Determinants of anemia among women and children in Nepal and Pakistan: An analysis of recent national survey data

Abstract Anemia remains one of the most intractable public health challenges in South Asia. This paper analyzes individual‐level and household‐level determinants of anemia among children and women in Nepal and Pakistan. Applying multivariate modified Poisson models to recent national survey data, we find that the prevalence of anemia was significantly higher among women from the poorest households in Pakistan (adjusted prevalence ratio [95% CI]: 1.10 [1.04–1.17]), women lacking sanitation facilities in Nepal (1.22 [1.12–1.33]), and among undernourished women (BMI < 18.5 kg/m2) in both countries (Nepal: 1.10 [1.00–1.21] and Pakistan: 1.07 [1.02–1.13]). Similarly, children in both countries were more likely to be anemic if stunted (Nepal: 1.19 [1.09–1.30] and Pakistan: 1.10 [1.07–1.14]) and having an anemic mother (Nepal: 1.31 [1.20–1.42] and Pakistan: 1.21 [1.17–1.26]). Policies and programs need to target vulnerable and hard‐to‐reach subpopulations who continue to bear a disproportionate burden of anemia. Covariates of poverty underpin rates of anemia among children and their mothers, but income growth alone will not suffice to resolve such deeply entrenched problems. Greater understanding of the relative role of various diet, health, sanitation, and educational factors by local context should guide investments to resolve anemia in tandem with stunting and maternal underweight.


| Analytic samples
WRA were included in this analysis if they (a) had data on hemoglobin concentration or data on presence of anemia, (b) had complete data on household-level and individual-level modules, and (c) were not pregnant. A total of 5,794 WRA were included in the analysis for Nepal and 8,324 in the analysis for Pakistan.
Children <5 years old were included in the analysis if they (a) had data on hemoglobin concentration or data on presence of anemia and (b) had complete data on household-level and individual-level modules.
In total, 2,088 children from Nepal and 8,968 children from Pakistan were included in the respective analyses.

| Anemia
Anemia was defined as hemoglobin concentration <120 g/L for WRA and <110 g/L for children <5 years old (UNICEF et al., 2001).
Hemoglobin status was evaluated using capillary blood and the HemoCue rapid testing in Nepal and evaluated using venous blood in

Key messages
• Anemia was prevalent among children and women in our sample from Nepal and Pakistan, indicating a moderate to severe public health concern in these populations.
• Stunted childrenover 40% both countrieswere more likely to be anemic compared with those not stunted, and having an anemic mother was associated with a higher odds of anemia among children. Among women, being thin increased the likelihood of anemia in both countries. Given the overlaps of anemia, stunting, and thinness, there are opportunities to target multiple conditions in the same individuals and households simultaneously.
• South Asia includes countries such as Bangladesh and Nepal that have strong policies to address anemia, while other countries in the region that suffer from a higher prevalence of anemia could benefit from a stronger policy and programme commitment to targeting anemia.
Pakistan. Hemoglobin concentration was adjusted for altitude in Nepal, but not Pakistan.

| Predictors of anemia
We conducted a literature review of predictors of anemia in South Asia and tested factors that have been considered previously, with emphasis on two recent studies that explored the correlates of anemia using India 's DHS 2005-2006and Bangladesh's DHS 2011(Balarajan et al., 2013Kamruzzaman et al., 2015). Improving the comparability of analyses across country studies in the region is an important goal underpinning the work reported here. The factors tested in Nepal and Pakistan differed slightly based on the availability of data. For example, the Pakistan NNS included dietary and micronutrient status data, whereas the Nepal DHS did not.
A variable for wealth index was created using principal component analysis. This derived variable was already included in the Nepal dataset and was created by us using the Pakistan dataset, using the method to create this variable from the final report for the Pakistan NNS. Food security scores were calculated using methods used in Pakistan's NNS (Aga Khan University, 2011).
The 2006 World Health Organization growth standards were used to calculate children's height-for-age, weight-for-age, weightfor-height, and body mass index (BMI)-for-age z-scores and define stunting, wasting, underweight, and overweight, respectively.
Variables for these z-scores existed in the Nepal dataset and were calculated in the Pakistan dataset using the zscore06 in Stata (Leroy, 2011).
Serum retinol, serum ferritin, plasma zinc, vitamin D, and calcium were evaluated in Pakistan to determine vitamin A deficiency, low ferritin, zinc, vitamin D, and calcium deficiency. Alpha 1 acid glycoprotein was used to adjust serum retinol, serum ferritin, and plasma zinc for inflammation (Engle-Stone et al., 2011;Thurnham et al., 2010).

| Statistical analysis
The prevalence of anemia among women and children was calculated for Nepal using reported sample weights. After log binomial models failed to converge, modified Poisson models were used to evaluate factors associated with anemia and estimate prevalence ratios, adjusting for the clustered design of the surveys. This approach was chosen because the outcome was not rare (>10%), and therefore, log binomial and modified Poisson models were more appropriate than logistic regression (Balarajan et al., 2013;Barros & Hirakata, 2003;Deddens & Petersen, 2008;Zou, 2004). Factors that were extremely rare or common (<5% or >95%) were not included in the model, and backwards stepwise processes were used to identify parsimonious models to explain predictors of anemia (all p < .05) in each country for each group of WRA and children <5 years old.

| RESULTS
The characteristics of the study populations included in this analysis are described in Table 1

| Factors associated with anemia in Nepal
Factors associated with anemia among children <5 years old and WRA in Nepal are presented in Tables 2 and 3. Being a stunted child was associated with a higher prevalence of anemia compared with not being stunted (APR [95% CI]: 1.19 [1.08-1.30]), as was lower age. That is, the prevalence of anemia was greatest among infants <6 month old. At the same time, children with an anemic mother were more likely to be anemic compared to children of nonanemic mothers (1.31 [1.20-1.42]).   Median split was used to dichotomize the variable.
*P < 0.05, **P < 0.01, ***P < 0.001. assessed during Baisakh (mid-April to mid-May) were more likely to be anemic than was the month with the highest likelihood of anemia, with women assessed in any other months of the year less likely to be anemic.

| Factors associated with anemia in Pakistan
Factors associated with anemia among children <5 years old and WRA in Pakistan are presented in Tables 4 and 5.
Age was associated with anemia among children; children aged 6-11 and 12-23 months were more likely to be anemic than children aged <6 months, whereas children 36-59 months old were less likely to be anemic than children <6 months old. Stunted children were more likely to be anemic than nonstunted children (1.10 [1.07-1.14]), and having been diagnosed with worms in the past 6 months also increased a child's likelihood of anemia (1.06 [1.00-1.13]) as well. Children with an anemic mother were more likely to be anemic than children with a Province of residence was also associated with anemia.
Examining the association between micronutrient deficiencies (vitamin A, vitamin D, iron, and zinc) and anemia among children in bivariate models adjusted for the survey design revealed that children with iron depletion were more likely to be anemic than children who  Median split was used to dichotomize the variable.
*P < 0.05, **P < 0.01, ***P < 0.001. The period of conception up to a child's second birthday is critical both to maternal health and nutrition, to healthy birth outcomes, and to subsequent child growth and development. It is therefore important to recognize that in South Asia, when mothers are anemic, infants are often born anemic. This suggests a need to focus on much policy attention aimed at resolving anemia to pregnant and lactating mothers, as well as to infants and young children in the context of introducing appropriate iron-rich complementary foods at 6 months of age as well as optimizing other infant and young child feeding practices.
It also suggests that tackling anemia should be a focus of broader initiatives aimed at improving child growth. The international community and many national governments have made a reduction in child stunting (another World Health Assembly target for 2025) a policy priority. Advocacy for this is strong, including by the influential Scaling up Nutrition movement and the Stop Stunting initiative that is focused on South Asia. Stunted children in Bangladesh have been shown to have a higher likelihood of being anemic than nonstunted children, but this was not observed in India (Baranwal, Baranwal, & Roy, 2014;Khan et al., 2016). This means that careful analysis has to underpin costeffective programing and that targeted actions with tailored activities are likely to gain more traction than conventional one-size-fits-all investments. Simply promoting child growth (while needed) may not in itself reduce anemia, and vice versa.
For example, studies have been inconsistent with regard to the effect of iron supplementation on linear growth. One systematic review concluded that iron supplementation provides no apparent benefit to linear growth among children (Ramakrishnan, Nguyen, & Martorell, 2009). However, that systematic review did not account for the difference in supplementing iron-deplete versus iron-replete children. As a result, although iron supplementation did not in their assessment appear to impact growth, it is likely that treating anemia among specifically anemic children could contribute to improving lin-  Median split was used to dichotomize the variable.
between indicators of socioeconomic status and nutrition (Adler & Ostrove, 1999;Adler et al., 1994), and it is well established that socioeconomics is an important determinant of anemia (Balarajan et al., 2011); thus, our findings reinforce the need for targeted efforts to focus on children from poor, hard to reach, and socially excluded households and population groups.

| Anemia among WRA
In our analysis, BMI was inversely associated with anemia in WRA in Nepal and Pakistan, which is in line with findings from India and Bangladesh (Balarajan et al., 2013;Kamruzzaman et al., 2015).
Overweight and obese women had a lower prevalence of anemia compared with women of normal weight. However, the prevalence of anemia in overweight women was 20-22% in Nepal and 46-47% in Pakistan, indicating that the double burden of malnutrition (overweight or obesity alongside underweight and nutritional deficiencies) is already present in these countries and needs serious policy attention.
In Pakistan, the likelihood of anemia was greater among women who had given birth more than twice, whereas in India, having five or more children was associated with a higher prevalence of anemia compared with having no children (Balarajan et al., 2013). Thus, the physiological burden of repeated pregnancy, likely coupled with low-quality diets, parasites, and poor sanitation, poses important risks to mother and their babies.
Women's years of schooling was not a determinant of anemia in Nepal or Pakistan. That contrasts with Bangladesh and India, where an inverse relationship was found between years of formal education and anemia (Kamruzzaman et al., 2015) (Balarajan et al., 2013). In a pooled analysis across low-income countries, education level along with wealth and cultural norms and behaviors were overarching determinants of anemia among women (Balarajan et al., 2011). The lack of association in our analysis for Nepal and Pakistan suggests that factors that are relatively less amenable to the influence of education (such as low quality of diet, parasite loads, and lack of access to clean water and sanitation facilities) may be playing a more dominant role in anemia.

| Policy implications
Iron deficiency anemia is one of the top 10 leading causes of YLD in South Asia (and within the top 5 in Afghanistan, Bangladesh, India, and Nepal, 2013), with notable human and economic costs to countries. The median cost of iron deficiency for 10 developing countries was found to be $2.32 per capita (0.57% of GDP) based on annual physical productivity losses alone (Horton & Ross, 2003). When both physical and cognitive productivity losses were accounted for, this increased to $16.78 per capita (over 4% of GDP).
Programs that target specific known forms of anemia can be costeffective. In calculating the cost-effectiveness of iron supplementation and fortification, Baltussen found that iron supplementation would avert~2.5 million disability-adjusted life years (DALYs) in Africa and Southeast Asia at 95% coverage (Baltussen, Knai, & Sharan, 2004). In Southeast Asia, the cost of iron supplementation at 95% coverage would be 115 international dollars per DALY averted. For iron fortification, the cost would be 35 international dollars per DALY averted.
In addition to supplementation and fortification, efforts to increase dietary intake of iron-rich foods across South Asia are needed. Availability and geographical and financial access may be barriers to these dietary changes, as well as religious and cultural beliefs. Homestead food production programs have shown to change diets and reduce anemia among women and children. This is a promising intervention that has been used in Bangladesh and Nepal, among other countries (Olney, Pedehombga, Ruel, & Dillon, 2015;Talukder et al., 2010).
Although fortification and supplementation are cost-effective approaches to preventing and treating anemia due to nutritional deficiencies, other potential causes of anemia cannot be overlooked. Certain infectious diseases and genetic disorders can also cause anemia.
In such cases, micronutrient supplementation or fortification will not correct anemia (Balarajan et al., 2011). Water, hygiene and sanitation interventions, deworming, and malaria prevention and treatment are interventions that may be considered alongside dietary improvement and micronutrient supplementation and fortification, depending on the context-specific conditions. Genetic hemoglobin disorders in developing countries pose an added barrier given the low diagnostic capabilities for these disorders in remote and resource poor areas.
The policy commitment and scope to combat anemia in children and women in South Asian countries vary. In response to a 75% Health Policy andStrategic Framework (2005-2015) that included little focus on anemia prevention, though the framework did include nutrition advice and micronutrient supplementation. A large-scale fortification program is underway, including the fortification of wheat flour with iron, which is expected to reduce the prevalence of anemia.
Although the fortification program is promising, Pakistan may need a more comprehensive strategy to reduce anemia-particularly among young children, who are likely to benefit less from large-scale fortification of wheat flour with iron-that considers the multifactorial etiology of anemia, as seen in countries such as Bangladesh and Nepal and the potential for impact of each of the interventions in the population groups at a higher risk of anemia (i.e., young children, adolescents, and WRA).

| CONCLUSIONS
Bangladesh is the only country in South Asia that is on track to meet the World Health Assembly goal to reduce by 50% the prevalence of anemia in WRA by 2025. This argues for much greater policy attention and programatic investment to tackle anemia across the South Asia region. Progress in Nepal was rapid between 1998 and 2006, falling from 65% to 34%, but since 2007 has stalled (World Health Organization, 2014). It is widely believed that iron supplementation activities helped secure the early progress but that other determinants are in need of being addressed. Efforts are currently underway in Nepal to determine how best to address the multifactorial etiology of anemia.
In Pakistan, as in neighboring Afghanistan and India, the prevalence of anemia appears to have increased over the past decade or so. Much more needs to be understood about the drivers of such regression. The achievement of the global target of cutting anemia by half among WRA requires a relative rate of reduction exceeding 5% per year. The target uses the 1993-2005 period as its reference point for the 50% reduction, which underlines the fact that many parts of South Asia have a huge task ahead of them-first managing to reverse recent trends and then accelerating reduction at rates that will need to reach almost 10% relative reduction per year by 2025 (World Health Organization, 2014).
With millions of anemic women and children in South Asia, governments of this region have to take urgent action. Part of this will involve tackling endemic malaria, but actions are needed across multiple sectors in coordinated ways. Targeting higher risk groups with wellimplemented evidence-based interventions that combine improvements in diets, micronutrient supplementation, and food fortification to address nutritional anemia while responding to the multifactorial etiology of anemia will utilize resources more effectively and yield larger impacts. Continued monitoring of risk factors for anemia and progress in program implementation will inform policymakers, program implementers, and researchers of changing and stagnant patterns across time and geographies.