Study population
In this hospital based cross-sectional observational study we utilize data from the Kilimanjaro Christian Medical College (KCMC). The hospital is one of four zonal referral hospitals in Tanzania. It serves the local community and receives medical referrals from the larger Kilimanjaro region (population ~10 million). Most patients are indigenous Chagga and Pare, although over 60 other ethnic groups including Sambaa, Sukuma, Rangi and Maasai are also represented [22].
Births occurring at KCMC after 1999 have been electronically recorded in a birth registry housed at KCMC. Methods for recruiting participants for this registry have been described in detail elsewhere [22]. Briefly, the birth registry is part of a collaboration between the University of Bergen, Norway and the KCMC, Tumaini University in Moshi Urban District of the Kilimanjaro region in Northern Tanzania. Women admitted for delivery are asked to bring a copy of their antenatal care (ANC) card, a medical record of visits during pregnancy. Within 24 hours of delivery and after maternal verbal consent, a registry midwife conducts a structured interview with the mother and abstracts information on antenatal visits for the medical record. The registry combines ANC data and pregnancy outcomes data.
From February 1st 1999 to December 31st 2008, 26,152 births were included in the birth registry. We excluded referrals due to pregnancy or labour complications (n = 3,582) as they generally do not receive antenatal care at KCMC. To ensure that each pregnancy was counted only once since our unit of analysis was the pregnancy and not the birth, we excluded second twins and second and third triplets (n = 851), leaving 21,889 pregnancies in the analyses sample. Each pregnancy was included as a separate observation for women who delivered more than once during the 10 year period. The overall registry study was approved by the KCMC Ethics Committee and the Ministry of Health in Tanzania. The protocol for the current analyses was approved by the Duke University Medical Center Institutional Review Board.
Description of variables
Folic acid and iron supplementation
Information on folic acid and iron was retrieved from the antenatal record where there are separate checkboxes for folic acid and iron supplementation. Information was further obtained by interview with responses “Yes” or “No” to the questions “Did you take any drugs during this pregnancy?” The questions allowed women to report intake of these supplements whether they were prescribed and issued at the clinic, which is a common practice, or purchased over the counter from somewhere else than the KCMC pharmacy. Neither supplementation dose nor timing of when supplementation began, were collected. Supplementation of folic acid and iron were treated as binary outcomes in all analyses.
Socio-demographic correlates of supplementation
Socio-demographic correlates of supplement use included marital status (married, single, divorced, widowed); maternal age in years (<18, 18–25, 26–35,>35); parity (nulliparous, parous); education level of the woman and her partner (none, primary (1–7 years), secondary (8–11 years), tertiary (12 or more years)); occupation in categories of professional denoting requiring training (teachers, doctors, nurses, midwives and other occupations normally requiring education beyond secondary school level), housewife (housework, care of offspring), farmer (plant or animal husbandry in family farm in addition to housework and caring for offspring), service (typically, ward attendant in hospital, cleaner, bus driver, shopkeeper and messenger), business (denotes shop owners or women who conduct other trades, with measurable incomes), and other (including students, hairdressers, or house-girls) and tribe of participant and her partner categorized as Chagga, Pare and other. To avoid many and small categories in the multiple regression analysis, we grouped some of the categories for maternal age, parental education and parental occupation.
Health indicators and healthcare utilization
Health indicators based on structured interview with the mother and with supplementary information from the antenatal record and standard protocols were episodes of malaria, anaemia, or infections. Infections were defined as any infectious condition other than schistosomiasis, tuberculosis, or malaria. “Other diseases before pregnancy” included diabetes, heart disease, epilepsy, liver disease, sickle cell, TB, and lung disease. “Other diseases during pregnancy” included gestational diabetes, tuberculosis, diabetes, hypertension, epilepsy, bleeding, jaundice, schistosomiasis, heart disease and lung disease. Health care utilization variables included number of prenatal care visits categorized into few or no visits (0–1), 2–3 visits, 4–8 visits (recommended number of visits in uncomplicated pregnancies [12], and 9–15 visits (more frequent than once a month), as well as knowledge of HIV infection status. The diagnosis of anaemia was based on haemoglobin levels below 8.5 grams per decilitre at any time during pregnancy.
Statistical analyses
We estimated the proportion of women who ever used folic acid (yes/no), iron (yes/no) or both supplements (yes/no) during the prenatal period (percentages and 95% confidence intervals). Contingency tables were employed to estimate crude odds ratios (ORs) with 95% confidence intervals for the association between each of the three dichotomous outcome variables, and socio-demographic, health services utilization and pre-pregnancy and pregnancy related morbidity. All factors associated with iron or folic acid supplementation in bivariate analyses (defined as p value ≤ 0.20) were included in a multiple logistic regression model to identify those independently associated with supplementation. Due to missing values, the number of observations in the multiple regression analysis was 21,027 (862 observations had missing values in one or more variables). All analytical tests were carried out using STATA/IC 10.1™ and SPSS 18.0 for Windows (SPSS Inc., Chicago, IL).