Prevalence of anemia and associated factors in children aged 659 months in Pernambuco , Northeastern Brazil

OBJECTIVE: To estimate the prevalence of anemia and to identify its associated factors in children aged 6-59 months. METHODS: This was a cross-sectional study on data from the Third Health and Nutrition Survey of the State of Pernambuco, Northeastern Brazil, with a representative sample of 1,403 children from urban and rural areas. Anemia was diagnosed by means of hemoglobin assays. Multivariate analysis was performed though a hierarchical model, using robust-variance Poisson regression to estimate the prevalence ratio as a function of the following variables: biological factors, morbidity, child nutritional state, socioeconomic factors, housing, sanitation and maternal factors. RESULTS: The weighted prevalence of anemia was 32.8% overall: 31.5% in urban areas and 36.6% in rural areas. In urban areas, anemia was signifi cantly associated with maternal education, consumer goods, number of children less than fi ve years old in the home, drinking water treatment, maternal age, maternal anemia and the child’s age. In rural areas, only maternal age and the child’s age were signifi cantly associated with anemia. CONCLUSIONS: The prevalence of anemia in children in Pernambuco was similar in urban and rural areas. The factors associated with anemia that are presented here should be taken into consideration in planning effective measures for its control. DESCRIPTORS: Infant. Child. Anemia, epidemiology. Risk Factors. Socioeconomic Factors. Cross-Sectional Studies.


INTRODUCTION
Diagnosing the health and nutritional situation of populations contributes towards developing strategies that aim to combat nutritional defi ciencies such as protein-energy malnutrition, iron defi ciency anemia and hypovitaminosis A.
Among nutritional defi ciencies, anemia affects 1.62 billion people worldwide, and preschool children are affected most, with prevalence of 47.4%.a In Brazil, anemia is increasing, unlike malnutrition and hypovitaminosis A. 5 Studies have indicated high prevalences of anemia among children under fi ve year of age in various regions of the country, reaching 46.9% in São Paulo (Southeastern Brazil]), 11 54% in Criciúma, Santa Catarina (Southern Brazil), 13 46.3% in Salvador, Bahia 2 and 40.9% in Pernambuco (both in Northeastern), 17 where the prevalence in rural areas (51.4%) was substantially greater than in urban areas (37.8%) in 1997. 17gh prevalences of anemia and its consequences for children's health, and especially for their growth and development, 22 have made anemia an important public health problem, given the diffi culty in implementing effective measures for controlling it.Its etiology involves many factors, such as socioeconomic, nutritional, biological, environmental and cultural characteristics, and the actions required encompass pertinent and relevant matters within the context of public health. 12,22us, investigation of determinants of anemia among children under the age of five years is backed by recommendations from the World Health Organization (WHO).a According to WHO, generation and maintenance of databases on anemia provide countries with a real instrument for developing control strategies for this problem.Therefore, the present study had the aims of estimating the prevalence of anemia among children aged six to 59 months and identifying associated factors.

METHODS
This study was developed using data from the Third Investigation on Health and Nutrition in the State of Pernambuco (III PESN/PE), b of 2006, which covered municipalities representing two geographical spaces in the state: urban (Metropolitan Region of Recife and urban areas in the interior of the state) and rural.
The sample calculation for each geographical area assumed an error of 3.7% and anemia prevalence of 40% for the state of Pernambuco, 17 with a confi dence interval of 95%.The whole population of children under the age of fi ve years in each geographical area was considered, c and 10% was added to the sample size to compensate for possible losses.This resulted in samples of 740 and 738 children aged six to 59 months in the urban and rural areas, respectively.The sample actually investigated was 717 children in the urban sector and 686 in the rural sector.The sampling process was similar to what was used in 1991, on the fi rst occasion that this investigation was conducted, d in order to ensure that the results would be comparable.The 18 municipalities investigated in I PESN/PE were selected by means of systematic sampling with probability proportional to the size of the population, thereby refl ecting the population distribution of the state.
In these previously selected municipalities, census tracts were drawn randomly and systematically: 20 in the urban area and 19 in the rural area, thereby ensuring that both areas would be representative.
A point on the edge of each census tract in the urban area was drawn and, from this point, households with children under the age of fi ve years were located in each block, until the intended number of 40 children per census tract had been completed.In the rural area, starting from a simple random draw to obtain a point on the edge of the census tract, the households closest to this initial mark were visited.
The data were gathered between May and October 2006, by means of forms to record the identifi cation of the household, socioeconomic characteristics, the children's characteristics, histories relating to pregnancy and morbidity, and anthropometric and clinicallaboratory records of the mothers and children.
The anthropometric assessment was done in accordance with the WHO recommendations.e Weight and height were measured twice, by different interviewers, and the mean value of the two measurements was recorded.The measurements were repeated if the two weights differed by more than 100 g or the two heights by more than 1 cm.All the children were weighed without shoes on and wearing a minimum of clothes, on a digital balance of capacity 150 kg and precision of 100 g (Plenna, model MEA-032000).Children under the age of two years were weighed together with the mother or other adult responsible for the child, and the fi nal weight was calculated by subtracting the respective adult's weight, and this was recorded on the form during the fi eldwork.The height of children of up to two years of age was measured in dorsal decubitus using an infantometer (CMS Weighing Equipment Ltd., Rollametre model, by Raven), with an amplitude of 100 cm and scale divisions of 0.1 cm.For children over the age of f Ministério da Saúde.Orientações para a coleta e análise de dados antropométricos em serviços de saúde: material preliminar da Coordenação Geral da Política de Alimentação e Nutrição/DAB/SAS/MS sobre a avaliação nutricional de crianças: norma técnica -SISVAN.Brasília, DF; 2008.two years, a portable stadiometer (Alturexata) made of a wooden column that could be disassembled was used.The scale was marked on both sides, with a fi eld of use from 35 cm to 213 cm and scale divisions of 0.1 cm.The children were measured in an upright position, without shoes, with their arms hanging down the sides of the body, and their heels, back and head touching the wooden column.
Venous blood samples were collected from the children aged six to 59 months and women aged 15 to 49 years on the day after the interview, in a place previously determined by the fi eld team.Hemoglobin assays were performed using the Hemocue equipment (Hemocue Limited, Sheffi eld, UK).
Anemia was diagnosed based on the WHO standards.a Individuals with hemoglobin levels lower than 11 g/dl (children) or 12 g/dl (women of reproductive age) were considered to be anemic.
The sample calculation was done separately for the urban and rural areas, and the proportionality of the population under the age of fi ve years in each area was not taken into consideration.Thus, the prevalence of anemia in Pernambuco was weighted according to the respective populations of the geographical areas.For this, the proportions of children under the age of fi ve years in each geographical area were summed, multiplied by their respective anemia prevalences and divided by the total number of children under the age of fi ve years in the state (based on the data in the 2000 census).c The biological characterization was done based on the children's ages and sex.Morbidity was characterized by the presence of diarrhea over the last 15 days.Birth weight and the indices of weight for age and height for age were used to assess the children's nutritional status.For both of these indices, the reference curves used were those of the Brazilian Ministry of Health, f as recommended by WHO. e The anthropometric assessment was done using the Anthro software, version 3.01 (WHO Anthro 2009, Geneva, Switzerland).In this, the weight/ age index was used in accordance with the following criteria: < -2 z-scores = low weight or very low weight and ≥ -2 z-scores = adequate weight or eutrophic state; and the height/weight index as follows: < -2 z-scores = short height and ≥ -2 z-scores = adequate height.
The socioeconomic factors were analyzed using the following variables: geographical area, family income, mother's schooling level and mother's remunerated work.The variables for analyzing the basic sanitation conditions were: origin of the water, treatment for drinking water, sewage system and garbage disposal.
The housing characteristics were investigated by means of the following variables: housing conditions, number of people in the household, number of children under the age of fi ve years in the household and consumer goods possessed.
Housing conditions and consumer goods were classifi ed as scores.For housing, the variables of housing type, housing regime, walls, fl oor, roof and number of rooms received scores between one and three points (worst to best condition, respectively).Thus, the scores for housing conditions ranged from 10 to 18 points in total, and these were brought together as group 1 (10-13 points) and group 2 (14-18 points).The consumer goods scores were established according to the availability of electricity and possession of a television, radio, refrigerator and stove in the child's household.Each item scored one point and thus, consumer goods were brought together as groups 1 (0-1 point), 2 (2-3 points) and 3 (4-5 points).
The maternal factors analyzed were the mother's age and maternal anemia.Health and nutritional care, and food intake were characterized as the number of prenatal consultations and duration of breastfeeding.
Bivariate analysis was performed for each geographical area by means of simple Poisson regression, in order to investigate associations between occurrences of anemia and the independent variables, with a signifi cance level of 5%.The variables that presented a statistical association with anemia with p < 0.20 were selected to form part of the multivariate regression models.
The multivariate analysis was performed by constructing conceptual models for anemia based on Osório et al 18 (2004).Six hierarchical levels for determining anemia were taken into consideration, in the following order: socioeconomic factors (geographical area, family income, mother's schooling level and mother's remunerated work); housing conditions and sanitation (housing conditions, consumer goods, number of people in the household, number of children under the age of fi ve years, origin of the water, treatment for drinking water, type of sewage system and garbage disposal); maternal factors (mother's age and maternal anemia); health and nutritional care (number of prenatal consultations); nutritional status and morbidity (diarrhea over the last 15 days, weight/age indicator, height/age indicator and birth weight); and children's biological characteristics (child's age).Two multivariate analysis models were constructed: one for the urban area and the other for the rural area.
Poisson's regression with robust variance was used to investigate how the prevalence of anemia might have been infl uenced by the various explanatory variables. 4,6he backward method was used: the variables of the fi rst hierarchical level were analyzed together and the variables with signifi cance greater than or equal to 20% were progressively excluded.The variables of the second hierarchical level were then added to the model, and the same procedure was followed, with progressive exclusion of the variables of this level with p ≥ 0.20.Thus, all the hierarchical levels were analyzed.Upon concluding the models for the urban and rural areas, variables with p < 0.05 were taken to be signifi cant.To control for possible confounding factors, variables with p < 0.20 were kept in the models at each hierarchical level.
The statistical signifi cance was determined by means of the Wald test for heterogeneity and linear trends for ordinal variables, and the adjusted prevalence ratios and respective 95% confi dence intervals were estimated.
The EpiInfo software, version 6.04, and the R software, version 2.10.1 were used for the data processing and analysis.
The project for III PESN/PE was approved by the Research Ethics Committee of the Instituto de Medicina Integral Prof. Fernando Figueira (Procedural no.1321, 2004).All the children under the age of fi ve years and/ or women aged 15-49 years who were diagnosed as anemic received treatment with oral ferrous sulfate.

RESULTS
The mean age of the children studied was 32.6 months (standard deviation, SD = 15.4).It was 32 months (SD = 15.3) in the urban area and 33 months (SD = 15.5) in the rural area.Male children predominated (51.7% in the urban area and 51.2% in the rural area).
The variables that were statistically associated with anemia in the urban area were: family income and mother's schooling level; housing conditions, consumer goods, treatment for drinking water, type of sewage system and number of children under the age of fi ve years; mother's age and maternal anemia; number of prenatal consultations; and the child's age (Tables 1, 2 and 3).
For the rural area, the variables associated with anemia were: housing conditions and consumer goods; mother's age; height/age index; and the child's age (Tables 1, 2 and 3).
The prevalence ratios for anemia among children in the urban and rural areas, according to the preestablished hierarchical model are shown in Table 4.The Pernambuco.High prevalence of anemia is not only found in emerging countries like Brazil.Some studies in other countries have recorded similar or higher prevalences.In New Zealand, the prevalence of anemia was found to be 49% among children aged 6 to 11 months and 22% among children aged 12 to 24 months. 20In Vietnam, 45.1% of the children under the age of fi ve years were found to present anemia. 15In sub-Saharan African countries, prevalences of 82% in Benin and 83% in Mali were found. 14e prevalences of anemia were very similar between the urban and rural areas.In the rural area, the prevalence of anemia among children under the age of 24 months was around three times greater than among the older children.In the urban area, children under the age of 24 months presented prevalence of anemia that was around twice what was found among children over this age.a In the multivariate analysis for the urban and rural areas, the child's age (which was a variable at the proximal level for determining anemia) was the determinant that most infl uenced the prevalence of anemia.
An association between anemia and nutritional status in rural areas has also been observed in other studies. 1,14ow intake of iron-rich foods and diminished nutrient absorption caused by changes to the gastrointestinal epithelium in malnourished individuals contribute towards development of anemia.
Establishment of nutritional defi ciencies is linked to the population's socioeconomic conditions. 21,23,a Thus, the association found between low family income and anemia among children in the urban area corroborates studies both in Brazil and in other countries that have indicated that there is an inverse association between families' purchasing power and the prevalence of anemia among children. 3,10e association between mothers' schooling level and the care provided for to children has been greatly discussed in the literature, given that education has a relationship with the capacity to grasp the knowledge needed for adequate healthcare and nutrition for children, just as it provides a chance to enter the labor market and probably better socioeconomic conditions. 14,18,23The results from the present study refl ect this relationship, through showing that the prevalence of anemia was lower among urban children whose mothers had been to school for more than eight years.
In the multivariate analyses, among the variables analyzed in the distal group of associations with anemia, family income lost its statistical significance and remained as an adjustment variable in the urban area, while the mother's schooling level was the only variable that kept its statistical signifi cance in the fi nal model.None of the variables at the distal level remained in the fi nal model for the rural area, which was adjusted for the mother's remunerated work, coming from this level.
Socioeconomic indicators such as housing conditions and possession of goods have been investigated in order to obtain better estimates of the relationship between living conditions and health problems. 1,18,21The results found indicated that the prevalence of anemia was more than 46% among the children whose households had lower scores for housing conditions and consumer goods in both the geographical areas studied, which suggests that this is a serious public health problem.a Greater numbers of children under the age of fi ve years in the household showed a statistically signifi cant association with anemia, in the urban area.Greater numbers of children in the household implies more domestic work and worse quality of care for the children, and also means greater demand for foods to supply the needs of all members of the family, which may contribute towards greater risk of anemia. 9,23In the rural area, the number of children under the age fi ve years in the household did not have any impact on the prevalence of anemia.However, this does not indicate that people living in the rural area had better living conditions.On the contrary: this lack of impact may have been due to the worse situation regarding other socioeconomic, health and nutrition factors, given that when these were analyzed within a hierarchical model for determining anemia, they seemed to be superimposed on the infl uence of the number of children in the household.
Precarious sanitation conditions are linked with anemia, even if indirectly, since such conditions are associated with greater numbers of infectious and parasitic diseases, which in turn contributed towards diminishing the hemoglobin levels.a In the present study, the highest prevalence of anemia in the urban area occurred in households that were not using treated drinking water and did not have an adequate sewage system.Studies conducted among preschool children in northeastern Brazil have reinforced this relationship. 2,21 the rural area, anemia did not present a statistical association with lack of treatment for drinking water and lack of sewage treatment.However, 74% of the children in the rural area were living in households that were not using treated water for drinking, and 57% did not have adequate sewage systems in their homes.With regard to maternal characteristics, anemia in children showed a statistical association with teenage mothers in both the rural and the urban areas.This association has also been observed in other studies and is generally correlated with childcare quality. 1,19he mother's age remained in the fi nal models for the urban and rural areas.
Based on the concept that children's health refl ects the family situation, the possibility that a child would present anemia if other members of this family (especially the mother) were anemic is theoretically high.Such an association was found in the present study, thus corroborating previous fi ndings in Pernambuco 19 and also descriptions by Agho et al 1 (2008) in East Timor.Maternal anemia had a greater association with anemia in children in the urban area and, although this factor did not remain in the fi nal model for the rural area, the prevalence of anemia among the children of anemic mothers exceeded 44% in both areas.
Access to quality healthcare services constitutes an indicator of a population's health conditions and thus prenatal consultations are of importance for ensuring that healthy children are born. 1,18,21In the present study, the association between lower numbers of prenatal consultations and the presence of anemia was statistically signifi cant in the urban area of the state.
The lack of data on food intake formed a limitation to interpretation of the results from this study.The most frequent cause of anemia in children is iron defi ciency, generally associated with insuffi cient intake of this nutrient or consumption of foods that inhibit iron absorption. 10,16,20 the fi nal models of the present study, a greater number of variables were statistically associated with anemia among the children in the urban area than among those living in the rural area.
The rural area presented socioeconomic, housing and sanitation conditions that were more homogenous, which made it diffi cult to establish statistical relationships between these variables and the presence of anemia.Mothers' income and schooling levels in the rural area possibly did not show up as factors associated with anemia because the majority of the families were in the categories of lowest income and lowest maternal schooling level.Likewise, housing and sanitation conditions in rural areas are generally poor.Nationwide data emphasize that rural areas in northeastern Brazil present a worse situation than the national mean, regarding women's schooling and household infrastructure, thus characterizing a gap between urban and rural areas that is still large.g Success in combating anemia depends on understanding its associated factors.This panorama of anemia in Pernambuco contributes towards planning effective measures for its control, including maintenance of government programs for fortifi cation of wheat and maize fl our and for iron supplementation, as well as actions that aim to improve the population's socioeconomic, housing, sanitation and health conditions.
Third Investigation on Health and Nutrition in the State of Pernambuco, in 2006, conducted following an initiative from the Department of Nutrition, Federal University of Pernambuco, the Instituto de Medicina Integral Prof. Fernando Figueira and the State of Pernambuco Health Department, with fi nancial support from the Ministry of Health and the National Council for Scientifi c and Technological Development (CNPq procedural no.505540/2004-5) [unpublished data].

Table 1 .
Prevalence of anemia in children aged six to 59 months, according to biological characteristics, morbidity and nutritional status, per geographical area.Pernambuco, Northeastern Brazil, 2006.

Table 2 .
Prevalence of anemia in children aged six to 59 months, according to socioeconomic factors, housing characteristics and type of sewage system, per geographical area.Pernambuco, Northeastern Brazil, 2006.
a Values grouped according to scores

Table 3 .
Prevalence of anemia in children aged six to 59 months, according to maternal factors, health and nutritional care, and food intake, per geographical area.Pernambuco, Northeastern Brazil, 2006.

Table 4 .
Adjusted prevalence ratios for anemia in children aged six to 59 months, per geographical area.Pernambuco, Northeastern Brazil, 2006.
a Adjusted according to the variables of family income, housing conditions and type of sewage system.b Adjusted according to the variables of mother's work, housing conditions, consumer goods, treatment for drinking water, maternal anemia and height/age index.