight major languages, namely Ateso, Ngakarimojong, Lugbara, Luganda, Luo, Runyoro-Rutoro, Lusoga, and Runyankole-Rukiga. Of 19,088 eligible women, 18,506 were successfully interviewed, representing a 97% response rate. To limit the effect of recall bias, and in keeping with the Roll Back Malaria indicator on IPTp-SP, the sample was restricted to women who had had a live birth within the 2 years preceding the survey. Therefore, 5901 women were included in this analysis.
The outcome variable was uptake of IPTp-SP and was categorized as taking less than three doses or at least three doses as recommended by the World Health Organization (WHO) .
Data on socio-demographic characteristics, including age, marital status, education status, area and region of residence, and wealth index were abstracted. Other variables considered for analysis were parity, exposure to radio messages about health, number of ANC attendances as well as timing of first attendance, and source of anti-malarials during last pregnancy. Age of participants was categorized into 15–24, 25–34 and older than 34 years. Marital status was categorized as never married, married and separated/widowed/divorced. Education status was categorized as no education, primary education, secondary education, and higher education. Participants’ area of residence was either urban or rural. Wealth index scores were derived using the principal component analysis. Households were given scores based on the number and kind of consumer goods they own, including a television set, bicycle or car, and housing characteristics such as a source of drinking water, toilet facilities and flooring material. The computed wealth scores were then divided into five equal categories called quintiles. The lowest quintile comprised the poorest and the highest quintile, the richest. Ugandawas divided into 12 different regions, namely Kampala, Central, Busoga, Bukedi, Teso, Karamoja, Bunyoro/Toro, West Nile, Acholi, Lango, Bugishu, and Ankole/Kigezi. Parity defined as the participants’ number of live births was categorized into 1 child, 2 children and ≥ 3 children. Exposure to radio messages about health was categorized as not at all, less than once a week, and at least once a week.
Data on attendance of ANC were also abstracted, and were assessed using three items, namely number of ANC visits for most recent pregnancy, timing of first ANC visit, and source of anti-malarials during pregnancy. The number of ANC visits was categorized into adequate (≥ 4 visits) and inadequate (< 4 visits) as per WHO recommendation . Timing of first ANC attendance was categorized into first trimester, second trimester and third trimester. Source of anti-malarials was either at ANC, at another facility visit or other sources.
Analyses were conducted using STATA 15 (StataCorp 2017) utilizing the svyset command to match the multistage cluster sampling design method. Distribution of socio-demographic characteristics by IPTp-SP uptake was assessed using the Chi square test. Differences with p-values < 0.05 were considered significant (2-tailed). Both bivariate and multivariable logistic regression analysis were conducted to obtain crude odds ratios (cOR) and adjusted odds ratios (aOR) respectively and their 95% confidence intervals (95% CI). Predictors that were statistically significant (p < 0.05) in the Chi square test were entered in the multivariable logistic regression model. All analyses considered the complex sample design. Sample weighting was used to adjust for the cluster sampling design and sampling probabilities across clusters and strata.
The survey protocol was reviewed and approved by the Inner City Fund International Inc., Fairfax, VA, USA (ICF) Institutional Review Board. To access the survey data, permission from the Demographic and Health Survey Program of the US Agency for International Development was sought and obtained. The data received did not have personal identifiers. In addition, clearance was obtained from the Centers for Disease Control and Prevention, Atlanta, Georgia, USA (CDC), Associate Director for Science, to conduct this study as non-research.