Anaemia in pregnancy among Aboriginal and Torres Strait Islander women of Far North Queensland: A retrospective cohort study

Aim: Anaemia during pregnancy is common worldwide. In Australia between 7.1% and 11% of mothers have been reported to have anaemia in pregnancy. Higher rates are reported for Aboriginal and Torres Strait Islander women (Townsville: 34.2%, remote Northern Territory: 50%). The present study describes anaemia in pregnancy among Aboriginal and Torres Strait Islander women of Far North Queensland. Methods: Health service information was analysed for 2076 Aboriginal and Torres Strait Islander women who gave birth between 2006 and 2010. The prevalence of anaemia in pregnancy, characteristics of the mothers and pregnancy outcomes were described. Logistic regression for bivariate analyses and multivariable linear modelling with and without imputed data were used to compare those mothers who had anaemia in pregnancy with those who did not. Results: More than half of Aboriginal and Torres Strait Islander women (54.5% (95% CI: 52.4%, 56.7%)) had anaemia in pregnancy. For mothers who gave birth in 2009 and 2010 (n = 1796) with more complete data, those who were iron de ﬁ cient during pregnancy were more likely to be anaemic (RR: 1.40, P = <0.001). Mothers (29.0%) from localities of relative socioeconomic advantage had lower risk of anaemia in pregnancy (RR: 0.86, P = 0.003), as did mothers (31.9%) who were obese (RR: 0.87, P = 0.013). Conclusions: The prevalence of anaemia in pregnancy among Aboriginal and Torres Strait Islander women of Far North Queensland is high. Prevention and treatment of anaemia will improve the health of these mothers, and possi-bly the health and early development of their children.


Introduction
The 'First Thousand Days' from conception to around age 2 years is a time of rapid growth and neurological development. 1,2 Anaemia in pregnancy-defined as blood haemoglobin levels below 110 g/L-is a concern because of poorer health and pregnancy outcomes of mothers, and also the potential detrimental effects on the health and development of their children. 1,3 Infections, inflammation and genetic conditions (e.g. thalassaemia) can cause anaemia, as well as iron deficiency and other nutritional deficiencies. 3 Although an essential nutrient, iron can have negative metabolic effects. 4 To prevent damage, iron absorption is tightly regulated but increases when iron requirements are high, as in pregnancy. 4 The pregnant mother requires iron not only for her immediate demands-increased blood, tissue growth-but to provide iron stores to her baby. 5 The main source of iron for a baby in the first months of life is not breast milk or infant formula but these iron stores acquired before birth. 5 The high maternal iron requirements mean that anaemia in pregnancy is usually due to iron deficiency and is a strong predictor of early onset anaemia in the child. 3,6 High rates of early childhood anaemia are a continuing concern in remote Aboriginal and Torres Strait Islander communities of northern Australia. 7,8 Australian studies reporting anaemia in pregnancy include studies from South Australia (unsupplemented control group: 11% anaemic, births 1997-1999, including 3.3% Aboriginal mothers), Western Australia (6.2% anaemic >18 years, births 2005-2006, 7.3% Aboriginal and Torres Strait Islander mothers) and for all births in South Australia 1999-2005 (7.1% anaemic, 2.5% Aboriginal and Torres Strait Islander mothers). [9][10][11] Higher rates of anaemia in pregnancy have been reported for Aboriginal and Torres Strait Islander women accessing antenatal care in Townsville (34.2% anaemic, births 2001-2003) and in two remote Northern Territory communities (50.0% anaemic, births [2004][2005][2006]. 12,13 Anecdotal reports by health service providers in Far North Queensland (Figure 1) indicate that anaemia among Aboriginal and Torres Strait Islander mothers and their children is also prevalent but published information is lacking. Consequently, research has been undertaken to investigate anaemia among Aboriginal and Torres Strait Islander mothers and their children in Far North Queensland from an intergenerational perspective. Here we describe anaemia in pregnancy, and investigate associations between anaemia and various maternal characteristics, health indicators, and pregnancy outcomes of these mothers for a pregnancy and birth between 2006 and 2010.

Methods
This is a retrospective cohort study using linked information extracted from three existing health service data collections for mothers resident in Far North Queensland, in respect of pregnancies and births of their babies born between 2006 and 2010.
Data sources: Electronic data systems used by health service providers store confidential client information with strict provisions for data security and confidentiality. However de-identified information may be released for research purposes, subject to stringent processes to ensure data security and confidentiality. The process of securing the necessary approvals and release of a linked, de-identified dataset has been described elsewhere. 14 Briefly, data collections accessed were the Queensland Perinatal Data Collection (PDC); the Queensland Health Pathology Services Data Collection (Auslab); and the community health services electronic record system, Ferret, used mainly in remote locations of Far North Queensland (Supporting information  Table S1). Information extracted from Auslab had been recorded from 2000 up to 2010, from Ferret from date of rollout (see Figure 1), up to 2010 and from PDC from 2006 to 2010. Individual records were linked and deidentified by the Queensland Health Statistical Services Branch for release to the research group in May 2017.
Participants: Study data were extracted from these data collections for two cohorts-the Cape York cohort and the 2009-2010 cohort.
The Cape York cohort includes mothers of Aboriginal and Torres Strait Islander children born between 2006 and 2008, where the child had previously been included in an unpublished health service review of childhood growth in remote Cape York communities.  Variables and definitions: Anaemia in pregnancy was defined as haemoglobin less than 110 g/L as used by Queensland Health. 15,16 Measurements of haemoglobin used here are results of pathology laboratory measurements. Iron deficiency was defined as Ferritin levels below 15 ug/L. 15 Information recorded on the PDC includes mothers' ethnicity, parity, pre-pregnancy weight, height, smoking in pregnancy; birth status of babies-live/still born, gestational age at birth. Other information was derived from PDC records (maternal age, teenage mothers, body mass index categories, prematurity and birthweight category) as defined by the Australian Institute of Health and Welfare and the National Health and Medical Research Council-see Table S2. [17][18][19] Pre-existing diabetes was defined as a fasting oral glucose tolerance test result ≥7.0 mmol/L and/or a glycated haemoglobin reading ≥6.5%. Gestational diabetes was defined as an oral glucose tolerance test result ≥5.1 (fasting) and/or 10.0 (1 hour) and/or 8.5 mmol/L (2 hours) among women without pre-existing diabetes. 16,20,21 For definitions of hypertension, iron deficiency, low red cell folate (RCF) and vitamin B12 levels, see Table S2. Information on food insecurity, diet or nutrient supplements are not recorded in these electronic data collections.
The Socio-Economic Index for Areas (SEIFA 2011) ranks Australian Bureau of Statistics Statistical Local Areas (SLAs) by deciles of relative socioeconomic advantage and disadvantage. 22 A ranking of '1' indicates a locality of greatest relative disadvantage while a ranking of '10' indicates a locality of greatest relative advantage. 22 The appropriate SEIFA decile ranking was allocated to each mother based on her usual place of residence. For the purpose of this analysis, SEIFA deciles (1-10) were reduced to two categories: SEIFA deciles 1 and 2 (the 20% most disadvantaged SLAs in Australia) or SEIFA decile 3 or higher. These categories were selected as most of the mothers in this study (71.1%) lived in SEIFA categories 1 and 2.
Statistical analysis: Categorical variables were described using absolute and relative frequencies. The distribution of numerical variables were assessed; symmetrically distributed numerical characteristics were described using mean values, SDs and ranges; numerical values with a skewed distribution (parity, Ferritin levels, baby's gestational age at birth) were described using median, interquartile ranges (IQRs) and ranges. The prevalence of anaemia was presented with 95% confidence intervals (95% CIs).
Bivariate analysis: Characteristics of the mothers and their pregnancy outcomes were compared between those mothers who had been anaemic in pregnancy and those who had not, using logistic regression.
Multivariable analysis: The following characteristics were considered during multivariable analyses (Cohort 1 '2009-2010 cohort' n = 1796; Cohort 2 'Cape York cohort' n = 280). Variables with complete dataset were ethnicity of mother, age of mother, SEIFA category for residence of mother, five or more antenatal care visits, pregnancy induced hypertension, birthweight of baby (Cohort 2: complete dataset). Variables with missing values were: BMI category of mother, parity, smoking during pregnancy, mother with preexisting diabetes, gestational diabetes, low RCF value before or during pregnancy, low vitamin B12 value before or during pregnancy, iron deficiency during pregnancy, birthweight of baby (missing values Cohort 1). The number of missing values for variables used in multivariable analyses is shown in Tables 1-2 and Tables S1-S4. 'Missing-ness': Examination of patterns of missing data showed data missing for some key variables; year of birth of baby (that is, the cohort) was significantly associated with missing body mass index (P < 0.001), missing parity (P = 0.042), and missing iron status (P < 0.001), resulting in more missing data for the Cape York cohort mothers. Consequently it was decided to conduct analysis stratified by cohort. Tables S3 and S4 provide more information on missing values and patterns of 'missing-ness'.
Multivariable general linear models for the binomial family using the log link to estimate relative risks (RRs) were used to identify independent risk factors for anaemia during pregnancy for the complete case analysis. Backward and forward stepwise modelling procedures were initially conducted to establish basic multivariable models for both cohorts. Characteristics that were not part of the basic models were assessed for potential confounding effects. A confounder was assumed to be a variable that changed estimates of characteristics in the basic model by 10% or more. 23 Once a model was established, all possible two-way interactions involving variables in the model were assessed for statistical significance.
Multiple imputation: Multivariate multiple imputation was conducted using Stata's MI commands for sequential imputation using chained equations. Missing values were imputed for BMI of mother; parity; smoking during pregnancy; mother with pre-existing or gestational diabetes; iron deficiency during pregnancy; and birthweight of baby. Low RCF and vitamin B12 values before or during pregnancy were not imputed because these characteristics were missing in close to 80% of cases in both cohorts and they did not show statistically significant associations during bivariate or multivariable complete case analyses. Before imputation, patterns of missing values were investigated and assumed to be 'missing at random' in each cohort. 24 Linear regression was used to impute missing values of continuous characteristics; logistic regression was used to impute missing values of dichotomous characteristics. Imputation models were based on the following variables: anaemia during pregnancy, pregnancy induced hypertension, ethnicity, age, SEIFA index and antenatal care received. Twenty imputed datasets were created for each cohort. Multivariable general linear models for the binomial family using the log link to estimate RRs were used to identify independent risk factors for anaemia during pregnancy for imputed data.
Results of multivariable models for complete case and imputed data analyses are presented as RRs and 95% CIs.
P-values of less than 0.05 were considered statistically significant. Analysis was conducted using Stata version 13 (Sta-taCorp, Lakeway Drive, College Station, Texas).

Results
Data provided in May 2017 included information for 2332 mothers who gave birth to 2548 Aboriginal and Torres Exclusions: For the purpose of this report, non-Indigenous mothers (n = 15) and mothers normally resident outside of Far North Queensland (n = 29) were excluded. Births that were not the first birth in the cohort years were excluded (n = 289). Mothers with missing information for haemoglobin levels during pregnancy (n = 119) were also excluded.    Both models were adjusted for the confounding effect of age of mother (no missing values imputed). Imputed data are averages of 20 imputations. * P-value less than 0.05.
The mean age of mothers was 25.2 years (SD = 6.4) ranging from 13 up to 48 years. One in five mothers (21.0%) were teenagers. Median parity was two, ranging from 0 to 16. Most mothers (78.9%) had at least five antenatal healthcare visits in pregnancy. More than half (57.1%) smoked in pregnancy. Mean body mass index of the mothers aged 18 years and over (measurements available n = 1535) was 27.7 (6.6) ranging from 16.0 to 56 kg/m 2 .
Among mothers with glucose tolerance results, 6.1% (n = 1239) had pre-existing diabetes and 17.8% (n = 794) had gestational diabetes. PDC records showed that 5.1% had pregnancy-induced hypertension. Among mothers with prior records of blood pressure 17.6% (n = 812), had hypertension. More than half of the mothers with measures of Ferritin had iron deficiency during (59.3%, n = 1133) or before (57.8%, n = 561) the cohort pregnancy.
The Cape York cohort: The majority (88.9%) of these mothers (n = 280) were Aboriginal and the remaining were Torres Strait Islander (5.7%) or both Aboriginal and Torres Strait Islander (5.4%) ( Table 2). Nearly all (93.2%) were usually resident in Cape York, 6.4% in Cairns and Hinterland Health Service District and one mother in the Torres Strait and Northern Peninsula Area (Figure 1). Most (91.1%) mothers lived in localities with a SEIFA ranking in the lowest or second lowest decile. 22 The mean age of these mothers was 25.0 years (SD = 6.4) ranging from 15 to 40 years. One in five (20.4%) were teenagers. Median parity was two, ranging from nil to eight.  Where glucose tolerance results were available 4.7% (n = 212) had pre-existing diabetes and 20.0% (n = 140) had gestational diabetes. PDC records showed that 7.5% had pregnancy-induced hypertension. Among mothers with prior records of blood pressure, 20.6% (n = 272) had hypertension. Of those with measures of Ferritin, 39.7% (n = 54) had iron deficiency during the cohort pregnancy while 48.1% (n = 77) had iron deficiency before the cohort pregnancy.
Cape York cohort-Pregnancy outcomes: Among the 283 babies born to these 280 mothers, 51.9% were boys (Table S8). No information was available for perinatal mortality for this cohort. Median gestational age was 39 weeks, ranging from 27 to 42 weeks. Mean birthweight was 3097 g (SD = 591 g) ranging from 800 to 5320 g. About one in seven babies (13.8%) were low birthweight, 11.7% premature and some (4.6%) macrosomic (birthweight ≥ 4000 g).
Cape York cohort: After controlling for existing or gestational diabetes, only body mass index remained statistically significant for these mothers (being obese: RR = 0.40, P = 0.019) ( Table 4; imputed data analysis).

Discussion
Among the Aboriginal and Torres Strait Islander mothers of Far North Queensland described here, over half had anaemia in pregnancy (54.5%, n = 2076). This is much higher than among pregnant women elsewhere in Australia but similar to findings from two remote Northern Territory communities, where 50% of mothers had anaemia in pregnancy. 13 These results reflect the higher rates of anaemia reported among Australian Indigenous people in recent national health surveys. 25 Although other conditions can cause anaemia, iron deficiency is the 'usual suspect' as the cause of anaemia in pregnancy. 15,26 Among the 2009 and 2010 birth mothers, for whom data were more complete, analysis confirmed that iron deficiency was strongly associated with anaemia in pregnancy.
Compared to the average Australian mother in 2016, the mothers described here were younger, more likely to smoke and they had less antenatal care. 27 The prevalence of obesity was higher among the 2009-2010 birth mothers, although not the Cape York mothers. About one in four of these Far North Queensland mothers had diabetes in pregnancy compared to about one in eight mothers Australiawide in 2016. 27 Previous reports of the poor health and nutrition of young Indigenous women in North Queensland also flagged the potentially detrimental intergenerational effects. 28 The pregnancy outcomes reported here, with more premature and low birthweight babies, reflect the poor nutrition and health status of these mothers. The association of anaemia of mothers with increased birth weight of the babies may reflect their marginal nutrition status, with the requirements of bearing a healthy weight baby depleting the limited nutritional reserves of these mothers.
An unexpected finding was that obese mothers were less likely to be anaemic than other mothers, though their rates of anaemia (46.1%) were still high. It may be that the higher food intake resulting in obesity provides a higher nutrient intake. However, maternal obesity is never recommended because of the negative health effects on the mother and her baby. 28 Instead, mothers need diets sufficiently nutritious to meet their requirements without excessive energy intakes.
There are limitations in this study, which used health service information recorded during provision of routine care. Some information of interest is not recorded on electronic data collections such as indicators of food insecurity or nutrient supplement use. Missing data for some key variables particularly in the early years was a limitation. Some measurements used in this analysis may have been available only for selected mothers if clinical protocols prescribed specific pathology tests for 'high risk' mothers. Examples include measurements of glucose tolerance and Ferritin levels. Because 'missing at random' is an assumption for multiple imputation where 'missing not at random' was suspected, respective characteristics were carefully analysed in alternative models.
Despite these limitations, the findings reported here are consistent with the high rates of anaemia among young Aboriginal and Torres Strait Islander children and pregnant women reported elsewhere in remote Australia. 7,13 A recent review has shown that food insecurity is associated with increased risk of anaemia among women and young children in high and low income countries. 29 Food insecurity and poor diet of Aboriginal and Torres Strait Islander people have been documented in remote settings in Australia. 30,31 There is increasing evidence of the importance of a nutrient-rich diet in pregnancy, not only for the mother but for the physical health and cognitive development of her child. 1,32 A nutrient-rich diet helps prevents anaemia and provides many nutrients needed for health. For mothers on low incomes, high cost is a barrier to healthy eating especially in remote settings. 33 Mothers with low iron status in pregnancy can benefit from supplements-in addition to a healthy diet-but reaching those with highest needs is challenging. 34 Fortification of flour with folate appears to have been especially effective at reaching vulnerable population groups. 35 But iron is a nutrient with potential for harm so iron supplementation must be targeted to those with specific needs. 4 Complementary interventions to improve food security, promote good nutrition, and provide targeted iron supplementation and/or fortification are needed, designed and implemented in partnership with the Aboriginal and Torres Strait Islander communities. 36 Also essential are the policy commitment and funding to develop, implement and evaluate these interventions. 37 To 'Close the Gap' 38 in the health, education and economic status of Aboriginal and Torres Strait Islander people in Far North Queensland compared to their non-Indigenous peers, these high rates of anaemia in pregnancy among Aboriginal and Torres Strait Islander mothers must be reduced.

Funding source
DL was supported by a National Health and Medical Research Council post-graduate scholarship APP1092732. Other agencies provided non-financial support to this research. We would like to acknowledge and thank the Aboriginal and Torres Strait Islander leaders of the key community controlled health service organisations in Far North Queensland who considered and endorsed the proposed research, providing the support that made this research possible. In addition, we acknowledge and thank the Queensland Health Data Custodians and their research and data management staff for their assistance and support for this work.

Supporting information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site: Table S1 Data collections used as information sources to describe rates of anaemia in pregnancy and health indicators among Aboriginal and Torres Strait Islander mothers in Far North Queensland  Cape York mothers -comparing mothers with anaemia during the cohort pregnancy with mothers who did not have anaemia during this pregnancy: Bivariate analysislogistic regression Table S7 Pregnancy outcomes -2009-2010 births: comparing those babies whose mother had anaemia in pregnancy with those babies whose mother did not have anaemia in pregnancy -results based on 1812 babies of whom 1795 were live born babies: Bivariate analysislogistic regression Table S8 Pregnancy outcomes -Cape York births: comparing those babies whose mother had anaemia in pregnancy with those babies whose mother did not have anaemia in pregnancy (n = 283 -all babies live-born): Bivariate analysis logistic regression