The burden of anemia among women in India


Objective: This research investigates the prevalence and determinants of anemia among women in Andhra Pradesh. We examined differences in anemia related to social class, urban/rural location and nutrition status body mass index (BMI). We hypothesized that rural women would have higher prevalence of anemia compared to urban women, particularly among the lower income groups, and that women with low body mass index (BMI; <18.5 kg/m2) would have a higher risk compared to normal or overweight women.

Design: The National Family Health Survey 1998/99 (NFHS-2) provides nationally representative cross-sectional survey data on women's hemoglobin status, body weight, diet, social, demographic and other household and individual level factors. Ordered logit regression analyses were applied to identify socio-economic, regional and demographic determinants of anemia.

Setting: Andhra Pradesh, a southern Indian state.

Subjects: A total of 4032 ever-married women aged 15–49 from 3872 households.

Results: Prevalence of anemia was high among all women. In all 32.4% of women had mild (100–109.99 g/l for pregnant women, 100–119.99 for non-pregnant women), 14.19% had moderate (70–99.99 g/l), and 2.2% had severe anemia (<70 g/l). Protective factors include Muslim religion, reported consumption of alcohol or pulses, and high socioeconomic status, particularly in urban areas. Poor urban women had the highest rates and odds of being anemic. Fifty-two percent of thin, 50% of normal BMI, and 41% of overweight women were anemic.

Conclusion: New program strategies are needed, particularly those that improve the overall nutrition status of women of reproductive ages. This will require tailored programs across socio-economic groups and within both rural and urban areas, but particularly among the urban and rural poor.

Sponsorship: Partial support for Margaret Bentley was provided by the Ford Foundation/India and the Carolina Population Center. Support for Paula Griffiths was provided by the Andrew W. Mellon Foundation.


Anemia continues to be a major public health problem worldwide, particularly among females of reproductive age in developing country settings. In 1992, World Health Organization global estimates of anemia prevalence averaged 56%, with a range of 35–75% depending on geographic location World Health Organisation 1992 (WHO, 1992). Prevalence of anemia in South Asia is among the highest in the world, mirroring overall high rates of malnutrition.

In India, recent nationally representative data from the National Family Health Survey 1998/1999 (International Institute of Population Sciences and ORC Macro 2000) on anemia of women of reproductive age describe the magnitude of the problem. More than one third of Indian women have a body mass undex (BMI) <18.5 kg/m2, reflecting chronic energy and micronutrient deficit. The prevalence of anemia among all women in the Indian sample is 52%. Fifteen percent of these women are classified as moderately anemic (Hb 70–99 g/l) and 2% as severely anemic (Hb <70 g/l). While there are regional differences, prevalence rates across the states are remarkably similar, reflecting underlying determinants that include diets low in heme-iron and high in phytates, high levels of malaria and other infectious diseases, and frequent reproductive cycling that decreases iron stores (Gillespie, 1998; Allen, 1997; Stoltzfus, 1997). Smaller-scale studies conducted in India of micronutrient deficiency confirm the high prevalence of anemia among adolescent girls and women (Sharma et al, 1996; Kanani & Poojara, 2000; Chakma et al, 2000; Rajaratnam et al, 2000; Kapil et al, 1999).

The consequences of anemia for women include increased risk of low birthweight or prematurity, perinatal and neonatal mortality, inadequate iron stores for the newborn, increased risk of maternal morbidity and mortality, and lowered physical activity, mental concentration, and productivity (Gillespie & Johnston, 1998; Stoltzfus, 1997; Allen, 1997). Women with even mild anemia may experience fatigue and have reduced work capacity (Gillespie, 1998).

Qualitative studies conducted on women's illness and health-seeking behavior in diverse socio-cultural settings across India report ‘weakness’ as one of the most common symptoms of ill-health that they experience, and weakness (often attributed to poor diet and poverty by the women) is often ranked at the top of illnesses that cause concern (Patel, 1994; Kanani et al, 1994; Kielmann, 2000; Amin, 2000; 1994; Bentley & Parekh, 1998).

In previous research to examine the determinants of body mass index (BMI) and under/over nutrition among women of reproductive age, we analyzed NFHS-2 data from one southern state of India, Andhra Pradesh (Griffiths & Bentley, 2001). We found large differentials in women's BMI related to women's residential status (urban vs rural), socioeconomic status, and other variables. Our analyses showed that more than 37% of rural women had a low BMI (<18.5 kg/m2) and 7.3% a BMI >25 kg/m2 (overweight or obese), compared with 12.1% and 37% of women living in large urban areas with low and high BMI, respectively. Moreover, as expected, there was a strong relationship between socioeconomic status and BMI. We concluded that these disparities in women's nutrition status were primarily related to women's access to resources and income, including better diets and access to health care, regardless of whether they lived in rural or urban areas.

This research investigates the prevalence and determinants of anemia among the same women in Andhra Pradesh. Following the results of our previous research related to the nutrition transition in India, we expected to find differences in anemia related to social class, urban/rural location, and nutrition status (BMI). We hypothesized that rural women would have a higher prevalence of anemia compared with urban women, particularly among the lower income groups. We also hypothesized that women with low BMI (<18.5 kg/m2) would have a higher risk of anemia compared to women of normal or overweight.

Materials and methods


We analyzed data from the second Indian National Family Health Survey 1998/1999 (International Institute of Population Sciences and ORC Macro 2000) for the state of Andhra Pradesh (NFHS 1998/1999). NFHS 2 is a demographic and health survey collected as part of the Demographic and Health Survey (DHS) program, which is funded primarily by the United States Agency for International Development. Additional funding for the nutritional components of the survey in India was provided by the United Nations Children's Fund (UNICEF).Footnote 1 The national survey covered a representative stratified random sample collected between November 1998 and December 1999 of approximately 95 000 women aged 15–49 from the 26 states of India. The main strata used in the sampling process were rural and urban areas. The primary sampling units (villages in rural areas and census enumeration blocks in urban areas) were selected with probability proportional to size from the rural and urban areas. Households were selected from within the selected primary sampling units. Andhra Pradesh, a state in southern India that was the first state to publicly release NFHS 2 data, provided the sample for the present analysis. It includes survey and nutrition status data on 4032 ever-married women aged 15–49 from 3872 households.Footnote 2


The second round of the National Family Health Survey (NFHS 2) included weights and heights of women of reproductive ages, as well as hemoglobin measures.Footnote 3 BMI was calculated for each of the women by dividing her weight (kg) by her height (m2). The Hemocue system was used to estimate the concentration of hemoglobin in capillary blood (International Institute of Population Sciences and ORC Macro 2000). A single drop of blood was taken from a finger prick after removing the first two drops of blood to ensure that the sample was based on fresh capillary blood and placed into a cuvette for measurement.

Data analysis

We used ordered logit models to identify socioeconomic, regional, diet, demographic and health determinants of anemia status. Primary outcome variables in the analyses were created from hemoglobin measurements. We classified women as mildly, moderately or severely anemic based upon their hemoglobin status and following international references (WHO, 1992). A hemoglobin concentration of less than 70 g/l was used to define severe anemia, 70–99.99 g/l for moderate anemia, and 100–109.99 g/l to correspond to mild anemia in pregnant women (n=170) and 100–119.99 g/l for non-pregnant women. However, only 2.2% of women were classified as being severely anemic. For the purposes of the regression modeling we have therefore combined the severe and moderate groups to create a severe/moderate category (a hemoglobin concentration of less than 100 g/l) to avoid problems with zero cell counts in estimating the models. In the regression model the outcome variable was coded so that women with no anemia were given a value of 0, those with mild anemia a value of 1 and moderate to severe anemia was coded as 2.

Variables tested for significance in their association with anemia, (anemia was classified using the WHO (1992) hemoglobin reference definitions), in the ordered logit models are presented in Table 1. The variables fall into four main categories: location and standard of living, socio-demographic variables; health and diet variables; and BMI. Model 1 included a variable that combined the standard of living indexFootnote 4 and urban location variables. This allowed a test of whether there were differences in the likelihood of being anemic between socioeconomic groups both within urban and rural areas. Model 2 introduced the socio-demographic variables in addition to the location and standard of living variables. Two additional models were tested, adding to the variables already included in the earlier models. Model 3 introduced the diet and health variables, and model 4 the variable classifying women as thin (BMI <18.5 kg/m2), normal-weight (BMI 18.5–24.9 kg/m2), or overweight (BMI ≥25 kg/m2). model 1—location of residence and standard of living variable; model 2—model 1+socio-demographic variables; model 3—model 2+health and diet variables; model 4—model 3+BMI.

Table 1 Variablesa tested for significant association with anemiab in logistic regression models 1–4

Building the models in this way allowed a test of the significance of the association of location and standard of living with anemia status, controlling for a range of other factors. In addition, this allowed the identification of factors that reduced the significance of the combined location and standard of living variable in each model, hence enabling the identification of variables associated with the standard of living within the urban living environment and a woman's hemoglobin status.

Model estimation

We used SPSS version 10 for the preliminary statistical analyses. All women with hemoglobin measures in the survey were included in the preliminary analysis (96% of the original sample). Individuals with missing data for hemoglobin were those who refused to have their hemoglobin levels measured. We tested for any significant difference between women with missing hemoglobin measures and those who had a hemoglobin measure in our main descriptive variables of interest: place of residence and standard of living. No significant difference (P<0.05) was detected in the distribution of missing cases across the categories of these two variables.

Descriptive statistics were produced for Andhra Pradesh using the individual sampling weights. In Andhra Pradesh the sampling weight corrects for differential non-response between the geographical regions in which the survey was administered. Using the sample weight in the analysis allows correction of disproportionate representation of women from certain regions because of non-response. Failure to account for weights in the analysis can produce misleading point estimates (Stata Corporation, 1999). Pearson's chi-square was used to determine statistically significant differences observed within the various categories of the WHO (1992) hemoglobin grouping variable in relation to standard of living and location of residence, or the grouped BMI (kg/m2) variable representing thin, normal-weight and overweight women. Differences were considered statistically significant at P<0.05.

Stata Release 6 (1999) was used to fit the ordered logit models using maximum likelihood estimation. To account for the complex survey design, we included the state-level individual sampling weight, strata (urban/rural), and clustering variable (primary sampling unit), using the survey option to estimate the models in Stata. Accounting for weighting in the analysis allows a design-based point estimate to be obtained. In addition, taking account of the stratification and clustering of the data provides more robust estimates of the associated standard errors than an analysis that ignores the survey design characteristics (Stata Corporation, 1997). The beta estimates produced by the ordered logit model correspond to the log-odds ratio of being over vs under any particular level of the outcome variable used. The model assumes that the effect of any of the independent variables should be the same regardless of the choice of category (non anemic, mildly anemic, moderately or severely anemic) for the outcome variable being considered (McCullagh & Nelder, 1989). In order to test this assumption we conducted a test on model 4 in SAS version 7Footnote 5 (SAS Institute, 1997).

We retained only significant variables, with a two-tailed P-value <0.05, in any one of models 1–4 (Table 3).

Table 3 Determinants of anemia status for women in Andhra Pradesh from the NFHS 1998/1999 (odds ratios and 95% confidence intervals obtained from ordinal logistic regression models)a


The percentage of women observed in each of the anemia classification groups is presented by the location and standard of living variable, and the BMI variable in Table 2. These variables are reflective of the main hypotheses we test, ie that differences in anemia prevalence are related to social class, urban/rural location, and nutrition status (BMI). Percentages were adjusted using the sample weights to account for variations in the response rate in the different geographical regions of the state in NFHS 2. Some 49.5% of women were classified as anemic, 32.4% as mildly anemic, 14.9% as moderately anemic, and 2.2% as severely anemic.

Table 2 Weighted percentages of women with anemia by standard of living, location of residence and body mass index

We observed statistically significant differences in the prevalence of anemia between groups based upon socioeconomic status and location of residence, indicated by the chi-square statistic displayed in Table 2.

Although statistically significant differences were observed in the percentage of women classified as anemic, with the higher standard of living groups having reduced risk (P<0.01), the prevalence of anemia was high among all groups. Prevalence of mild anemia ranged from 40% in the urban high standard of living group to 62% in the urban low standard of living group. Similarly the lowest prevalence of moderate anemia was found in the urban high standard of living group (10%) and the highest in the urban low standard of living group (18%). For severe anemia the highest and lowest prevalences were observed in rural areas, with the rural high standard of living group exhibiting the lowest (1%), and the rural low standard of living group the highest (3%), prevalence.

Statistically significant differences in anemia classification by BMI are shown in Table 2 (P<0.01). Although overweight women had a lower prevalence of any degree of anemia compared to normal-weight and thin women, high prevalence was observed across all of the BMI groups. The percentage of women observed to be anemic was 52% for thin women, 50% for normal-weight women, and 41% for overweight women. Prevalence of moderate anemia ranged from 9% for the overweight women to 17% for thin women, and for severe anemia from 1% for the overweight women to 4% for the thin women.

The results of the ordered logit models are presented in Table 3. Results are presented as odds ratios with 95% confidence intervals. Women from the urban low standard of living group were observed to have the highest odds of being mildly, moderately or severely anemic (OR=1.76, 95% Cl=1.25, 2.46) when compared to the high urban standard of living group after controlling for socio-demographic, diet and health, and BMI variables in model 4. The only other group in the socioeconomic and location variable observed to be significantly more likely to have increased odds of being anemic in model 4 was the rural low sample (OR=1.34, 95% CI=10.3, 1.74). In the unadjusted model (model 1), the medium rural (OR=1.48, 95% CI=1.17, 1.87) and medium urban group (OR=1.43, 95% CI=1.11, 1.83) as well as the rural poor (OR=1.87, 95% CI=1.48, 2.36) and urban poor groups (OR=2.36, 95% CI=1.69, 3.30) had significantly higher odds of being anemic when compared to the urban high group.

Other socio-demographic factors where significant differences were observed in anemia status were religion and maternal education. In model 4, Muslim women were observed to be significantly less likely to be mildly, moderately or severely anemic than Hindu women (OR=0.65, 95% CI=0.49, 0.86). Women from religious groups other than Muslim or Hindu (OR=1.59, 95% CI=1.28, 1.98) were significantly more likely to be anemic than those in the Hindu group. Respondents who reported having received at least a high school education were significantly less likely to be mildly, moderately or severely anemic (OR=0.65, 95% CI=0.45, 0.94).

Of the health and diet variables tested for significance in the models, two were found to be significant: drinking alcohol and eating pulses at least daily. Drinking alcohol was the most significant predictor of any type of anemia. Women who drank alcohol were significantly less likely to be mildly, moderately, and severely anemic than their counterparts who did not drink alcohol after controlling for all of the other significant variables in the models (OR=0.53, 95% CI=0.41, 0.68). In model 4, respondents who reported eating pulses daily were significantly less likely to be mildly, moderately or severely anemic than those who reported eating pulses less than daily or not at all (OR=0.83, 95% CI=0.72, 0.96).

Respondents with a BMI less than 18.5 kg/m2 were observed to be marginally significantly more likely to be anemic (OR=1.14, 95% CI=1.00, 1.29) than those with a normal BMI (18.5–24.9 kg/m2). In contrast overweight respondents with a BMI ≥25 kg/m2 were observed to be significantly less likely to be anemic than those with a normal BMI (OR=0.76, 95% CI=0.62, 0.93). The BMI variable also reduced the significance of the location and standard of living index variable. Significant differences were reduced for the rural low group when compared to the urban high group after controlling for BMI.

We conducted bivariate analyses to better understand the characteristics of the rural and urban women with and without anemia. Rural poor and urban poor women are similar on a number of indicators: 33% of rural poor and 30% of urban poor women were from scheduled castes and less than 1% of both groups had achieved secondary school education. Daily pulse consumption, a proxy for socioeconomic status and a strong predictor of anemia, was reported by less than 25% of rural and urban poor women, compared to 36% among rural high, 43% among urban medium, and 52% among urban high income women. Occupational status also differed among the sample, with 81% of rural poor women and 50% of urban poor women reporting working outside of the home. This compares to 70 and 38% of rural medium- and high-income women, and only 28 and 12% of urban medium- and high-income women, reporting working outside of the home.


The results on the prevalence and determinants of anemia among Indian women should be interpreted in their economic and socio-cultural context. By nearly any measure, India remains one of the poorest countries in the world, with a population of over one billion and a fertility rate well above replacement level (World Bank, 2000). There have been impressive improvements in most health indicators in the last two decades, including a reduction in infant mortality rate from 115 in 1980 to 70 in 1998 and a drop in the fertility rate from 5 to 3.2 during the same period (World Bank, 2000). Improvements in nutritional status, however, have been less impressive. More than half of the world's undernourished population lives in India (Krishnaswami, 2000) and half of Indian children are malnourished (Measham & Chatterjee, 1999; Kumar, 2000). Apart from overall poverty, the health status of women in India reflects gender discrimination from birth (Miller, 1981; Murthi et al, 1995; Kishor, 1995), inequitable distribution of health resources (Arnold et al, 1996; Basu, 1995), and early and frequent reproductive cycling and reproductive tract infections (Koblinsky, 1995; Brabin et al, 1998; Bhatia & Cleland, 1995; Bang et al, 1989). The high rates of anemia among Indian women, therefore, reflect their social and biological vulnerability both within society and the household.

We conducted this analysis as a companion to our previous paper, which investigated the determinants of under and overweight among women living in Andhra Pradesh, India (Griffiths & Bentley, 2001). Following the earlier work, we hypothesized that disadvantaged and undernourished Indian women would be more likely to be anemic, reflecting health disparities that are on the rise because of increasing urbanization and improvements in economic development (Diwaker & Qureshi, 1992; Visaria, 1997; Shariff, 1999). We also hoped to identify risk factors that would be helpful for program purposes to prevent anemia among Indian women. What have we learned?

Anemia among women in this large, southern Indian state cuts across social class, place of residence, and other factors that normally discriminate health status. Rich or poor, fat or thin, urban or rural—the prevalence of anemia is high among women in all these groups and differences are only relative. More than 40% of women in the highest socioeconomic group are anemic, as are 62% of urban poor and 54% of rural poor women.

Our hypotheses related to socioeconomic status, urban/rural location and anemia were partially supported. We had expected to find the highest prevalence of anemia among rural women, who are also the poorest, based on the standard of living index. However, the poorest rural and urban women both had the greatest risk of anemia and had similar probabilities of being anemic, with the exception of the poorest urban women who were more likely than poor rural women to be anemic. This supports the findings of other studies in the 1980s and 1990s which revealed a great diversity in the extent and depth of poverty within the urban sector in developing countries and poor health outcomes for the most marginalized urban groups (Harpham et al, 1988; Harpham, 1997; Rossi-Espagnet, 1984; Satherthwaite, 1993; Tazibzadeh et al, 1989; United Nations, 1998). Rates of urban growth are most intense in developing countries, often resulting in poor housing, overcrowding, pollution and increased exposure to infectious diseases (Harpham et al, 1988). The direct effects of poverty that result in low income, limited education and insufficient diet have all been associated with poor health outcomes for the urban poor in developing countries.

Despite greater opportunities for health care in urban areas, the urban poor are often more marginalized than rural populations in their ability to access health services because of constraints in financial and administrative resources that are necessary to access the services in urban areas (Yesudian, 1988; Kakar, 1988; Griffiths & Stephenson, 2001). Likewise, although urban areas theoretically have greater access to a wide variety of food and nutrients through close access to markets, extreme poverty limits the ability of the urban poor to purchase them.

One reason that urban poor women may have higher risk of anemia than rural poor women is their lack of access to their own income or resources because of lower rates of extra-household employment and reduced economic power within the household (Basu, 1995; Bennett, 1991; Sen, 1991; Banerjee, 1995). The urban poor may also experience higher rates of infection related to poor sanitation or high rates of reproductive tract infections, gynecological morbidity, or sexually transmitted diseases (Bhatia & Cleland, 1995; Brabin et al, 1998). Dimensions of autonomy such as freedom of movement, decision-making power and control over finances can also exert a strong influence over service use and service choice in the South Asian setting (Bloom et al, 2001). This results in inappropriate treatment of illnesses.

The apparent protective effect of alcohol on anemia is an interesting finding and we explored this in fuller detail through bivariate analyses. Poorer, rural women and rural women from scheduled tribes (also the poorest) were more likely to consume alcohol, compared to all the other groups. These women are also the thinnest and have higher rates of anemia compared to other women, with the exception of the urban poor. Of scheduled tribe women who do not drink alcohol, 52% are not anemic, 32% are mildly anemic, and 14 and 2% are moderately and severely anemic. Of women from this group who do consume alcohol, the prevalence of all types of anemia is lower, particularly in the most serious classification: 54% are not anemic, 41% are mildly anemic, 5% are moderately anemic, and 0% are severely anemic. Alcohol consumption, therefore, does appear protective against anemia among this vulnerable group. There is an abundant literature on increased iron status and absorption related to alcohol consumption (Turnbull, 1974; Hallberg & Hulthen, 2000; Millman & Kirchhoff, 1996). However, we do not know what it is about alcohol consumption among these women that reduces their risk of anemia (eg fermentation process, iron vessels, alcohol stimulation of gastric acid secretion, promoting solubility and reduction of ferric iron; iron present in the alcohol; enhanced iron absorption, or an unmeasured variable that changes anemia status not related to alcohol consumption).

Our hypothesis on the relationship of BMI to anemia was partially supported. Thin women (BMI <18.5 kg/m2) were marginally significantly more likely to be anemic compared to women of normal-weight. Although being classified at normal or overweight is somewhat protective, more than 10% (see Table 2) of women with a high BMI (≥25 kg/m2) are moderately or severely anemic, suggesting dietary deficiency or other problems among women who have no apparent resource constraints (Griffiths & Bentley, 2001). Other problems may include hookworm or malaria infection, acute infections, micronutrient deficiency that interferes with iron metabolism, or poor dietary patterns that compromise adequate iron intake. Among the urban, overweight and higher-income women with moderate to severe anemia, we suspect that diet may play an important role but we are unable to assess its role with the current data.

These results on anemia in women are both similar and different to our previous work related to the emerging nutrition transition in India (Griffiths & Bentley, 2001). In this analysis, we find similar relationships between socioeconomic status and anemia to those found between socioeconomic status and BMI. Poorer women are more likely to be anemic and to be underweight, while better off women are less likely to be anemic and more likely to be overweight. However, the urban/rural patterns for anemia differ from the results we found between urban/rural residence and BMI. Our previous analysis showed a strong association between urban/rural location and BMI, with urban women more likely to be overweight or obese and rural women more likely to be thin or underweight. While the majority of overweight and obese women live in urban areas, socioeconomic status was more important than urban/rural residence in predicting whether women were fat or thin. For the anemia analysis, we found that poor, urban women have the highest rates and risk of anemia compared to the other groups, including poor rural women. The results reflect the effect of poverty on women's nutrition and anemia status, regardless of whether they live in the rural or urban areas.

What have we learned that is helpful for program and policy makers in India? The ‘bad news’ is that many of the risk or protective factors are not amenable to change or rapid intervention. Protective factors include being Muslim, an urban or rural woman of middle or high class, having graduated from high school, consuming alcohol or pulses, or being overweight. Promoting alcohol consumption is not an appropriate public health intervention, particularly in this setting. With the exception of the protective factor of daily pulse consumption, none of the other dietary variables explain any of the anemia outcomes in our multivariate analysis. We conducted a cross-tabulation of daily pulse consumption with socioeconomic status and found a strong correlation. Pulse consumption is likely to be a proxy for higher income. Pulses also have a high iron content and could be directly providing protection against anemia. The lack of relationship between the other dietary factors and anemia, however, does not imply that a poor-quality diet is not part of the problem, nor that interventions to improve dietary quality, quantity or iron bioavailability are unwarranted. This is because our results are greatly limited by the type of dietary data available in the NFHS data and we are not adequately able to explore how dietary patterns or nutrient intake may influence anemia status among the women in the sample.

The finding of the highest risk of anemia among very poor urban women should focus attention on this group for intervention purposes. While most health indicators in India do show that rural women are disadvantaged relative to urban women (International Institute of Population Science and ORC Macro, 2000), our findings for anemia suggest a more complex scenario that is controlled not only by location of residence but also socioeconomic status. This emphasizes the need for within-group analyses.

We believe this analysis strengthens the recommendation of Stoltzfus (1997) for a re-examination of policy on the international cut-offs to assess anemia prevalence for purposes of surveillance and treatment. She argues that prevalence data should distinguish between ‘any’ anemia (mild, moderate and severe combined) to report prevalence in all three degrees. Although a policy change has not yet occurred, one important example of a change in practice is that the new global burden of disease estimates for anemia will consider hemoglobin as a continuous variable rather than a yes/no variable for mild anemia, as had been done previously (Stoltzfus, personal communication). In this sample, the prevalence of ‘any’ anemia ranges from 40 to 62% across all socioeconomic and residence groups. The high prevalence of ‘any’ anemia in this study (49.5%) is similar to that in other developing countries and makes it difficult for governments to target women at risk or to allocate sufficient resources for prevention or treatment (Stoltzfus, 1997). While the majority of these anemic Indian women have mild anemia, 17% are moderately or severely anemic. This compares to 13% of non-pregnant women in Zanzibar, 21% in Nepal, and 3% in Java, Indonesia (Stoltzfus, 1997). In India, this represents millions of women who are walking around with symptoms and risks associated with a serious degree of anemia or anemia related to other causes (eg malaria, hookworm or subclinical infections, or other factors). The NFHS 2 in India allows an analysis by degree of anemia and this provides a benchmark for program evaluation, for examining changes across regions or sub-groups over time, and for mobilizing programs to target interventions to women at most risk.

We believe that improving women's overall nutrition status and their access to resources (income) will have the greatest impact on reducing anemia in India (World Bank, 1993). Very thin and very poor women, particularly in urban areas, have the highest risk of moderate and severe anemia. In the short-term, combined food and iron supplementation programs would be most effective to address both anemia and underweight. The alarming 10% prevalence of moderate and severe anemia among obese women is cause for concern and we need to learn more about the determinants of this outcome among this subgroup of women for prevention purposes. Integrated programs for hookworn eradication, malaria prophylaxis, or that address other micronutrient deficiencies would also be important for reducing the burden of anemia (Stoltzfus, 1997; Gillespie & Johnston, 1998). The results presented here and in our earlier analysis clearly show that making improvements in household socioeconomic status and maternal education will affect maternal health and nutrition in a sustained way.

In conclusion, the high prevalence of anemia among women in India is a burden for them, for their families, and for the economic development and productivity of the country. Iron supplementation programs, for a variety of reasons, have not been effective in reducing anemia prevalence (Galloway & McGuire, 1991; Vijayaraghaven et al, 1990) and operational research on how best to improve existing iron supplementation programs is needed (Yip, 1994). New and innovative strategies are needed, particularly those that improve the overall health and nutrition status of adolescent girls before they enter their reproductive years (Gillespie & Johnston, 1998; Kurz & Johnson-Welch, 1994; Kanani & Poorjara, 2000; Creed-Kanashiro et al, 2000). This will require tailored programs that target women in all socioeconomic groups and who live within both rural and urban areas, but particularly in need of intervention are the urban poor, who are a rapidly growing marginalized segment of the Indian population.


  1. 1.

    The International Institute for Population Sciences based in Mumbai, India coordinated the data collection with technical assistance from MEASURE DHS+at ORC Macro, Maryland, USA and the East–West Center, Hawaii, USA. Researchers can apply for permission to analyze the data through MEASURE DHS+, who make data collected in all of the DHS surveys available for analysis through their website; The survey was approved by the institutional review board at ORC, Macro, and the entire questionnaire and all of the procedures were approved by a multi-agency technical advisory committee in India, which considered human subject protections and ethical issues. Informed consent was obtained from participants both to take part in the survey and a separate, more detailed consent was obtained for hemoglobin and lead measures. The main objectives of the survey were to provide estimates of fertility, family planning practices, infant and child mortality, maternal and child health and nutrition, the utilization of maternal and child health services, the quality of these services, the status of women, women's reproductive health problems and domestic violence.

  2. 2.

    The data from NFHS 2 are being made available state by state by ORC Macro between 2000 and 2002.

  3. 3.

    All women were given the results of the hemoglobin test and had them explained to them. In addition, women with severe anemia (Hb <70 g/l) were read a statement asking whether they would give permission for the health investigator to inform a local health official about the problem.

  4. 4.

    The standard of living index is a composite index calculated by ORC Macro and the International Institute of Population Sciences and is based upon household ownership of possessions/consumer durables and land/livestock (NFHS, 1998/1999). The index was divided into low, medium and high groups based upon scores obtained. The high group contains those with a score of 25–66, the medium group scores between 15–24, and the low group 0–14.

  5. 5.

    STATA does not have a procedure for checking the proportional odds assumption. This resulted in us having to utilize SAS to test this assumption, but we were unable to test the model, which controls for survey design as SAS does not have the facility to control for survey design effects. However, by testing model 4 without survey design effects we were able to ascertain whether the proportional odds assumption was met. The results presented in Table 3 are from the models estimated in STATA which account for the survey design effects.


  1. Allen, L (1997). Pregnancy and iron deficiency: unresolved issues. Nutr. Rev., 55, 4

    Google Scholar 

  2. Amin, A (2000). Gynecological morbidity and decision-making among rural Maharashtra women. Doctoral DissertationJohns Hopkins University, Baltimore, Maryland

  3. Arnold, F, Choe, MK & Roy, TK (1996). Son preference, the family-building process and child mortality in India. Presented at the Annual Meetings of the Population Association of America, 9–11, May, New Orleans

  4. Banerjee, N (1995). Household dynamics and women in a changing economy. In: ed. M Krishnaraj, R Sudarshan & A SharriffGender, Population, and Development, New Delhi: Oxford University Press

    Google Scholar 

  5. Bang, RA, Bang, AT, Baitule, Y, Choudhary, S, Sarmukkadam, S & Tale, O (1989). High prevalence of gynecological diseases in rural Indian women. Lancet, 1, (8629) 85–88.

    CAS  Article  Google Scholar 

  6. Basu, AM (1995). Women's roles and the gender gap in health and survival. In: ed. M Gupta, LC Chen & TN KrishnanWomen's health in India, pp153–174, Bombay: Oxford University Press

    Google Scholar 

  7. Bennett, L (1991). Gender and Poverty in India, Washington, DC: World Bank

    Google Scholar 

  8. Bentley, M & Parekh, A (1998). Perceptions of anemia and health seeking behavior among women in four Indian states, Technical Working Paper no. 9Washington, DC: Mothercare/John Snow Inc

    Google Scholar 

  9. Bhatia, JC & Cleland, J (1995). Self-reported symptoms of gynecological morbidity and their treatment in South India. Stud. Family Planning, 26, 203–216.

    CAS  Article  Google Scholar 

  10. Bloom, SS, Wypij, D & Das Gupta, M (2001). Dimensions of women's Autonomy and the influence of maternal health care utilization in a northern Indian city. Demography, 38, 67–78.

    CAS  Article  Google Scholar 

  11. Brabin, L, Nicholas, S, Gogate, K & Karande, A (1998). High prevalence of anemia among women in Mumbai. India Food Nutr. Bull., 19, 3 205–222.

    Article  Google Scholar 

  12. Chakma, T, Rao, PV & Tiwary, RS (2000). Prevalence of anemia and worm infestation in tribal areas of Madhya Pradesh. J. Indian Med. Assoc., 98, 567–561.

    CAS  PubMed  Google Scholar 

  13. Creed-Kanashiro, HM, Uribe, TG, Bartolini, RM, Fukumoto, MN, Lopez, TT, Zavaleta, NM & Bentley, ME (2000). Improving dietary intake to prevent anemia in adolescent girls through community kitchens in a periurban population of Lima, Peru. J. Nutr., 130, 459S–261S.

    CAS  Article  Google Scholar 

  14. Diwaker, A & Qureshi, MH (1992). Demographic processes of urbanization in Delhi. Population Geog., 14, 27–36.

    Google Scholar 

  15. Galloway, R & McGuire, J (1994). Determinants of compliance with iron supplementation: supplies, side effect, or psychology?. Soc. Sci. Med., 39, 381–390.

    CAS  Article  Google Scholar 

  16. Gillespie, S (1998). Major Issues in the Control of Iron Deficiency, The Micronutrient Initiative/UNICEF

    Google Scholar 

  17. Gillespie, S & Johnston, J (1998). Expert Consultation on Anemia Determinants and Interventions, Ottawa: The Micronutrient Initiative

    Google Scholar 

  18. Griffiths, PL & Bentley, ME (2001). The nutrition transition is underway in India. J. Nutr., 131, 2692–2700.

    CAS  Article  Google Scholar 

  19. Griffiths, PL & Stephenson, R (2001). Understanding users perspectives of barriers to maternal health care use in Maharashtra, India. J. Biosoc. Sci., 33, 339–359.

    CAS  Article  Google Scholar 

  20. Hallberg, L & Hulthen, L (2000). Prediction of dietary iron absorption: an algorithm for calculating absorption and bioavailability of dietary iron. Am. J. Clin. Nutr., 71, 1147–1160.

    CAS  Article  Google Scholar 

  21. Harpham, T (1997). Urbanisation and health in transition. Lancet, 39, SIII 11–SIII 13.

    Article  Google Scholar 

  22. Harpham, T, Lusty, T & Vaughan, P (1988). In the Shadow of the City: Community Health and the Urban Poor, Oxford: Oxford University Press

    Google Scholar 

  23. International Institute for Population Sciences & ORC Macro (2000). National Family Health Survey India 1998/1999, Mumbai: Andhra Pradesh International Institute for Population Sciences

  24. Kakar, DN (1988). Primary Health Care and Traditional Medical Practitioners, New Delhi: Sterling Publishers

    Google Scholar 

  25. Kanani, S, Latha, K & Shah, M (1994). Application of qualitative methodologies to investigate perceptions of women and health practitioners regarding women's health disorders in Baroda slums. In: Gittelsohn J, Bentley ME, Pelto PJ, Nag M, Pachauri S, Harrison AD, & Landman LTListening to Women Talk about their Health Issues and Evidence from India, pp116–130, New Dehli: The Ford Foundation, Har-Anand Publications

    Google Scholar 

  26. Kanani, SJ & Poojara, RH (2000). Supplementation with iron and folic acid enhances growth in adolescent Indian girls. J. Nutr., 130, 452S–455S.

    CAS  Article  Google Scholar 

  27. Kapil, U, Pathak, P, Tandon, M, Singh, C, Pradhanb, R & Dwivedi, SN (1999). Micronutrient deficiency disorders amongst pregnant women in three urban slum communities of Delhi. Indian Pediat., 36, 983–989.

    CAS  Google Scholar 

  28. Kielmann, K (2000). Perceived morbidity and the dynamics of ‘women's health’ in low-income areas of Maharashtra, India. Doctoral DissertationJohns Hopkins University, Baltimore, MD

  29. Kishor, S (1995). Gender differentials in child mortality: a review of the evidence. In: ed. M Das Gupta, LC Chen & TN KrishnanWomen's Health in India: Risk and Vulnerability, pp19–54, Bombay: Oxford University Press

    Google Scholar 

  30. Koblinsky, MA (1995). Beyond maternal mortality—magnitude, interrelationship and consequences of women's health, pregnancy related complications and nutritional status on pregnancy outcomes. Int. J. Gynecol. Obstetr., 48, s21–s32.

    Article  Google Scholar 

  31. Krishnaswami, K (2000). Country profile: India. nutritional disorders—old and changing. Lancet, 351, 1268–1269.

    Google Scholar 

  32. Kumar, S (2000). India health survey finds too many women and children in poor health. Lancet, 356, 1830–1830.

    CAS  Article  Google Scholar 

  33. Kurz, K & Johnson-Welch, C (1994). The Nutrition and Lives of Adolescents in Developing Countries: Findings from the Nutrition of Adolescent Girls Research Program, Washington, DC: International Center for Research on Women

    Google Scholar 

  34. McCullagh, P & Nelder, JA (1989). Generalized Linear Models, 2nd ednLondon: Chapman and Hall

    Google Scholar 

  35. Measham, AR & Chatterjee, M (1999). Wasting Away: the Crisis of Malnutrition in India, Washington, DC: The World Bank

    Google Scholar 

  36. Miller, BD (1981). The endangered sex: neglect of female children in rural north India. Cornell University Press, Ithaca N.Y.

  37. Millman, N & Kirchhoff, M (1996). Relationship between serum ferritin, alcohol intake, and social status in 2235 Danish men and women. Ann. Hematol., 72, 145–151.

    Article  Google Scholar 

  38. National Nutrition Monitoring Bureau (1989). Report of repeat survey phase I, Table 6Hyderabad: National Institute of Nutrition, Indian Council of Medical Research

  39. Patel, P (1994). Illness beliefs and health-seeking behaviour of the Bhil women of Panchamahal District, Gujarat State. In:Listening to Women Talk about their Health Issues and Evidence from India, eds. Gittelsohn, Bentley et alDelhi Har-Anand

    Google Scholar 

  40. Rajaratnam, J, Abel, R, Asokan, JS & Jonathan, P (2000). Prevalence of anemia among adolescent girls of rural Tamil Nadu. Indian Pediat, 27, 532–536.

    Google Scholar 

  41. Rossi-Espagnet, A (1984). Primary health care in urban areas: reaching the urban poor in developing countries, A state of the art report by UNICEF and WHO. Report no. 2499MGeneva: WHO

    Google Scholar 

  42. SAS Institute (1997). SAS/STAT Software: changes and Enhancements through Release 6.12, Cary, NC: SAS Institute

  43. Satherthwaite, D (1993). The impact on health of urban environments. Environ. Urbanis., 15, 87–111.

    Article  Google Scholar 

  44. Sen, AK (1991). Gender and co-operative conflicts. In:Persistent Inequalities: Women and World Development, ed. I TinkerNew Delhi: Oxford University Press

    Google Scholar 

  45. Shariff, A (1999). India Human Development Report, pp69–71, New York: Oxford University Press

    Google Scholar 

  46. Sharma, A, Prasad, K & Rao, KV (1996). Identification of an appropriate strategy to control anemia in adolescent girls of poor communities. Indian Pediatr., 37, 261–267.

    Google Scholar 

  47. Stata Corporation (1999). Stata User's Guide: Release 6, College Station, TX: Stata Press. Texas

  48. Stoltzfus, RJ (1997). Rethinking anemia surveillance. Lancet, 349, 1764–1766.

    CAS  Article  Google Scholar 

  49. Tazibzadeh, I, Rossi-Espagnet, A & Maxwell, R (1989). Spotlight on the Cities: Improving Urban Health in Developing Countries, Geneva: WHO

    Google Scholar 

  50. Turnbull, A (1974). Iron absorption. In:Iron in Biochemistry and Medicine, edn. A Jacobs M Worwood pp369–403, London: Academic Press

    Google Scholar 

  51. United Nations (1998). World Urbanization Prospects: The 1996 Revisions, New York: UN

  52. Vijayaraghavan, K, Brahmam, GNV, Nair, KM, Akbar, D & Praalhad Rao, N (1990). Evaluation of National Nutritional Anemia Prophylaxis Programme. Ind. J. Pediatr., 57, 183–190.

    CAS  Article  Google Scholar 

  53. Visaria, P (1997). Urbanization in India: an overview. In:Urbanization in Large Developing Countries, ed. GW Jones & Visaria pp259–288, Oxford: Clarendon Press

    Google Scholar 

  54. World Bank (1993). Invest in Health, World Development Report. pp195–324, Oxford: Oxford University Press

  55. World Bank (2000). World Development Report 2000/2001: Attacking Poverty, New York, Oxford University Press

  56. World Health Organization (1992). The Prevalence of Anaemia in Women: a Tabulation of Available Indormation, WHO, Geneva Switzerland

  57. Yesudian, CAK (1988). Health Service Utilization in Urban India, Delhi: Mittal

    Google Scholar 

  58. Yip, R (1994). Iron deficiency: contemporary scientific issues and international programmatic approaches. J. Nutr., 124, (S8) S1479–S1490.

    Article  Google Scholar 

Download references


The authors would like to acknowledge the individuals who contributed to the collection of the Andhra Pradesh National Family Health Survey 1998/1999; Sumati Kulkarni, Fred Arnold, TK Roy, Arvind Pandey, Robert Retherford, kamla Gupta, M Vivekananda Murty, Sunita Kishor, Vinod Mishra, Sushil Kumar, Zaheer Ahmad Khan, and Sidney, B Wesley. We thank Dr Usha Ramakrishna, Emory University, for reviewing an earlier draft of this paper and for providing analytical suggestions that greatly improved the presentation of results. Dr Anna Maria Siega-Riz of the University of North Carolina, Chapel Hill, also provided analytical guidance and support.

Author information



Corresponding author

Correspondence to M E Bentley.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Bentley, M., Griffiths, P. The burden of anemia among women in India. Eur J Clin Nutr 57, 52–60 (2003).

Download citation


  • India
  • women
  • iron deficiency
  • anemia
  • nutrition
  • urban/rural
  • socioeconomic status
  • underweight/overweight

Further reading