- Top of page
- Materials and methods
Anaemia is a widespread public health problem that affects 25% of the global population (WHO 2008). It reduces physical work capacity (Haas & Brownlie 2001) and is a major contributor to death and disability worldwide (Ezzati et al. 2004). Women of childbearing age are at an increased risk of anaemia: the global prevalence is 42% in pregnant women and 30% in non-pregnant women of reproductive age. The prevalence varies by geographical region and is highest in Africa, where 57% of pregnant and 48% of non-pregnant women are anaemic (WHO 2008). Anaemia in women is associated with maternal mortality, preterm delivery, stillbirths and low birth weight (Allen 2000; Kidanto et al. 2009; Zhang et al. 2009; Ali et al. 2011). Additionally, maternal anaemia increases the risk of iron deficiency in infants (Kilbride et al. 1999; Kalaivani 2009).
Anaemia is a major public health problem in Tanzania, and understanding more about its determinants will inform how best to direct new and existing prevention measures (WHO 2008) and contribute towards achieving the millennium development goal 5 of reducing maternal mortality (United Nations 2006). However, previous studies conducted in Tanzania to assess determinants of anaemia among women have found conflicting results, and most have been carried out in geographically restricted regions of the country or among subgroups of women, and hence are not generalizable to the entire country (Matteelli et al. 1994; Shulman et al. 1996; Hinderaker et al. 2001, 2002; Marchant et al. 2002b; Massawe et al. 2002b; Msuya et al. 2011; Saathoff et al. 2011; Finkelstein et al. 2012). Low income, older maternal age, hookworm infestation, malaria infection and micronutrient deficiencies are some of the risk factors for anaemia among pregnant women in Tanzania (Antelman et al. 2000; Hinderaker et al. 2001; Marchant et al. 2002b; Massawe et al. 2002b; Msuya et al. 2011; Finkelstein et al. 2012). Additionally, Iron deficiency has been shown to be a risk factor for anaemia in adolescent girls (Massawe et al. 2002a).
Also, although some factors are known to increase anaemia risk, other potential risk factors, such as type of cooking fuel and body mass index (BMI), have been less well investigated. Information on the modifying effects of education and wealth, and contextual factors such as geographical area and place of residence on the impact of key risk factors is lacking. A better understanding of the relationship between different risk factors will allow identification of high-risk groups and facilitate improved targeted prevention.
Iron supplementation for pregnant women is a key national strategy for anaemia prevention in Tanzania, but coverage varies by education, wealth and place of residence (NBS & ICF Macro 2011). The effect of pregnancy on anaemia may therefore vary by these factors. The effect of education on anaemia may be weaker among the wealthiest women, who are more likely to have better access to iron-rich foods such as meat and eggs irrespective of their education (NBS & ICF Macro 2011).
We therefore sought to identify determinants of moderate-to-severe anaemia and to explore a priori interactions between key risk factors, among women of reproductive age in Tanzania using a recent large national health survey.
- Top of page
- Materials and methods
Moderate-to-severe anaemia among Tanzanian women of reproductive age was associated with pregnancy, education, geographical zone, regional prevalence of iron supplementation, use of contraceptives and type of toilet facility. The effect of pregnancy on anaemia varied significantly by place of residence, wealth and education level. Pregnancy increased the risk of anaemia to a greater extent among non-educated women. The effect of education on anaemia was also modified by wealth and pregnancy status, with education reducing the risk of anaemia more among the poorest women, and among pregnant women. Women without a toilet were at an increased risk of anaemia. Among non-pregnant women, hormonal contraception use was associated with a decreased risk of anaemia.
Our finding that education is associated with a decreased risk of anaemia is in keeping with the results of other studies which observed similar associations between education and moderate-to-severe anaemia (Ngnie-Teta et al. 2008; Ghosh 2009). In our study, the effect of education differed according to wealth and was particularly strong and significant among the poorest women. This could be explained by educated women being more likely to have better access to health information (Tanzania Commission for AIDS (TACAIDS), Zanzibar AIDS Commission (ZAC), National Bureau of Statistics (NBS), Office of the Chief Government Statistician (OCGS) & ICF International 2013) and to have better healthcare seeking behaviour than non-educated women (Ghosh et al. 2013). Our results suggest that educating women may help reduce the risk of anaemia, particularly among those in the lowest wealth strata.
Our finding that the risk of anaemia does not vary by urban/rural area of residence is consistent with that of a study conducted in Mali which similarly found no association between area of residence and moderate-to-severe anaemia (Ngnie-Teta et al. 2008). In contrast, a study conducted in India found an increased risk of anaemia among rural women, especially poorer rural women (Bentley & Griffiths 2003).
Pregnancy is an established risk factor for anaemia (Balarajan et al. 2011). The strength of association in our study is lower than that reported from analysis of the Mali Demographic and Health Survey data (Ngnie-Teta et al. 2008). This is probably because, unlike in the Mali study, we used different cut-off points for defining anaemia in pregnant and non-pregnant women. Pregnant women usually receive routine doses of iron supplements to prevent anaemia and coverage of iron supplementation varies throughout Tanzania. By showing that the risk of anaemia fell with increasing regional iron supplementation coverage, our study provides further evidence on the protective effect of iron supplementation on anaemia (Yakoob & Bhutta 2011). The variation in the effect of pregnancy on anaemia by wealth quintile could be due to differences in individual use of iron supplements, perhaps as a result of differences in healthcare access and health-seeking behaviour (NBS & ICF Macro 2011). Education may be associated with reduced anaemia risk in pregnant women but in not non-pregnant women because iron supplements are usually distributed during pregnancy and educated women are more likely to access the supplements than the uneducated (NBS & ICF Macro 2011).
Use of a copper IUD has been associated with increased risk of anaemia (Kivijarvi et al. 1986; Hassan et al. 1999b) through the side effect of heavy menstrual blood loss (Hassan et al. 1999a). However, some hormone-coated IUDs may reduce excessive menstrual blood loss and may be more appropriate to use, to limit risk of anaemia. We did not find an increased risk of anaemia among IUD users, although this was based on small numbers. Further studies that are adequately powered to investigate the effect of IUDs on anaemia are needed, as if an association does exist, women using IUDs may benefit from iron supplementation.
As expected, we found that hormonal contraception use was associated with a reduced risk of anaemia. Prevention of iron deficiency anaemia is known to be one of the non-contraceptive benefits of oral contraceptives (Mishell 1993).
The type of toilet facility is a measure of the general level of sanitation. In Tanzania, 15.9% of households have no toilet (NBS & ICF Macro 2011), and schistosomiasis and soil-transmitted helminths are found countrywide. In a study in northern Tanzania, 65.3% of pregnant women were infected with S. mansoni and 56.3% with hookworms (Ajanga et al. 2006). Women in households without toilets are at increased risk of infection by hookworms and other parasites, which may explain our findings that lack of toilet facilities increases anaemia risk. Hookworms increase the risk of anaemia in poor communities where sanitation standards are low (Brooker et al. 2008; Finkelstein et al. 2012), and treatment of S. mansoni or hookworm infections with chemotherapy improved Hb levels among anaemic children in Uganda (Koukounari et al. 2006). Parasite control through deworming and improved sanitation might therefore reduce the burden of anaemia.
Surprisingly, we did not find that ITNs reduced anaemia risk, but our results are in agreement with those of other studies, including a multicountry study which found that use of ITNs was not associated with moderate-to-severe anaemia in most nationally representative household surveys (Ngnie-Teta et al. 2008; Florey 2012). In contrast, in a study of pregnant women, Marchant et al. (2002a) found that ITNs did reduce anaemia. However, this study used a different Hb threshold for anaemia, included pregnant women only, and was undertaken in the context of social marketing of ITNs. In the TDHS, women are asked whether they slept under an ITN the night before the survey, which may not accurately reflect the usual pattern of mosquito net usage.
Finally, although use of biofuels for cooking increases the risk of anaemia in children (Mishra & Retherford 2007), we did not find any association between type of cooking fuel and anaemia risk. This lack of association with biofuels in adult women could be explained by the body's compensatory response to chronic Hb reduction, whereby the kidneys produce erythropoietin which stimulates production of more red blood cells (Leifert 2008).
Our study has a number of strengths. Demographic and Health Surveys use standardised methodology for data collection and adopt strict quality assurance procedures in data management (Macro International 1996; Rutstein & Rojas 2006) resulting in nationally representative and high-quality data, which increases the reliability of our results. To our knowledge, this is the first study to explore risk factors for moderate-to-severe anaemia among Tanzanian women using nationally representative data. The representativeness of the data increases the generalizability of the results to the country (and to similar settings), provides a better understanding of the determinants of moderate-to-severe anaemia among Tanzanian women and improves the usefulness of the findings, especially in identifying at risk groups and developing targeted interventions. Our study population was large, which enabled us to assess multiple associations between anaemia and various risk factors and to examine possible interactions. Our study is novel in that we were able to examine the effect of individual, household and contextual factors on anaemia risk in Tanzania.
There are some limitations. It was not possible to control for the effect of some factors associated with anaemia, such as HIV and malaria, and micronutrient intake, which may have led to residual confounding. Data on HIV prevalence by zone are consistent with some, but not all, of the geographical differences in anaemia observed in our study (Tanzania Commission for AIDS (TACAIDS), Zanzibar AIDS Commission (ZAC), National Bureau of Statistics (NBS), Office of the Chief Government Statistician (OCGS) & ICF International 2013). Malaria prevalence among children (a proxy for prevalence in women) (Tanzania Commission for AIDS (TACAIDS), Zanzibar AIDS Commission (ZAC), National Bureau of Statistics (NBS), Office of the Chief Government Statistician (OCGS) & ICF International 2013) and the consumption pattern of iron-rich foods among women with children aged <3 years (NBS & ICF Macro 2011) also vary by zone, but do not appear to explain observed differences in anaemia. However, the consumption of vitamin A-rich foods is consistent with variations in anaemia risk in some zones (NBS & ICF Macro 2011). For instance, Zanzibar, which has the highest anaemia risk, has the lowest proportion of women who consumed vitamin A-rich foods. As the TDHS is a cross-sectional study, some observations may be the result of reverse causality. For instance, women who are already anaemic may tend to sleep under a bed net. Finally, as our reference group included women who are mildly anaemic, some of the observed associations might have been underestimated. However, there is evidence that well-known socio-demographic and environmental risk factors for anaemia, including pregnancy, are associated with moderate-to-severe, but not mild, anaemia among African women (Ngnie-Teta et al. 2008). This observation supports the suggestion by some researchers to lower the cut-off point for anaemia among black women to 10 g/dl (Johnson-Spear & Yip 1994), which is in line with our definition of anaemia among pregnant women.