Prevalence and factors associated with anemia among children under five years of age in Rombo district, Kilimanjaro region, Northern Tanzania

Background: Anemia is a severe public health problem affecting more than half of children under five years of age in low-, middle- and high-income countries. The study aimed to determine the prevalence and factors associated with anemia among children under five years of age in northern Tanzania. Methods: This community-based cross-sectional study was conducted in Rombo district, Kilimanjaro region, northern Tanzania, in April 2016. Multistage sampling technique was used to select a total of 602 consenting mothers and their children aged 6-59 months and interviewed using a questionnaire. Data were analyzed using Stata version 15.1. We used generalized linear models (binomial family and logit link function) with a robust variance estimator to determine factors associated with anemia. Results: Prevalence of anemia was 37.9%, and it was significantly higher among children aged 6-23 months (48.3%) compared to those aged 24-59 months (28.5%). There were no significant differences in anemia prevalence by sex of the child. Adjusted for other factors, children aged 6-23 months had over two times higher odds of being anemic (OR=2.47, 95% CI 1.73, 3.53, p<0.001) compared to those aged 24-59 months. No significant association was found between maternal and nutritional characteristics with anemia among children in this study. Conclusion: Prevalence of anemia was lower than the national and regional estimates, and it still constitutes a significant public health problem, especially among children aged 6-23 months. The study recommends iron supplementation, food fortification, dietary diversification, and management of childhood illnesses interventions for mothers and children under two years.


Introduction
In children under five years, anemia is a significant public health problem in the low-, middle-and high-income countries. The world health organization (WHO) defines anemia as a low blood hemoglobin concentration of less than 11g/dl in children under five years of age 1,2 . Anemia in children is a major cause of adverse health consequences such as stunted growth, impaired cognitive development, compromised immunity, disability and increased risk of morbidity and mortality [2][3][4][5][6][7][8][9][10][11] . Globally, about 43% of children under-five are anemic, and there is a marked variation in the prevalence of anemia between low-and middleincome countries (LMIC). Over 50% of anemic children live in LMIC 12 , and the highest prevalence rate (78%) was reported in Ghana and the lowest (26%) in Cuba 13,14 . According to WHO, the African region has the highest proportion (62%) of anemic children 12 .
A variety of factors causes anemia, but the most common cause is iron deficiency 1,3,12 . Iron deficiency can result from inadequate dietary intake or poor absorption, increased needs for iron during the high growth periods, and increased iron loses due to helminths infection 3 . Other causes of anemia can be infections like malaria, genetic makeup, and nutritional deficiencies of vitamins B12, A, C and folate 3 . Factors associated with anemia also vary from region to region. The factors include the area of residence (whereby children living in rural areas beingmore at risk), low education level of the mother, child's sex (high among males), child's age (below 24 months) and history of infections, high birth order and maternal history of anemia 1,4,[13][14][15][16][17][18][19][20] . Unemployment, low family income, low wealth quartile and high poverty index have also been associated with anemia in children under five 5,9,15,17 . In addition, poor breastfeeding practices and complementary feeding leads to anemia 7,14-16 .
To combat anemia in children, WHO recommends combined strategies such as iron supplementation, especially to vulnerable populations, food-based approaches to increase iron intake through food fortification and dietary diversification and management of infectious diseases, particularly malaria and helminth infections 21 . These strategies are recommended to be built into the primary health care system and existing programs such as maternal and child health, integrated management of childhood illness, adolescent health, safe motherhood, roll-back malaria, deworming and tuberculosis 21 . Improved quality of anemia care is also among key strategies to accelerate progress towards addressing this problem 22 . Although Tanzania is implementing these strategies 23 , the Demographic and Health Survey (DHS) report shows no improvement in reducing anemia prevalence. For the two consecutive DHS rounds, 2010 and 2015, the prevalence of anemia was 58%.
The results of the DHS show that the country is still far from reaching the set target of reducing anemia prevalence to 20% by 2020. In the Kilimanjaro region, Same District, anemia prevalence was 70% 19 . Since studies show variations in factors that are associated with anemia, there was a need to conduct this study in the Rombo district as an important step towards evidence-based decision-making when planning for interventions. Geographically Same is semi-arid district while Rombo is located around Mount Kilimanjaro, hence having different topographic conditions.

Study design and setting
This study utilized data from a community-based cross-sectional study conducted in Rombo district, Kilimanjaro  Study population, sample size, and sampling The study included consenting mothers and their children aged 6-59 months. A single proportion formula was used for sample size calculation. Using a standard normal value of 1.96 under 95% confidence interval, a 48% prevalence of anemia among children 6-59 months in Kilimanjaro region 2 , a margin of error of 5% and multiplying by a design effect of 1.5 to account for cluster design, the minimum required sample size was 575 mother-child pairs.
Multistage sampling technique was used to select 708 motherchild pairs from households with children aged 6-59 months. Two villages were randomly selected from each randomly selected ward. A listing of households with children under five years was generated with the help of village leaders or link persons, followed by a random selection of households. Systematic random sampling was used to select households. When the visited household had no child under five years of age, the next household was selected until the minimum required sample size was reached. If there were more than one child aged 6-59 months, the younger one was selected to represent the rest of the children in the household. If the child's mother was not at home, the research team visited the house a minimum of three times before declaring that the participant could not be reached. Children whose mothers were not available on the day of data collection were excluded from the study as it was not possible to verify child information if next in kin or neighbor was interviewed. In addition, after excluding 89 children aged <6 months and 17 with missing hemoglobin concentrations, we analyzed data for 602 mothers-child pairs Figure 1.

Amendments from Version 2
There are minor edits in the citations while some parts of the discussion have been revised with two additional citations for clarity.
Any further responses from the reviewers can be found at the end of the article

REVISED
Data collection methods A questionnaire, shared as extended data 26 , was used to collect data during face-to-face interviews. Although the questionnaire has not been validated in Tanzania, we adopted questions from the DHS and added some from previous literature. The following information was collected; maternal reproductive health, breastfeeding history, feeding patterns, initiation of complementary feeding, use of health facilities during pregnancy and child nutrition status. The questionnaire was in both English and Swahili languages but administered using the Swahili language, a language spoken by all the local people in this setting. Trained medical student at the Kilimanjaro Christian Medical University College collected data under the Institute of Public Health supervision.

Study variables and measurements
The dependent variable in this study was anemia. Anemia was defined as a blood hemoglobin concentration below 11.0 g/dl in children under five years of age 1 . Blood samples were drawn among children from a drop of blood taken from a finger prick or heel prick (for children aged 6-11 months) and collected in a microcuvette strip. Hemoglobin (Hb) was measured on-site using a portable HemoCue rapid testing method (HemoCue  Hb 301 Analyzer -HemoCue AB, Kuvettgatan 1, SE-262 71 Angelholm, Sweden). The anemia results were given on-site and children with severe anemia (hemoglobin level <7 g/dL) were referred to the nearby health facilities.
The independent variables included socio-demographical characteristics such as age of the mother in years (<20, 20-29 and 30+), education level, occupation (Peasant/farmer, Employed and Others), marital status (single, married/cohabiting and divorced/ separated/ widowed), area of residence (rural and urban depending on how the locals define them), alcohol consumption (Yes and No), body mass index (BMI) of the mother (underweight (<18.5Kg/m 2 ); normal weight (18.5-24.9 Kg/m 2 ), overweight (25-29.9 Kg/m 2 ), and obese (≥30 Kg/m 2 )); and child's age and sex. Nutritional characteristics included exclusive breastfeeding (Yes and No) 27 , colostrum feeding (Yes and No), meal frequency per day (≤3 meals and >3 meals), age at initiation of complimentary feeding (<6 months and 6+ months), and use of deworming drugs past six months (Yes and No).
Measurement of weight was performed using a SECA weighing scale (SECA GmbH & Co. KG, Hamburg, Germany) while recumbent length was measured for children aged <24 months and standing height was measured for older children using stadiometers. At least two measurements were taken then the average was calculated. Stunting, wasting, and underweight (height-for-age, weight-for-height z-score, and weight-for-age z-score below minus two standard deviations (-2 SD), respectively) from the median of the WHO reference population 2 . Child anthropometric z-scores were calculated using the 2006 WHO child growth standards through the "zscore06" package in Stata 28 .

Ethical consideration
Ethical approval was obtained from Kilimanjaro Christian Medical University College Research and Ethics Review Committee (KCMU-CRERC). Permission to conduct the study was also sought from the Rombo District Authority. Before data collection, logistics meetings were held with ward and village leaders of selected sites to inform them about the study's purpose. The study purpose was explained to mothers before enrolment. Those who agreed to participate provided written informed consent. Unique identification numbers were used to ensure the anonymity of participant information.

Statistical analysis
Data were analyzed using Stata version 15.1, StataCorp LLC. Means and standard deviations were used to summarize numeric variables while frequency and percentages for categorical variables. Chi-square (χ 2 ) test was used to compare the prevalence of anemia by participant characteristics. Odds ratio (OR) and 95% confidence intervals (CIs) were used to determine factors associated with anemia in children using generalized linear models (GLM) with binomial family and logit link function adjusted for potential confounding. Akaike information criteria (AIC) was used to select the best model. The GLM model with binomial family and log link function was favored against the log-linear model, i.e., Poisson family with log link function hence all the analyses were performed using the former model. A robust variance estimator was used to account for model misspecification hence improving precision of estimates. The stepwise regression method was used to select variables included in the adjusted analysis at the 10% threshold level. The age of the child remained the only significant predictor of anemia at this stage. Maternal age, alcohol use (statistically significant in the crude analysis), sex of the child, and child's nutritional characteristics, specifically exclusive breastfeeding, wasting, stunting, and underweight, were considered potential confounders, hence included in the final model.

Background characteristics of mothers and children
Data were analyzed for a total of 602 mothers and children aged 6-59 months. The mean age (SD) of mothers in this study was 29.9±7.6 years. More than half (52%) of all mothers were aged between 20-29 years, 70% had primary school education level, 81.3% were married or cohabiting with their partners. The prevalence of obesity among women was 14.3%. The median age (IQR) of children in this study was 24 (14, 36) months while more than half (52.5%) were aged between 24-59 months. Also, more than half (52.7%) of all children were males Table 1 29 .
Feeding practices and nutritional status of children The vast majority (96.3%) were given colostrum while the overall prevalence of exclusive breastfeeding up to six months was 40.1%. Less than half (45.2%) of children in this study were given more than three meals per day and 69.7% were initiated complimentary feeding before six months. Also, 70.5% of children in this study were given deworming drugs. This study's prevalence of wasting, stunting, and underweight was 10%, 38.5%, and 6%, respectively Table 2 29 .

Prevalence of anemia by child's age and sex
In this study, the mean (SD) hemoglobin level of children aged 6-59 months was 11.2±1.6g/dl and the prevalence of anemia (hemoglobin level less than 11g/dl) was 37.9%. Prevalence was slightly higher among females (39.7%) compared to 36.2% among males Figure 2 29 , but this difference was not significant (p=0.40). Prevalence was much higher among children aged 6-23 months (48.1%) compared to 28.5% among those aged 24-59 months Figure 3 29 . These differences in the prevalence by age were statistically significant (p<0.001).

Factors associated with anemia
The study performed crude and adjusted analyses to determine factors associated with anemia in children aged 6-59 months. In the crude analysis, factors associated with anemia were whether the mother consumed alcohol, exclusive breastfeeding, and child's age Table 3 29 . Lower odds of anemia were observed among children whose mothers consumed alcohol (OR=0.68, 95%CI 0.48, 0.95, p=0.03). Higher odds of anemia were observed among children who were breastfed exclusively (OR=1.53, 95%CI 1.09, 2.14, p=0.02) and children aged 6-23 months (OR=2.34, 95%CI 1.67, 3.28) compared to those aged 24-59 months which showed a much stronger association with anemia (p<0.001). There was a positive association between stunting and the odds of anemia (OR=1.39, 95%CI 0.99, 1.95) but this association was not strong (p=0.06), Table 3 29 .
Adjusted analysis for factors associated with anemia in children is shown in Table 4 29 . A multivariable model was developed by adding and later removing one variable after another to assess the presence and effect of confounding. Age of the child was the only variable that remained to be strongly (p<0.001) associated with higher odds of anemia. Adjusted for mother's age categories (years), whether a mother consumed alcohol during pregnancy, exclusive breastfeeding,  Table 4 29 .

Discussion
The prevalence of anemia among children aged 6-59 months in this study was 37.9%. Age of the child was the only factor significantly associated with anemia among children. This study's prevalence of anemia in this study is much lower than the national and regional estimates 2 and other sub-population studies in Tanzania 9,19 . One of these studies was hospital-based 9 , while the other included children aged In this study, adjusted for the background and nutritional characteristics, children aged 6-23 months had higher odds of having anemia than those aged 24-59 months. Infants  (<24 months) are consistently reported to be at higher odds of being anemic in other studies 2,4,5,13,14,31,32 . Infants have a higher demand for nutrients needed for their growth, hence need proper complementary feeding. In this setting, there is a practice of giving porridge (a mixture of water, maize flour, and added sugar), cow's milk and less diversified foods at a younger age 33 . This practice could be one of the factors that leads to poor anemia status in children 33,34 . Also, conflicting advice on infant and young child feeding from various sources, including close relatives, community members, and health care providers affects breastfeeding practices, impacting the child's anemia status 34 . Receiving quality anemia care, particularly nutrition advice about healthy foods and the minimum acceptable diet to the care giver, and routine hemoglobin measurement is critical in reducing anemia burden for children 6-23 months, who are most at risk 22 . There were no significant differences in the prevalence of anemia by sex of the child in this study which is consistent with findings from other studies 13,14,18,30 . On the contrary, females have been reported to be less likely to be anemic in Ethiopia 16 , which is contrary to findings from Kenya where the risk was high in male children (aged 6 months to 14 years) 31 , which could account for these differences. We did not find an association between maternal characteristics such as age categories, education level, occupation and ANC visits among others contrary to other studies. ANC visit and mother's occupation have been associated with anemia elsewhere 7,16 . The higher education level of mothers is protective against childhood anemia 15,19,31 .
Likewise, there was no association between nutritional characteristics such as deworming drugs uptake, exclusive breastfeeding (EBF), colostrum feeding, complementary feeding, and feeding frequency with anemia. However, other studies reported an association between nutritional characteristics with a higher risk of anemia in under-five children 4,5,14,18,33 . On the contrary, Meinzen-Derr et al. 20 reported that, infants exclusively breastfed for six months in developing countries might be at increased risk of anemia, especially among mothers with poor iron status. In addition, there is evidence that the longer the infant is exclusively breastfed, the worse the severity of childhood anemia due to low iron content in breast milk 35,36 . The positive association between EBF and anemia was observed in this study but was not statistically significant. The effect of EBF on anemia in children is an area that needs further research. Despite the observed association in this study, nutritional interventions (EBF included) are among the key strategies to reduce the burden of anemia in under-five children 21,23,27 .
The study involved participants from most wards in the Rombo district, providing a picture of anemia in children under five. However, the findings in this study may not be generalized to other districts in Kilimanjaro and regions across the country. Also, the study might have been prone to recall and social desirability bias due to the self-reporting of nutritional practices associated with anemia. These may under or over-estimate these practices in the district.

Conclusion
The prevalence of anemia was lower than the national and regional prevalence but it still constitutes a significant public health problem especially among children aged 6-23 months. There were no significant differences in anemia prevalence by sex of the child and any of the nutritional characteristics. The study recommends iron supplementation, food fortification, dietary diversification, and management of childhood illnesses interventions for mothers and children under two years. Future studies should apply mixed methods, including longitudinal to children 6-59 months.

Discussion:
The authors don't discuss thoroughly the differences between their findings and the literature:

○
For example: Age of the child was the only factor significantly associated with anemia among children. What could have been the confounding factors? 1.
High prevalence in other studies could be linked to differences in the study population and wider population coverage since most utilized nationally representative data such as DHS data. What are these differences?

2.
After the statement of high prevalence in other studies, one would expect that the authors would cite those studies. Instead, they mention Ayoya et al. and Pita et al whose findings do not reflect high prevalence of anemia.

References:
Reference 2: is related to the Tanzania DHS, not to a WHO report. Therefore, the DHS is not the primary source of the definition of anemia, but only used it.
Reference 3: This policy brief focuses on women of reproductive age, not on children 6-59 months of age. It may not be relevant here. In fact, it refers to the general WHO definition of anemia, without even indicating the cut-off points.

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes Title: the methodology and results sections specify that the study focuses on children 6-59 months old. The title should reflect the same age category.

Are the conclusions drawn adequately supported by the results? Yes
Response: We thank the reviewer for this comment. However, we would like the title to be stated the same way because children 6-59 months still constitute the under-five-year category. Indeed, previous studies stated the same way even though their target population was children 6-59 months [1].

Results:
The study shows no association with exclusive breastfeeding, colostrum feeding. These two variables are not appropriate in this context, since the methodology section limits the study to children 6-59 months.
○ Response: Previous studies established an association between exclusive breastfeeding with anemia in children 0-6 months. Women in this study were asked whether they breastfed the child exclusively for six months (regardless of the child's current age), the responses being Yes/No, but no association was observed with anemia. As indicated in the last but one paragraph of our discussion, more studies are necessary to establish a causeeffect relationship between exclusive breastfeeding and the risk of anemia in children. Nevertheless, there is evidence that the longer the infant is exclusively breastfed, the worse the severity of childhood anemia due to low iron content in breast milk [2,3]. We have also added this information in the discussion section.

Discussion:
The authors don't discuss thoroughly the differences between their findings and the literature: Response: Several analyses were performed to determine the factors associated with anemia among children 6-59 months in this study. Details are also contained in the data analysis section. Age was the only statistically significant predictor of anemia even after adjusting for background and nutritional characteristics. We have added this description in the discussion section.

High prevalence in other studies could be linked to differences in the study population and wider population coverage since most utilized nationally representative data such as DHS data.
What are these differences?
Response: By differences here we refer to the methodological variations in the included population. Cross-sectional surveys in a small geographical area are likely to give a slight difference estimate compared to the nationally representative sample such as the DHS. The first paragraph of the discussion section specifically highlights these issues.
3. After the statement of high prevalence in other studies, one would expect that the authors would cite those studies. Instead, they mention Ayoya et al. and Pita et al whose findings do not reflect high prevalence of anemia.
Response: Those studies are cited just before this statement. The sentence citing Ayoya et al is slightly edited for clarity.

References:
Reference 2: is related to the Tanzania DHS, not to a WHO report. Therefore, the DHS is not the primary source of the definition of anemia, but only used it.
Response: Reference 2 is the Tanzania Demographic and Health Survey report of 2015/16 and it contains the cited information. These anthropometric measurements were also assessed following a similar methodology documented in this report. In addition, reference 1 cites the WHO document defining anemia in children 6-59 months. In table 1, single mother is mentioned, how single mother has child, it may be separated/divorcee of widow, this classification whether used in Africa the prevalence of underweight mentioned is 6% and wasting 10%, i still have some doubt about this percent as stunting is 38%. While considering reference in logistic regression, always we consider good practices as reference, but this is not followed in this table. Age at Complementary feeding should be below 6 month, 6-8 months and after 8 months and likely analysis should be done.

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? In method: Under study design, it is mentioned that he study aimed to assess the nutritional status of children under five years in the district. but in abstract, only anemia prevalence is mentioned and also title does not reflect this under study population, it is still not clear how many wards were selected and why systematic sampling was used for selection of HHs.

Response:
The current study used secondary data from a community-based cross-sectional study conducted in the district whose primary purpose was to assess the nutritional status of children under five years of age. Our analysis focused primarily on anemia and its associated factors given the high burden of this problem in the region and Tanzania in general.
Under study variable, para 3, it is mentioned reference 2 for WHO reference population, but reference 2 is not WHO reference.

Response:
We acknowledge the reviewer's comment. Please note that reference 2 is the Tanzania Demographic and Health Survey report of 2015/16 and it contains the cited information. These anthropometric measurements were also assessed following a similar methodology documented in this report.
Under statistical analysis, nothing is mentioned about how nutritional status was decided.

Response:
We have this information at the end of paragraph 3 of the study variables section which reads; "Child anthropometric z-scores were calculated using the 2006 WHO child growth standards through the "zscore06" package in Stata". This function in Stata software allows the specification of the child's height, weight, age, and sex among other parameters for the calculation of z-scores.
In table 1, single mother is mentioned, how single mother has child, it may be separated/divorcee of widow, this classification whether used in Africa the prevalence of underweight mentioned is 6% and wasting 10%, i still have some doubt about this percent as stunting is 38%. While considering reference in logistic regression, always we consider good practices as reference, but this is not followed in this table. Age at Complementary feeding should be below 6 month, 6-8 months and after 8 months and likely analysis should be done.

Responses:
Regarding marital status, it is possible for a single mother to have a child even without being married. However, the reason for having two groups (< 6 months and ≥ 6 months) as shown in Table 2 and 3 is to gain more statistical power as a function of sample size. It can be seen in Table 2 that there were 163 (30.3%) children initiated complementary feeding at ≥ 6 months. Breaking this variable further will not improve the results observed in Table 3 and may increase the probability of committing type I error [4].

Version 1
Reviewer

Response:
The reviewer comment is acknowledged. The first and second paragraphs of the discussion section has been revised to clearly explain the observed results, which also compare with previous studies and shows the public health implications.
○ A question about deworming drugs was included in the questionnaire. What was the prevalence of STH in the study area? The impact of intestinal helminth infection should be discussed in this manuscript. ○ Response: Unfortunately, the study did not assess the prevalence of STH which may be associated with increased anemia risk, which remains an area for future studies.
○ Malaria is prevalent in Tanzania. Why wasn't malaria data included in the study, given that it contributes to anemia? ○ Response: We thank the reviewer for this comment. Although malaria is prevalent in Tanzania, not all regions are highly affected. At the regional level, malaria prevalence is high in the Southern slopes of Mount Kilimanjaro compared to the highlands [1]. Secondly, the primary study aim was to assess prevalence of anemia in children under five in the Rombo district. A more comprehensive study that capture the socialeconomic, demographic, behavioral, and clinical determinants of anemia and other child nutritional-related characteristics is essential in this setting. The benefits of publishing with F1000Research: Your article is published within days, with no editorial bias • You can publish traditional articles, null/negative results, case reports, data notes and more • The peer review process is transparent and collaborative • Your article is indexed in PubMed after passing peer review • Dedicated customer support at every stage • For pre-submission enquiries, contact research@f1000.com